.github/workflows/code-runner-precommit.yml

@@ -0,0 +1,73 @@
+name: code-runner pre-commit
+
+on:
+  push:
+    branches: [main]
+    # The workflow's own file counts as code_runner: editing only this file would
+    # otherwise trigger nothing, so a fix to the job itself could never prove
+    # itself.
+    paths: ['code_runner/**', '.github/workflows/code-runner-precommit.yml']
+  pull_request:
+    branches: [main]
+    paths: ['code_runner/**', '.github/workflows/code-runner-precommit.yml']
+
+jobs:
+  pre-commit:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.x"
+      # Only pre-commit itself. Every linting hook runs in its own isolated
+      # env -- ruff, mypy, pylint, bandit and codespell each come from their
+      # own repo+rev, and the local hooks need nothing beyond bash and
+      # python3 -- so installing the full requirements.txt spent 9s a run on
+      # packages no pre-commit-stage hook imports. Verified by running all
+      # three configs green in a venv holding pre-commit and nothing else.
+      # The pin is read out of requirements.txt rather than repeated here, so
+      # it cannot drift from what the dependency-freshness gate governs.
+      - name: Install pre-commit
+        run: |
+          pip install "$(grep '^pre-commit==' code_runner/requirements.txt)"
+
+      # Keyed on THIS package's config. pre-commit/action@v3.0.1 caches on
+      # hashFiles('.pre-commit-config.yaml') -- the repo ROOT's -- so all three
+      # package jobs shared one key, raced to overwrite it with hook sets built
+      # from four different configs, and editing a per-package config did not
+      # invalidate it. The cache was therefore silently wrong rather than
+      # missing, and the hooks reinstalled from scratch on every run (38-49s).
+      - uses: actions/cache@v4
+        with:
+          path: ~/.cache/pre-commit
+          key: >-
+            pre-commit-code-runner-${{ env.pythonLocation }}-${{
+            hashFiles('code_runner/.pre-commit-config.yaml') }}
+          # Prefix fallback, or this fix would evaporate on the very
+          # event that motivated it: bumping a hook rev changes the
+          # exact key, and a total miss means reinstalling every hook
+          # again. pre-commit's store is keyed internally by hook
+          # repo+rev, so a superseded entry is ignored disk rather
+          # than a wrong hook, and cache still re-saves under the new
+          # exact key after a prefix-only hit.
+          restore-keys: pre-commit-code-runner-${{ env.pythonLocation }}-
+      # mypy re-analyses the stdlib and every third-party stub from scratch
+      # without this, which measured 10s of the hook step in CI against 0.25s
+      # warm locally -- a 19MB cache dir does the whole difference. Keyed on
+      # the config, not the sources: mypy's own cache metadata records the
+      # interpreter, the flag set and a hash per source file, so it re-checks
+      # what changed and cannot pass stale code. Deliberately NOT keyed on
+      # github.sha -- that re-saves 19MB x3 every commit and would evict the
+      # 68MB hook caches under the repo's 10GB LRU budget. A fixed key keeps
+      # the expensive part, the dependency analysis, which is what costs 10s.
+      - uses: actions/cache@v4
+        with:
+          path: .mypy_cache
+          key: >-
+            mypy-code-runner-${{ env.pythonLocation }}-${{
+            hashFiles('code_runner/.pre-commit-config.yaml') }}
+          restore-keys: mypy-code-runner-${{ env.pythonLocation }}-
+      - name: Run the hooks
+        run: >-
+          pre-commit run --config code_runner/.pre-commit-config.yaml
+          --all-files --show-diff-on-failure

.github/workflows/code-runner-tests.yml

@@ -0,0 +1,34 @@
+name: code-runner tests
+
+on:
+  push:
+    branches: [main]
+    # The workflow's own file counts as code_runner: editing only this file would
+    # otherwise trigger nothing, so a fix to the job itself could never prove
+    # itself.
+    paths: ['code_runner/**', '.github/workflows/code-runner-tests.yml']
+  pull_request:
+    branches: [main]
+    paths: ['code_runner/**', '.github/workflows/code-runner-tests.yml']
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: code_runner
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.x"
+          cache: pip
+          cache-dependency-path: code_runner/requirements.txt
+
+      - name: Install dependencies
+        run: pip install -r requirements.txt && pip install --no-deps -e .
+
+      - name: Run tests with coverage
+        run: python -m pytest -v

.pre-commit-config.yaml

@@ -13,7 +13,7 @@
 # linters, and carry this same hook so committing with one of those configs is
 # gated too. This root config is what `pre-commit install` wires up, and it is
 # the only one that covers the subprojects without a config of their own
-# (guard-lib/, crdt_sync_dart/, file_length/, ...).
+# (aur-publish/, guard-lib/, crdt_sync_dart/, file_length/, ...).
 #
 # The path is repo-relative on purpose. pre-commit always runs hooks from the
 # git root, so this resolves the same locally and on a CI runner -- an absolute

CLAUDE.md

@@ -16,4 +16,3 @@ overwrite it with `repo_contract/templates/check_repo_contract.shim.sh`.
 - lint: `ruff check agent_ready coverage_gaps flutter_web_drive repo_contract`
 - coverage: `PYTHONPATH=. python -m pytest -q agent_ready/tests flutter_web_drive/tests repo_contract/tests unified-design-system/tests --cov=agent_ready --cov=flutter_web_drive --cov=repo_contract --cov=conformance --cov=conformance_diff --cov-branch --cov-report=xml --cov-fail-under=100`
 - coverage-gaps: `coverage-gaps coverage.xml`
-- deploy: `git -C ~/src/utils pull --ff-only && ~/src/utils/scripts/deploy.sh` (installs the sickdays CLI + launcher from the live checkout, then exits 1 unless every released library's version tag is on origin and matches HEAD and the live `sickdays` runs the text-free checkout; `--verify-only` for the health check alone)

aur-publish/README.md

@@ -0,0 +1,45 @@
+# aur-publish
+
+Hands-off publishing of kuhy's Flutter apps to the AUR.
+
+```bash
+~/src/utils/aur-publish/publish.sh              # wait for registration, then publish
+~/src/utils/aur-publish/publish.sh --dry-run    # everything except the push
+~/src/utils/aur-publish/publish.sh --skip-wait  # account already works
+~/src/utils/aur-publish/publish.sh --only habit-stack
+```
+
+## What it does
+
+1. Polls `/register` until AUR signup reopens, opens it in a browser and
+   prints the SSH key to paste. **Signup is the only manual step** — it
+   requires email confirmation.
+2. Waits for `ssh aur@aur.archlinux.org` to authenticate.
+3. Then, unattended, per package: resolve the newest `vX.Y.Z` tag → set
+   `pkgver` → `updpkgsums` → `makepkg -C` → `namcap` → AUR rule check →
+   `pacman -U` → launch check → `.SRCINFO` → commit → push.
+
+Every package is built, installed and **launched** before it is pushed, so a
+broken package cannot reach the AUR. A package failing any gate is skipped and
+reported; the others still publish. Safe to re-run.
+
+## anubis_gate.py
+
+The AUR sits behind [Anubis](https://anubis.techaro.lol/), which answers HTML
+requests with a proof-of-work challenge page carrying **HTTP 200**. A plain
+`curl` therefore reports 200 for a page that is actually unreachable — which is
+exactly how "is registration back?" gets answered wrong. This module solves the
+challenge and reports the status of the *real* page behind it.
+
+```bash
+python3 anubis_gate.py /register     # prints the true status code
+```
+
+## AUR rules enforced before pushing
+
+The AUR does not review uploads, but per the submission guidelines "packages
+that violate the rules may be deleted without warning". `check_aur_rules()`
+blocks a push on the mechanically checkable ones: `# Maintainer:` line,
+`x86_64` in `arch`, no `replaces=` (reserved for renames — use `conflicts`),
+not already in an official repo, `pkgdesc`/`url`/`license` present, and
+`.SRCINFO` matching the PKGBUILD.

aur-publish/anubis_gate.py

@@ -0,0 +1,138 @@
+"""Fetch an AUR URL from behind the Anubis anti-bot gate.
+
+The AUR sits behind Anubis (https://anubis.techaro.lol/), which answers every
+HTML request with a proof-of-work challenge page carrying HTTP **200**. A bare
+`curl` therefore reports 200 for a page that is actually unreachable, which is
+exactly how "is /register back up?" gets answered wrong.
+
+This module solves the challenge and reports the status of the *real* page
+behind it, so the caller can distinguish:
+
+  * 200  -> genuinely available
+  * 503  -> reached the site; that endpoint is disabled server-side
+  * other-> whatever the app actually said
+
+Exit codes mirror the HTTP status so shell callers can branch on them:
+0 for 200, 1 for anything else, 2 for a local/transport failure.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import http.cookiejar
+import json
+import re
+import sys
+import time
+import urllib.error
+import urllib.parse
+import urllib.request
+
+BASE = "https://aur.archlinux.org"
+UA = "Mozilla/5.0 (X11; Linux x86_64; rv:128.0) Gecko/20100101 Firefox/128.0"
+CHALLENGE_RE = re.compile(
+    r'<script id="anubis_challenge" type="application/json">(.*?)</script>',
+    re.S,
+)
+PASS_PATH = "/.within.website/x/cmd/anubis/api/pass-challenge"
+# A solve at difficulty 4 takes ~0.01s; this ceiling only guards a server-side
+# difficulty spike, so the poller degrades instead of spinning forever.
+MAX_NONCE = 50_000_000
+
+
+class _NoRedirect(urllib.request.HTTPRedirectHandler):
+    """Keep redirects visible instead of silently following them."""
+
+    def redirect_request(self, *_args, **_kwargs):  # noqa: D102 - see class doc
+        return None
+
+
+def _opener() -> urllib.request.OpenerDirector:
+    jar = http.cookiejar.CookieJar()
+    opener = urllib.request.build_opener(
+        urllib.request.HTTPCookieProcessor(jar), _NoRedirect
+    )
+    opener.addheaders = [
+        ("User-Agent", UA),
+        ("Accept", "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8"),
+    ]
+    return opener
+
+
+def _solve(random_data: str, difficulty: int) -> tuple[int, str, float]:
+    """Find a nonce whose sha256 digest starts with `difficulty` zero hex chars."""
+    prefix = "0" * difficulty
+    started = time.time()
+    for nonce in range(MAX_NONCE):
+        digest = hashlib.sha256(f"{random_data}{nonce}".encode()).hexdigest()
+        if digest.startswith(prefix):
+            return nonce, digest, time.time() - started
+    raise RuntimeError(f"no nonce found below {MAX_NONCE} at difficulty {difficulty}")
+
+
+def fetch(path: str, timeout: int = 30) -> tuple[int, str]:
+    """Return (status, body) for `path`, clearing Anubis first if challenged."""
+    opener = _opener()
+    url = f"{BASE}{path}"
+
+    try:
+        with opener.open(url, timeout=timeout) as resp:
+            body = resp.read().decode("utf-8", "replace")
+            status = resp.status
+    except urllib.error.HTTPError as exc:
+        return exc.code, exc.read().decode("utf-8", "replace")
+
+    match = CHALLENGE_RE.search(body)
+    if not match:
+        return status, body
+
+    challenge = json.loads(match.group(1))["challenge"]
+    nonce, digest, elapsed = _solve(
+        challenge["randomData"], int(challenge["difficulty"])
+    )
+    query = urllib.parse.urlencode(
+        {
+            "id": challenge["id"],
+            "response": digest,
+            "nonce": str(nonce),
+            "redir": path,
+            "elapsedTime": str(int(elapsed * 1000)),
+        }
+    )
+    try:
+        opener.open(f"{BASE}{PASS_PATH}?{query}", timeout=timeout)
+    except urllib.error.HTTPError:
+        # A redirect or non-2xx here is fine; what matters is the cookie the
+        # opener's jar just captured.
+        pass
+
+    try:
+        with opener.open(url, timeout=timeout) as resp:
+            return resp.status, resp.read().decode("utf-8", "replace")
+    except urllib.error.HTTPError as exc:
+        return exc.code, exc.read().decode("utf-8", "replace")
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("path", help="AUR path, e.g. /register")
+    parser.add_argument(
+        "--print-body", action="store_true", help="write the page body to stdout"
+    )
+    args = parser.parse_args()
+
+    try:
+        status, body = fetch(args.path)
+    except (urllib.error.URLError, RuntimeError, TimeoutError) as exc:
+        print(f"transport error: {exc}", file=sys.stderr)
+        return 2
+
+    print(status)
+    if args.print_body:
+        sys.stdout.write(body)
+    return 0 if status == 200 else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())

aur-publish/lib/aur_publish_one.sh

@@ -0,0 +1,107 @@
+#!/bin/bash
+
+# ============================================================================
+# Building, verifying and pushing one package to the AUR.
+#
+# Sourced by publish.sh, never executed: these functions read the readonly
+# config (AUR_HOST, AUR_ROOT, DRY_RUN, PUBLISHED, FAILED) that the entry script declares, and
+# a `readonly` re-run inside a twice-sourced lib would abort under `set -e`.
+# ============================================================================
+
+set -euo pipefail
+
+publish_one() {
+    local pkg="$1" repo="$2" wrapper="$3" binary="$4" expected="$5"
+    local dir="$AUR_ROOT/$pkg"
+    local version
+
+    log "$pkg"
+
+    [[ -d "$repo" ]] || { warn "$pkg: no repo at $repo"; return 1; }
+    [[ -f "$dir/PKGBUILD" ]] || { warn "$pkg: no PKGBUILD at $dir"; return 1; }
+
+    version="$(newest_tag "$repo")"
+    [[ -n "$version" ]] || { warn "$pkg: no vX.Y.Z tag on the remote"; return 1; }
+    info "newest release tag: v$version"
+
+    cd "$dir"
+    sed -i -E "s/^pkgver=.*/pkgver=$version/" PKGBUILD
+    # The tarball for a tag contains the PKGBUILD, so the checksum can only be
+    # computed after the tag exists — never committed ahead of time.
+    updpkgsums >/dev/null 2>&1 || { warn "$pkg: updpkgsums failed (tag missing?)"; return 1; }
+
+    info "building"
+    if ! env PATH=/usr/bin:/bin:/usr/local/bin makepkg -Cf --noconfirm >/dev/null 2>&1; then
+        warn "$pkg: makepkg failed"
+        return 1
+    fi
+
+    local built
+    built="$(find . -maxdepth 1 -name "$pkg-$version-*.pkg.tar.zst" | head -1)"
+    [[ -n "$built" ]] || { warn "$pkg: no package produced"; return 1; }
+
+    # namcap's "ELF files outside of a valid path ('opt/')" is a false positive
+    # for /opt-installed apps, so only genuinely new errors are worth failing on.
+    local findings
+    findings="$(namcap "$built" 2>&1 | grep -E ' E: ' | grep -v "outside of a valid path" || true)"
+    if [[ -n "$findings" ]]; then
+        warn "$pkg: namcap errors:"
+        printf '      %s\n' "$findings"
+        return 1
+    fi
+
+    info "installing and launching"
+    sudo pacman -U --noconfirm "$built" >/dev/null 2>&1 \
+        || { warn "$pkg: pacman -U failed"; return 1; }
+    verify_installed "$binary" "$expected" "$wrapper" || { warn "$pkg: launch check failed"; return 1; }
+
+    # Generated before the rule check, which verifies it matches the PKGBUILD.
+    makepkg --printsrcinfo > .SRCINFO
+    check_aur_rules "$pkg" || return 1
+
+    # The AUR repo is a separate remote living alongside the build dir; adding
+    # it here (rather than cloning into a fresh path) keeps the verified
+    # PKGBUILD and the thing being pushed byte-identical.
+    if [[ ! -d .git ]]; then
+        git init -q -b master .
+        git remote add origin "ssh://aur@$AUR_HOST/$pkg.git"
+    fi
+    git remote get-url origin >/dev/null 2>&1 \
+        || git remote add origin "ssh://aur@$AUR_HOST/$pkg.git"
+
+    # Fetch first: an existing AUR package must be built on, not overwritten.
+    if git fetch -q origin master 2>/dev/null; then
+        git reset -q --soft FETCH_HEAD 2>/dev/null || true
+    fi
+
+    cat > .gitignore <<'EOF'
+# Build products; the AUR tracks only the recipe.
+*.pkg.tar.zst
+*.tar.gz
+src/
+pkg/
+EOF
+
+    git add PKGBUILD .SRCINFO .gitignore
+    if git diff --cached --quiet; then
+        info "already up to date on the AUR; nothing to push"
+        PUBLISHED+=("$pkg v$version (unchanged)")
+        return 0
+    fi
+
+    git -c user.name='Krzysztof Rudnicki' \
+        -c user.email='krzysztofrudnicki0@gmail.com' \
+        commit -q -m "$pkg $version-1"
+
+    if ((DRY_RUN)); then
+        info "DRY RUN — not pushing"
+        PUBLISHED+=("$pkg v$version (dry run)")
+        return 0
+    fi
+
+    info "pushing to the AUR"
+    git push -q origin master 2>&1 | tail -3 || { warn "$pkg: push failed"; return 1; }
+    PUBLISHED+=("$pkg v$version")
+    return 0
+}
+

aur-publish/lib/aur_registration.sh

@@ -0,0 +1,107 @@
+#!/bin/bash
+
+# ============================================================================
+# Waiting for AUR registration and SSH access to come up.
+#
+# Sourced by publish.sh, never executed: these functions read the readonly
+# config (AUR_HOST, SSH_KEY, GATE, POLL_SECONDS) that the entry script declares, and
+# a `readonly` re-run inside a twice-sourced lib would abort under `set -e`.
+# ============================================================================
+
+set -euo pipefail
+
+register_status() {
+    python3 "$GATE" /register 2>/dev/null | head -1
+}
+
+open_browser() {
+    local url="$1"
+    local browser
+    for browser in xdg-open thorium-browser chromium google-chrome-stable firefox librewolf; do
+        if command -v "$browser" >/dev/null 2>&1; then
+            "$browser" "$url" >/dev/null 2>&1 &
+            info "Opened $url in $browser"
+            return 0
+        fi
+    done
+    warn "No browser found; open this yourself: $url"
+}
+
+ensure_key() {
+    if [[ ! -f "$SSH_KEY" ]]; then
+        log "Generating an AUR SSH key (none at $SSH_KEY)"
+        ssh-keygen -t ed25519 -f "$SSH_KEY" -N '' -C 'kuhy@aur'
+    fi
+    if ! grep -q "Host $AUR_HOST" "$HOME/.ssh/config" 2>/dev/null; then
+        log "Adding $AUR_HOST to ~/.ssh/config"
+        mkdir -p "$HOME/.ssh"
+        cat >> "$HOME/.ssh/config" <<EOF
+
+# AUR uses a dedicated key so it stays separate from the GitHub identity.
+Host $AUR_HOST
+    User aur
+    IdentityFile $SSH_KEY
+    IdentitiesOnly yes
+EOF
+        chmod 600 "$HOME/.ssh/config"
+    fi
+}
+
+ssh_works() {
+    # The AUR refuses interactive shells; "Interactive shell is disabled" on a
+    # successful key auth is the success signal, not an error.
+    ssh -o BatchMode=yes -o StrictHostKeyChecking=accept-new \
+        -o ConnectTimeout=15 "aur@$AUR_HOST" help 2>&1 |
+        grep -qiE 'interactive shell is disabled|Welcome to AUR'
+}
+
+wait_for_registration() {
+    local status
+    status="$(register_status)"
+    if [[ "$status" == "200" ]]; then
+        log "AUR registration is OPEN"
+    else
+        log "AUR registration is closed (HTTP $status). Polling every ${POLL_SECONDS}s."
+        info "Ctrl-C to stop; re-run any time, nothing is lost."
+        while :; do
+            sleep "$POLL_SECONDS"
+            status="$(register_status)"
+            if [[ "$status" == "200" ]]; then
+                log "Registration is BACK (HTTP 200)"
+                break
+            fi
+            printf '    %s still HTTP %s\n' "$(date +%H:%M)" "$status"
+        done
+    fi
+
+    open_browser "https://$AUR_HOST/register"
+    cat <<EOF
+
+    ------------------------------------------------------------------
+    Sign up, and paste this into the "SSH Public Key" field:
+
+$(sed 's/^/      /' "$SSH_KEY.pub")
+
+    You will need to confirm the address from your email before the
+    account works. This script waits for that automatically.
+    ------------------------------------------------------------------
+EOF
+}
+
+wait_for_ssh() {
+    log "Waiting for the SSH key to authenticate against the AUR"
+    local waited=0
+    until ssh_works; do
+        sleep 30
+        waited=$((waited + 30))
+        if ((waited % 300 == 0)); then
+            info "still waiting ($((waited / 60)) min) — key not accepted yet"
+        fi
+    done
+    log "SSH authentication OK"
+}
+
+# ---------------------------------------------------------------------------
+# Phase 2: build, verify and publish one package
+# ---------------------------------------------------------------------------
+

aur-publish/lib/aur_verify.sh

@@ -0,0 +1,122 @@
+#!/bin/bash
+
+# ============================================================================
+# Gates a package must pass before it is allowed near the AUR.
+#
+# Sourced by publish.sh, never executed: these functions read the readonly
+# config (AUR_ROOT, DRY_RUN) that the entry script declares, and
+# a `readonly` re-run inside a twice-sourced lib would abort under `set -e`.
+# ============================================================================
+
+set -euo pipefail
+
+newest_tag() {
+    # Newest clean vX.Y.Z on the remote. Build tags (vX.Y.Z-build.N) and the
+    # grafted upstream Flutter SDK tags in diet-guard are excluded on purpose.
+    git -C "$1" ls-remote --tags origin 2>/dev/null |
+        awk '{print $2}' |
+        sed 's|refs/tags/||' |
+        grep -E '^v[0-9]+\.[0-9]+\.[0-9]+$' |
+        sed 's/^v//' |
+        sort -V |
+        tail -1
+}
+
+verify_installed() {
+    local binary="$1" expected="$2" wrapper="$3" output
+
+    # An instance left over from a previous run would make the wrapper print
+    # "already running" and hand off, which passes trivially and turns the
+    # strongest gate in this script into the weakest. Clear it first so the
+    # check always exercises a real cold start.
+    #
+    # Resolved through /proc/<pid>/exe rather than any form of name or
+    # command-line matching. Every argv-based approach selects the wrong
+    # processes here:
+    #   * `pkill -x` compares against the kernel's 15-character comm field, and
+    #     two of these wrappers are longer (habit_stack_desktop is 19), so a
+    #     name match would silently never fire.
+    #   * `pkill -f`, and equally `ps -eo args` piped through a pattern, match
+    #     any process whose command line merely *mentions* the path — this
+    #     script, the shell running it, and the matching tool itself all
+    #     qualify, so the run kills its own shell instead of the app. A
+    #     $$/$PPID guard does not cover it, because the offending shell is
+    #     often a more distant ancestor.
+    # /proc/<pid>/exe is a kernel-maintained symlink to the binary actually
+    # being executed, so it cannot collide with an incidental mention.
+    # The wrapper name is unique per app, so the /opt/<pkg>/ segment is left as
+    # a glob instead of threading the package name through another parameter.
+    #
+    # The link is read without `-f`: it is already fully resolved, and by the
+    # time this runs `pacman -U` has replaced the binary, so a leftover process
+    # points at an unlinked inode that the kernel reports with a " (deleted)"
+    # suffix. That suffix has to be stripped or the match silently misses the
+    # one process this exists to kill.
+    local pid exe killed=()
+    for pid in /proc/[0-9]*; do
+        pid="${pid#/proc/}"
+        exe="$(readlink "/proc/$pid/exe" 2>/dev/null)" || continue
+        exe="${exe% (deleted)}"
+        [[ "$exe" == /opt/*/bin/"$wrapper" ]] || continue
+        kill -TERM "$pid" 2>/dev/null && killed+=("$pid")
+    done
+    if ((${#killed[@]})); then
+        info "stopped a running instance so the launch check is meaningful"
+        # `kill` only reports that the signal was sent. Waiting for the process
+        # to actually go keeps a slow exit from failing the cold start below on
+        # a port that is still bound.
+        local waited=0
+        while ((waited < 100)) && kill -0 "${killed[@]}" 2>/dev/null; do
+            sleep 0.1
+            ((waited++))
+        done
+    fi
+
+    output="$(timeout 15 "/usr/bin/$binary" 2>&1 | head -3 || true)"
+    if grep -qF "$expected" <<<"$output"; then
+        return 0
+    fi
+    warn "unexpected startup output: $output"
+    return 1
+}
+
+# The AUR does not review uploads, but "packages that violate the rules may be
+# deleted without warning" (AUR submission guidelines). These are the rules that
+# are mechanically checkable, enforced here so a violation blocks the push
+# rather than surfacing later as a deletion request.
+check_aur_rules() {
+    local pkg="$1" problems=()
+
+    grep -q '^# Maintainer:' PKGBUILD \
+        || problems+=("missing '# Maintainer:' comment on line 1")
+
+    grep -qE "^arch=\(.*x86_64" PKGBUILD \
+        || problems+=("x86_64 not in arch=() — packages without it are not allowed")
+
+    # `replaces` makes pacman rip out the named package on the next -Sy, which
+    # is only correct for a genuine rename. `conflicts` is the right tool.
+    grep -q '^replaces=' PKGBUILD \
+        && problems+=("uses replaces=; AUR rules reserve it for renames — use conflicts=")
+
+    # A package already in an official repo must never be duplicated.
+    if pacman -Si "$pkg" >/dev/null 2>&1; then
+        problems+=("$pkg exists in an official repo — must not be submitted")
+    fi
+
+    grep -qE '^(pkgdesc|url|license)=' PKGBUILD \
+        || problems+=("missing pkgdesc/url/license")
+
+    # A stale .SRCINFO is the single most common upload rejection.
+    if ! diff -q <(makepkg --printsrcinfo) .SRCINFO >/dev/null 2>&1; then
+        problems+=(".SRCINFO does not match PKGBUILD")
+    fi
+
+    if ((${#problems[@]})); then
+        warn "$pkg: AUR rule violations:"
+        printf '      - %s\n' "${problems[@]}"
+        return 1
+    fi
+    info "AUR rule checks passed"
+    return 0
+}
+

aur-publish/publish.sh

@@ -0,0 +1,148 @@
+#!/bin/bash
+
+# ============================================================================
+# Publish kuhy's three Flutter apps to the AUR, hands-off.
+#
+# Run this once AUR registration is back. It:
+#   1. Waits for https://aur.archlinux.org/register to stop returning 503,
+#      then opens it in a browser and prints the SSH key to paste.
+#      (Signup is the ONLY manual step: it needs an email confirmation, and
+#      creating accounts is not something this script will do for you.)
+#   2. Waits for `ssh aur@aur.archlinux.org` to authenticate.
+#   3. From there, fully unattended, for each package:
+#        newest release tag -> pkgver -> updpkgsums -> makepkg -C -> namcap
+#        -> pacman -U -> launch check -> .SRCINFO -> clone AUR repo
+#        -> commit -> push.
+#
+# Every package is BUILT, INSTALLED and LAUNCHED before it is pushed, so a
+# broken package cannot reach the AUR. A package that fails any gate is
+# skipped and reported; the others still publish.
+#
+# Idempotent and resumable: re-running skips work that is already done, and
+# an already-published package is updated rather than duplicated.
+# ============================================================================
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+readonly SCRIPT_DIR
+readonly GATE="$SCRIPT_DIR/anubis_gate.py"
+readonly AUR_HOST="aur.archlinux.org"
+readonly SSH_KEY="$HOME/.ssh/aur"
+readonly AUR_ROOT="$HOME/sdk/aur"
+
+# package:repo:wrapper-process:cli-name:expected-startup-string
+# The wrapper process name is NOT derivable from the package name
+# (diet-guard-app runs diet_guard_desktop), so it is listed explicitly.
+readonly PACKAGES=(
+    "todo-flutter:$HOME/src/todo:todo_desktop:todo:serving on http://localhost:8730"
+    "habit-stack:$HOME/src/habit_stack:habit_stack_desktop:habit-stack:serving on http://localhost:8731"
+    "diet-guard-app:$HOME/src/diet-guard:diet_guard_desktop:diet-guard-app:serving on http://localhost:8732"
+)
+
+POLL_SECONDS=300
+SKIP_WAIT=0
+DRY_RUN=0
+ONLY=""
+FAILED=()
+PUBLISHED=()
+
+usage() {
+    cat <<EOF
+Usage: $(basename "$0") [OPTIONS]
+
+Options:
+  --skip-wait        Assume the account + SSH key already work; go straight
+                     to building and publishing.
+  --dry-run          Do everything except the final 'git push' to the AUR.
+  --only NAME        Publish just one package (todo-flutter, habit-stack,
+                     diet-guard-app).
+  --poll SECONDS     How often to re-check /register (default: $POLL_SECONDS).
+  -h, --help         Show this help.
+EOF
+    exit 0
+}
+
+log()  { printf '\n\033[1;34m==>\033[0m %s\n' "$*"; }
+info() { printf '    %s\n' "$*"; }
+warn() { printf '\033[1;33m  ! %s\033[0m\n' "$*" >&2; }
+die()  { printf '\033[1;31m==> ERROR: %s\033[0m\n' "$*" >&2; exit 1; }
+
+require_tools() {
+    local missing=()
+    local tool
+    for tool in git makepkg updpkgsums namcap ssh python3 curl; do
+        command -v "$tool" >/dev/null 2>&1 || missing+=("$tool")
+    done
+    if ((${#missing[@]})); then
+        info "Installing missing tools: ${missing[*]}"
+        # namcap/updpkgsums come from namcap + pacman-contrib; the rest are base.
+        sudo pacman -S --needed --noconfirm namcap pacman-contrib git openssh python \
+            || die "could not install: ${missing[*]}"
+    fi
+}
+
+# ---------------------------------------------------------------------------
+# Phase 1: wait for registration, then for the key to work
+# ---------------------------------------------------------------------------
+
+# The rest lives in lib/, sourced (not executed) so every function sees the
+# readonly config above and the mutable FAILED/PUBLISHED arrays main() reads.
+# shellcheck source=lib/aur_registration.sh
+source "$SCRIPT_DIR/lib/aur_registration.sh"
+# shellcheck source=lib/aur_verify.sh
+source "$SCRIPT_DIR/lib/aur_verify.sh"
+# shellcheck source=lib/aur_publish_one.sh
+source "$SCRIPT_DIR/lib/aur_publish_one.sh"
+
+main() {
+    require_tools
+    [[ -f "$GATE" ]] || die "missing $GATE"
+    mkdir -p "$AUR_ROOT"
+
+    ensure_key
+    if ((SKIP_WAIT)); then
+        ssh_works || die "SSH key is not accepted yet; drop --skip-wait to wait for it"
+    else
+        wait_for_registration
+        wait_for_ssh
+    fi
+
+    local entry pkg repo wrapper binary expected
+    for entry in "${PACKAGES[@]}"; do
+        IFS=: read -r pkg repo wrapper binary expected <<<"$entry"
+        [[ -z "$ONLY" || "$ONLY" == "$pkg" ]] || continue
+        if ! publish_one "$pkg" "$repo" "$wrapper" "$binary" "$expected"; then
+            FAILED+=("$pkg")
+        fi
+    done
+
+    log "Summary"
+    if ((${#PUBLISHED[@]})); then
+        printf '    published: %s\n' "${PUBLISHED[@]}"
+    fi
+    if ((${#FAILED[@]})); then
+        printf '\033[1;31m    FAILED:    %s\033[0m\n' "${FAILED[@]}"
+        info "Nothing broken was pushed; fix and re-run (safe to repeat)."
+        return 1
+    fi
+    if ((${#PUBLISHED[@]} == 0)); then
+        warn "nothing to do"
+        return 1
+    fi
+    printf '\n    Check: https://aur.archlinux.org/packages/todo-flutter\n'
+    return 0
+}
+
+while [[ $# -gt 0 ]]; do
+    case $1 in
+        --skip-wait) SKIP_WAIT=1; shift ;;
+        --dry-run)   DRY_RUN=1; shift ;;
+        --only)      ONLY="$2"; shift 2 ;;
+        --poll)      POLL_SECONDS="$2"; shift 2 ;;
+        -h|--help)   usage ;;
+        *)           die "Unknown option: $1" ;;
+    esac
+done
+
+main "$@"

code_runner/.pre-commit-config.yaml

@@ -1,4 +1,4 @@
-# Pre-commit Configuration for feedback_inbox
+# Pre-commit Configuration for code_runner
 # Install:  pre-commit install && pre-commit install --hook-type pre-push
 # Run:      pre-commit run --all-files
 # Update:   pre-commit autoupdate
@@ -8,8 +8,8 @@
 # below is written relative to THAT, not to this directory, and the top-level
 # `files` key keeps these hooks off the sibling subprojects. CI invokes this
 # config exactly the same way:
-#   pre-commit run --config feedback_inbox/.pre-commit-config.yaml --all-files
+#   pre-commit run --config code_runner/.pre-commit-config.yaml --all-files
-files: ^feedback_inbox/
+files: ^code_runner/
 
 default_language_version:
   python: python3
@@ -47,7 +47,7 @@ repos:
         pass_filenames: true
       - id: subproject-scope
         name: subproject scope regex still matches
-        entry: scripts/check_subproject_scope.sh feedback_inbox
+        entry: scripts/check_subproject_scope.sh code_runner
         language: script
         # always_run so the top-level `files` filter cannot skip the one hook
         # whose job is to notice that filter has stopped matching anything.
@@ -65,7 +65,7 @@ repos:
         types: [python]
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.17.4
+    rev: v0.17.2
     hooks:
       - id: ruff
         args: [--fix, --unsafe-fixes, --exit-non-zero-on-fix, --show-fixes]
@@ -105,7 +105,7 @@ repos:
     hooks:
       - id: pylint
         args:
-          - --rcfile=feedback_inbox/pyproject.toml
+          - --rcfile=code_runner/pyproject.toml
           - --fail-under=10
           - --jobs=4
         additional_dependencies:
@@ -117,18 +117,18 @@ repos:
       - id: bandit
         args:
           - -c
-          - feedback_inbox/pyproject.toml
+          - code_runner/pyproject.toml
           - --severity-level=high
           - --confidence-level=medium
           - --skip=B113
         additional_dependencies: ["bandit[toml]"]
-        exclude: ^(feedback_inbox/feedback_inbox/tests/|.*test.*\.py$)
+        exclude: ^(code_runner/code_runner/tests/|.*test.*\.py$)
 
   - repo: local
     hooks:
       - id: pytest-coverage
         name: pytest with coverage enforcement
-        entry: scripts/run_subproject_tests.sh feedback_inbox
+        entry: scripts/run_subproject_tests.sh code_runner
         language: script
         types: [python]
         pass_filenames: false

code_runner/README.md

@@ -0,0 +1,12 @@
+# code_runner
+
+Runs a learner's solution (Python 3, JavaScript, TypeScript) against a problem statement's examples inside bubblewrap: no network, read-only `/usr`, scratch dir, CPU/memory/process/file/time limits.
+
+```python
+import code_runner as cr
+
+statement = ...  # markdown from ~/.claude/skills/leetcode/fetch_problem.py
+results = cr.run("python3", code, cr.extract_starter(statement), cr.parse_examples(statement))
+```
+
+`run(lang, code, starter, examples) -> list[CaseResult]` with `n, input, expected, actual, ok, stdout, error, ms`. One sandbox process per example. TypeScript runs through node's type stripping (no type check, no enums). Needs `bwrap`, `prlimit`, `/usr/bin/python3`, node >= 22.18. Install: `./install.sh`.

code_runner/code_runner/__init__.py

@@ -0,0 +1,20 @@
+"""Run a learner's solution against a problem's examples in a bubblewrap sandbox."""
+
+from __future__ import annotations
+
+from code_runner._examples import parse_examples
+from code_runner._model import CaseResult, Example, Limits, RunnerError
+from code_runner._run import run
+from code_runner._starter import Entry, extract_starter, parse_starter
+
+__all__ = [
+    "CaseResult",
+    "Entry",
+    "Example",
+    "Limits",
+    "RunnerError",
+    "extract_starter",
+    "parse_examples",
+    "parse_starter",
+    "run",
+]

code_runner/code_runner/_compare.py

@@ -0,0 +1,89 @@
+"""Output comparison: exact JSON equality, float tolerance, order rules."""
+
+from __future__ import annotations
+
+import json
+import math
+
+TOLERANCE = 1e-5  # LeetCode accepts floating answers within 1e-5
+MAX_DEPTH = 64  # real answers nest a handful of levels; deeper is hostile
+
+
+class TooDeepError(ValueError):
+    """A compared value nests deeper than ``MAX_DEPTH``."""
+
+
+def too_deep(value: object, limit: int = MAX_DEPTH) -> bool:
+    """Whether ``value`` nests lists/dicts deeper than ``limit`` (iterative)."""
+    stack: list[tuple[object, int]] = [(value, 0)]
+    while stack:
+        node, depth = stack.pop()
+        if isinstance(node, list | dict):
+            if depth >= limit:
+                return True
+            children = node if isinstance(node, list) else node.values()
+            stack.extend((child, depth + 1) for child in children)
+    return False
+
+
+def _as_number(value: object) -> float | None:
+    if isinstance(value, bool) or not isinstance(value, int | float):
+        return None
+    return value
+
+
+def _key(value: object) -> str:
+    return json.dumps(value, sort_keys=True, ensure_ascii=False)
+
+
+def _deep_sorted(value: object) -> object:
+    if isinstance(value, list):
+        return sorted((_deep_sorted(item) for item in value), key=_key)
+    return value
+
+
+def _equal(expected: object, actual: object) -> bool:
+    want, got = _as_number(expected), _as_number(actual)
+    if want is not None and got is not None:
+        return math.isclose(want, got, rel_tol=TOLERANCE, abs_tol=TOLERANCE)
+    if isinstance(expected, list) and isinstance(actual, list):
+        return len(expected) == len(actual) and all(
+            _equal(e, a) for e, a in zip(expected, actual, strict=True)
+        )
+    if isinstance(expected, dict) and isinstance(actual, dict):
+        return expected.keys() == actual.keys() and all(
+            _equal(expected[k], actual[k]) for k in expected
+        )
+    return type(expected) is type(actual) and expected == actual
+
+
+def _top_sorted(value: object) -> object:
+    if isinstance(value, list):
+        return sorted(value, key=_key)
+    return value
+
+
+def matches(expected: object, actual: object, order: str) -> bool:
+    """Whether ``actual`` is an accepted answer for ``expected``.
+
+    Args:
+        expected: the statement's answer (JSON value).
+        actual: what the solution returned (JSON value).
+        order: "exact", "any" (list order free) or "deep" (nested order free too).
+
+    Raises:
+        ValueError: ``order`` is none of those.
+        TooDeepError: a value nests deeper than ``MAX_DEPTH`` (the recursive
+            walks below would blow the stack).
+    """
+    if too_deep(expected) or too_deep(actual):
+        msg = f"value nested deeper than {MAX_DEPTH} levels"
+        raise TooDeepError(msg)
+    if order == "exact":
+        return _equal(expected, actual)
+    if order == "any":
+        return _equal(_top_sorted(expected), _top_sorted(actual))
+    if order == "deep":
+        return _equal(_deep_sorted(expected), _deep_sorted(actual))
+    msg = f"unknown order {order!r}"
+    raise ValueError(msg)

code_runner/code_runner/_examples.py

@@ -0,0 +1,75 @@
+from __future__ import annotations
+
+import ast
+import json
+import logging
+import re
+from typing import Final
+
+from code_runner._model import Example, Order
+from code_runner._split import split_top_level
+
+_logger: Final = logging.getLogger(__name__)
+_LABEL: Final = re.compile(r"^(Input|Output)\s*:\s*(.*)$")
+_PAIR: Final = re.compile(r"^([A-Za-z_]\w*)\s*=\s*(.+)$", re.DOTALL)
+_ANY_ORDER: Final = re.compile(r"any order|order of the output", re.IGNORECASE)
+_DEEP_ORDER: Final = re.compile(
+    r"order of the (?:triplets|quadruplets|subsets|groups|pairs|elements)",
+    re.IGNORECASE,
+)
+
+
+def _value(text: str) -> object:
+    try:
+        return json.loads(text)
+    except ValueError:
+        try:
+            return ast.literal_eval(text)
+        except (ValueError, SyntaxError) as exc:
+            raise ValueError(text[:40]) from exc
+
+
+def _arguments(text: str) -> dict[str, object]:
+    args: dict[str, object] = {}
+    for part in split_top_level(text):
+        pair = _PAIR.match(part)
+        if pair is None:
+            msg = f"not a name = value pair: {part[:40]!r}"
+            raise ValueError(msg)
+        args[pair.group(1)] = _value(pair.group(2))
+    return args
+
+
+def _labels(body: str) -> list[tuple[str, str]]:
+    cleaned = (
+        line.replace("**", "").replace("`", "").strip() for line in body.splitlines()
+    )
+    matched = (_LABEL.match(line) for line in cleaned)
+    return [(m.group(1), m.group(2).strip()) for m in matched if m is not None]
+
+
+def _order(body: str) -> Order:
+    if _DEEP_ORDER.search(body):
+        return "deep"
+    return "any" if _ANY_ORDER.search(body) else "exact"
+
+
+def parse_examples(statement: str) -> list[Example]:
+    body = re.split(r"^## ", statement, maxsplit=1, flags=re.MULTILINE)[0]
+    order = _order(body)
+    found: list[Example] = []
+    pending: str | None = None
+    number = 0
+    for label, text in _labels(body):
+        if label == "Input":
+            pending = text
+            continue
+        if pending is None:
+            continue
+        number += 1
+        try:
+            found.append(Example(number, _arguments(pending), _value(text), order))
+        except ValueError as exc:
+            _logger.warning("example %d skipped, not runnable: %s", number, exc)
+        pending = None
+    return found

code_runner/code_runner/_languages.py

@@ -0,0 +1,119 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from importlib import resources
+import json
+import os
+from pathlib import Path
+import shutil
+from typing import TYPE_CHECKING
+
+from code_runner._model import RunnerError
+from code_runner._sandbox import WORK
+
+if TYPE_CHECKING:
+    from code_runner._starter import Entry
+
+MEGABYTE = 1024 * 1024
+_NODE_FLAGS = (
+    "--permission",
+    f"--allow-fs-read={WORK}",
+    f"--allow-fs-write={WORK}",
+    "--max-old-space-size=128",
+)
+
+
+@dataclass(frozen=True)
+class Language:
+    name: str
+    solution_file: str
+    harness_file: str
+    program: str
+    binary_env: str
+    binary_default: str
+    flags: tuple[str, ...]
+    memory_bytes: int
+
+    def resolve(self) -> Path:
+        for candidate in (
+            os.environ.get(self.binary_env),
+            self.binary_default,
+            shutil.which(self.program),
+        ):
+            if candidate and Path(candidate).is_file():
+                return Path(candidate).resolve()
+        msg = f"{self.program} not found (set {self.binary_env})"
+        raise RunnerError(msg)
+
+    def command(self, binary: Path) -> list[str]:
+        return [str(binary), *self.flags, f"{WORK}/{self.harness_file}", WORK]
+
+    def ro_binds(self, binary: Path) -> tuple[str, ...]:
+        if binary.is_relative_to("/usr"):
+            return ()
+        return (str(binary.parent.parent),)
+
+    def prepare(
+        self, scratch: Path, code: str, entry: Entry, args: list[object]
+    ) -> None:
+        harness = resources.files("code_runner").joinpath("harness", self.harness_file)
+        (scratch / self.harness_file).write_text(harness.read_text(encoding="utf-8"))
+        text = code
+        if self.name == "typescript":
+            text += f"\nmodule.exports = {{ __entry: {entry.cls or entry.name} }};\n"
+        (scratch / self.solution_file).write_text(text, encoding="utf-8")
+        spec = {
+            "cls": entry.cls,
+            "fn": entry.name,
+            "args": args,
+            "typescript": self.name == "typescript",
+        }
+        (scratch / "input.json").write_text(json.dumps(spec), encoding="utf-8")
+
+
+PYTHON = Language(
+    "python3",
+    "solution.py",
+    "harness.py",
+    "python3",
+    "CODE_RUNNER_PYTHON",
+    "/usr/bin/python3",
+    ("-I", "-B"),
+    512 * MEGABYTE,
+)
+JAVASCRIPT = Language(
+    "javascript",
+    "solution.js",
+    "harness.js",
+    "node",
+    "CODE_RUNNER_NODE",
+    "/usr/bin/node",
+    _NODE_FLAGS,
+    2048 * MEGABYTE,
+)
+TYPESCRIPT = Language(
+    "typescript",
+    "solution.cts",
+    "harness.js",
+    "node",
+    "CODE_RUNNER_NODE",
+    "/usr/bin/node",
+    _NODE_FLAGS,
+    2048 * MEGABYTE,
+)
+_BY_NAME = {
+    "python3": PYTHON,
+    "python": PYTHON,
+    "javascript": JAVASCRIPT,
+    "js": JAVASCRIPT,
+    "typescript": TYPESCRIPT,
+    "ts": TYPESCRIPT,
+}
+
+
+def get_language(name: str) -> Language:
+    try:
+        return _BY_NAME[name.lower()]
+    except KeyError:
+        msg = f"unsupported language {name!r}; use python3, javascript or typescript"
+        raise RunnerError(msg) from None

code_runner/code_runner/_model.py

@@ -0,0 +1,40 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+from typing import Literal
+
+# "any": list order is free; "deep": the order inside its items is free too.
+Order = Literal["exact", "any", "deep"]
+MAX_CAPTURE = 64 * 1024
+
+
+class RunnerError(Exception):
+    pass
+
+
+@dataclass(frozen=True)
+class Example:
+    n: int
+    args: dict[str, object]
+    expected: object
+    order: Order = "exact"
+
+
+@dataclass(frozen=True)
+class CaseResult:
+    n: int
+    input: dict[str, object]
+    expected: object
+    actual: object
+    ok: bool
+    stdout: str = ""
+    error: str | None = None
+    ms: float = 0.0
+
+
+@dataclass(frozen=True)
+class Limits:
+    timeout_s: float = 2.0
+    nproc: int = 64
+    fsize_bytes: int = 1024 * 1024
+    nofile: int = 256

code_runner/code_runner/_run.py

@@ -0,0 +1,134 @@
+from __future__ import annotations
+
+from collections.abc import Callable, Sequence
+from dataclasses import dataclass
+import json
+import os
+from pathlib import Path
+import stat
+import tempfile
+
+from code_runner._compare import TooDeepError, matches, too_deep
+from code_runner._languages import Language, get_language
+from code_runner._model import MAX_CAPTURE, CaseResult, Example, Limits
+from code_runner._sandbox import BoxResult, execute
+from code_runner._starter import Entry, parse_starter
+
+Box = Callable[[list[str], Path, Limits, int, tuple[str, ...]], BoxResult]
+_OUT_OF_MEMORY = ("out of memory", "allocation failed")
+_MAX_ERROR = 200
+# Everything hostile JSON can raise on the host: deep nesting, huge numbers.
+_HOSTILE = (ValueError, TypeError, RecursionError, MemoryError, OverflowError)
+
+
+def _positional(example: Example, entry: Entry) -> list[object]:
+    if set(example.args) == set(entry.params):
+        return [example.args[name] for name in entry.params]
+    if len(example.args) == len(entry.params):
+        return list(example.args.values())
+    given, wanted = len(example.args), len(entry.params)
+    msg = f"ArgumentError: example gives {given} arguments, {entry.name} takes {wanted}"
+    raise ValueError(msg)
+
+
+def _read_result(scratch: Path) -> dict[str, object] | None:
+    # The scratch dir is the solution's: never follow a link it planted there.
+    flags = os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK
+    try:
+        fd = os.open(scratch / "result.json", flags)
+    except OSError:
+        return None
+    with os.fdopen(fd, "rb") as handle:
+        if not stat.S_ISREG(os.fstat(handle.fileno()).st_mode):
+            return None
+        raw = handle.read(MAX_CAPTURE * 16)
+    try:
+        payload = json.loads(raw)
+    except _HOSTILE:  # ValueError, or RecursionError on "[" * 200000
+        return None
+    if not isinstance(payload, dict) or "ok" not in payload:
+        return None
+    return None if too_deep(payload) else payload
+
+
+def _crash(box: BoxResult) -> str:
+    low = box.stderr.lower()
+    if any(marker in low for marker in _OUT_OF_MEMORY):
+        return "MemoryLimitExceeded"
+    return f"Crashed (exit {box.returncode})"
+
+
+def _took(ms: object, fallback: float) -> float:
+    if isinstance(ms, bool) or not isinstance(ms, int | float):
+        return fallback
+    try:
+        return float(ms)
+    except OverflowError:  # an int too big for a float
+        return fallback
+
+
+def _outcome(scratch: Path, box: BoxResult) -> tuple[object, str | None, float]:
+    if box.timed_out:
+        return None, "TimeLimitExceeded", box.elapsed_ms
+    payload = _read_result(scratch)
+    if payload is None:
+        return None, _crash(box), box.elapsed_ms
+    took = _took(payload.get("ms"), box.elapsed_ms)
+    if payload["ok"] is True:
+        return payload.get("value"), None, took
+    return None, str(payload.get("error", "Error"))[:_MAX_ERROR], took
+
+
+@dataclass(frozen=True)
+class _Job:
+    language: Language
+    binary: Path
+    code: str
+    entry: Entry
+    limits: Limits
+    box: Box
+
+
+def _case(job: _Job, example: Example) -> CaseResult:
+    try:
+        args = _positional(example, job.entry)
+    except ValueError as exc:
+        return CaseResult(
+            example.n, example.args, example.expected, None, ok=False, error=str(exc)
+        )
+    language = job.language
+    with tempfile.TemporaryDirectory(prefix="code-runner-") as tmp:
+        scratch = Path(tmp)
+        language.prepare(scratch, job.code, job.entry, args)
+        done = job.box(
+            language.command(job.binary),
+            scratch,
+            job.limits,
+            language.memory_bytes,
+            language.ro_binds(job.binary),
+        )
+        actual, error, ms = _outcome(scratch, done)
+    ok = False
+    if error is None:
+        try:
+            ok = matches(example.expected, actual, example.order)
+        except (TooDeepError, RecursionError, MemoryError) as exc:
+            actual, error = None, f"InvalidResult: {type(exc).__name__}"
+    return CaseResult(
+        example.n, example.args, example.expected, actual, ok, done.stdout, error, ms
+    )
+
+
+def run(
+    lang: str,
+    code: str,
+    starter: str,
+    examples: Sequence[Example],
+    *,
+    limits: Limits | None = None,
+    box: Box = execute,
+) -> list[CaseResult]:
+    language = get_language(lang)
+    entry = parse_starter(language.name, starter)
+    job = _Job(language, language.resolve(), code, entry, limits or Limits(), box)
+    return [_case(job, example) for example in examples]

code_runner/code_runner/_sandbox.py

@@ -0,0 +1,147 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+import functools
+import math
+import os
+from pathlib import Path
+import signal
+import subprocess
+import tempfile
+import time
+from typing import IO
+
+from code_runner._model import MAX_CAPTURE, Limits, RunnerError
+
+WORK = "/work"
+PRLIMIT = "/usr/bin/prlimit"
+_ROOT_LINKS = ("bin", "sbin", "lib", "lib64")
+
+
+@dataclass(frozen=True)
+class BoxResult:
+    returncode: int
+    timed_out: bool
+    stdout: str
+    stderr: str
+    elapsed_ms: float
+
+
+def _root_links() -> list[str]:
+    args: list[str] = []
+    for name in _ROOT_LINKS:
+        path = Path("/") / name
+        if path.is_symlink():
+            args += ["--symlink", str(path.readlink()), str(path)]
+        elif path.is_dir():
+            args += ["--ro-bind", str(path), str(path)]
+    return args
+
+
+def _base_argv() -> list[str]:
+    return [
+        "bwrap",
+        "--unshare-all",  # no --share-net: no network at all, loopback included
+        "--die-with-parent",
+        "--new-session",
+        "--clearenv",
+        "--ro-bind",
+        "/usr",
+        "/usr",
+        *_root_links(),
+        "--proc",
+        "/proc",
+        "--dev",
+        "/dev",
+        "--tmpfs",
+        "/tmp",
+    ]
+
+
+def bwrap_argv(
+    command: list[str],
+    scratch: Path,
+    limits: Limits,
+    memory_bytes: int,
+    ro_binds: tuple[str, ...] = (),
+) -> list[str]:
+    argv = _base_argv()
+    for bind in ro_binds:
+        argv += ["--ro-bind", bind, bind]
+    argv += ["--bind", str(scratch), WORK, "--remount-ro", "/", "--chdir", WORK]
+    for name, value in (
+        ("HOME", WORK),
+        ("PATH", "/usr/bin"),
+        ("LANG", "C.UTF-8"),
+        ("PYTHONIOENCODING", "utf-8"),
+    ):
+        argv += ["--setenv", name, value]
+    cpu = math.ceil(limits.timeout_s) + 1
+    return [
+        *argv,
+        "--",
+        PRLIMIT,
+        f"--as={memory_bytes}",
+        f"--nproc={limits.nproc}",
+        f"--fsize={limits.fsize_bytes}",
+        f"--nofile={limits.nofile}",
+        f"--cpu={cpu}",
+        "--core=0",
+        "--",
+        *command,
+    ]
+
+
+@functools.cache
+def check_sandbox() -> None:
+    if not Path(PRLIMIT).is_file():
+        msg = f"{PRLIMIT} is missing (util-linux)"
+        raise RunnerError(msg)
+    try:
+        probe = subprocess.run(
+            [*_base_argv(), "/usr/bin/true"],
+            capture_output=True,
+            check=False,
+            timeout=20,
+        )
+    except OSError as exc:
+        msg = f"bubblewrap (bwrap) cannot start: {exc}"
+        raise RunnerError(msg) from exc
+    if probe.returncode != 0:
+        detail = probe.stderr.decode(errors="replace").strip()
+        msg = f"bubblewrap cannot create the sandbox: {detail}"
+        raise RunnerError(msg)
+
+
+def _read(handle: IO[bytes]) -> str:
+    handle.seek(0)
+    return handle.read(MAX_CAPTURE).decode("utf-8", errors="replace")
+
+
+def execute(
+    command: list[str],
+    scratch: Path,
+    limits: Limits,
+    memory_bytes: int,
+    ro_binds: tuple[str, ...] = (),
+) -> BoxResult:
+    check_sandbox()
+    argv = bwrap_argv(command, scratch, limits, memory_bytes, ro_binds)
+    with tempfile.TemporaryFile() as out, tempfile.TemporaryFile() as err:
+        started = time.monotonic()
+        with subprocess.Popen(
+            argv,
+            stdin=subprocess.DEVNULL,
+            stdout=out,
+            stderr=err,
+            env={"PATH": "/usr/bin:/bin"},
+            start_new_session=True,
+        ) as proc:
+            timed_out = False
+            try:
+                proc.wait(timeout=limits.timeout_s)
+            except subprocess.TimeoutExpired:
+                timed_out = True
+                os.killpg(proc.pid, signal.SIGKILL)
+        elapsed = (time.monotonic() - started) * 1000
+        return BoxResult(proc.returncode, timed_out, _read(out), _read(err), elapsed)

code_runner/code_runner/_split.py

@@ -0,0 +1,29 @@
+from __future__ import annotations
+
+_OPEN = "([{<"
+_CLOSE = ")]}>"
+
+
+def split_top_level(text: str) -> list[str]:
+    parts: list[str] = []
+    depth = 0
+    quote = ""
+    escaped = False
+    start = 0
+    for i, ch in enumerate(text):
+        if escaped:
+            escaped = False
+        elif quote:
+            escaped = ch == "\\"
+            quote = "" if ch == quote else quote
+        elif ch in "\"'":
+            quote = ch
+        elif ch in _OPEN:
+            depth += 1
+        elif ch in _CLOSE and not (ch == ">" and text[i - 1] in "=-"):
+            depth = max(depth - 1, 0)
+        elif ch == "," and depth == 0:
+            parts.append(text[start:i])
+            start = i + 1
+    parts.append(text[start:])
+    return [part.strip() for part in parts]

code_runner/code_runner/_starter.py

@@ -0,0 +1,108 @@
+from __future__ import annotations
+
+from dataclasses import dataclass
+import re
+
+from code_runner._model import RunnerError
+from code_runner._split import split_top_level
+
+_IDENT = r"[A-Za-z_$][\w$]*"
+_BLOCK_COMMENT = re.compile(r"/\*.*?\*/", re.DOTALL)
+_LINE_COMMENT = re.compile(r"^\s*(?://|#).*$", re.MULTILINE)
+_PY_CLASS = re.compile(r"^class\s+(\w+)", re.MULTILINE)
+_PY_DEF = re.compile(r"\bdef\s+(?!__)(\w+)\s*\(")
+_PY_FUNC = re.compile(r"^def\s+(\w+)\s*\(", re.MULTILINE)
+_JS_CLASS = re.compile(rf"^\s*class\s+({_IDENT})", re.MULTILINE)
+_JS_METHOD = re.compile(
+    rf"^\s+(?:async\s+)?(?!constructor\b)({_IDENT})\s*\(", re.MULTILINE
+)
+_JS_ASSIGNED = re.compile(
+    rf"\b(?:var|let|const)\s+({_IDENT})\s*(?::[^=]+)?=\s*(?:async\s+)?(?:function\b\s*\w*\s*)?\("
+)
+_JS_DECLARED = re.compile(rf"\bfunction\s+({_IDENT})\s*\(")
+_NOT_METHODS = {"if", "for", "while", "switch", "catch", "return", "function"}
+
+
+@dataclass(frozen=True)
+class Entry:
+    cls: str | None
+    name: str
+    params: tuple[str, ...]
+
+
+def _no_entry(lang: str) -> RunnerError:
+    return RunnerError(f"no entry point found in the {lang} starter code")
+
+
+def _balanced(text: str, open_at: int) -> str:
+    depth = 0
+    for i in range(open_at, len(text)):
+        ch = text[i]
+        if ch in "([{" or (ch == "<" and depth > 0):
+            depth += 1
+        elif ch in ")]}" or (ch == ">" and text[i - 1] != "=" and depth > 1):
+            depth -= 1
+            if depth == 0:
+                return text[open_at + 1 : i]
+    return text[open_at + 1 :]
+
+
+def _params(code: str, open_at: int) -> tuple[str, ...]:
+    names: list[str] = []
+    for part in split_top_level(_balanced(code, open_at)):
+        found = re.match(_IDENT, part.strip())
+        if found:
+            names.append(found.group())
+    return tuple(names)
+
+
+def _python(text: str) -> Entry | None:
+    code = _LINE_COMMENT.sub("", text)
+    owner = _PY_CLASS.search(code)
+    if owner is None:
+        found = _PY_FUNC.search(code)
+        return found and Entry(None, found.group(1), _params(code, found.end() - 1))
+    found = _PY_DEF.search(code, owner.end())
+    if found is None:
+        return None
+    params = _params(code, found.end() - 1)
+    return Entry(
+        owner.group(1), found.group(1), tuple(p for p in params if p != "self")
+    )
+
+
+def _javascript(text: str) -> Entry | None:
+    code = _LINE_COMMENT.sub("", _BLOCK_COMMENT.sub("", text))
+    owner = _JS_CLASS.search(code)
+    if owner is not None:
+        for method in _JS_METHOD.finditer(code, owner.end()):
+            if method.group(1) not in _NOT_METHODS:
+                return Entry(
+                    owner.group(1), method.group(1), _params(code, method.end() - 1)
+                )
+        return None
+    found = _JS_ASSIGNED.search(code) or _JS_DECLARED.search(code)
+    return found and Entry(None, found.group(1), _params(code, found.end() - 1))
+
+
+def parse_starter(lang: str, text: str) -> Entry:
+    entry = _python(text) if lang == "python3" else _javascript(text)
+    if entry is None:
+        raise _no_entry(lang)
+    return entry
+
+
+def extract_starter(statement: str) -> str:
+    lines = statement.splitlines(keepends=True)
+    heads = [i for i, line in enumerate(lines) if line.startswith("## Starter code")]
+    if not heads:
+        return ""
+    body = lines[heads[0] + 1 :]
+    if body and body[0].startswith("```"):
+        body = body[1:]
+        end = next(
+            (i for i, line in enumerate(body) if line.startswith("```")), len(body)
+        )
+        return "".join(body[:end])
+    end = next((i for i, line in enumerate(body) if line.startswith("## ")), len(body))
+    return "".join(body[:end])

code_runner/code_runner/harness/harness.js

@@ -0,0 +1,45 @@
+"use strict";
+// Runs inside the sandbox: call the solution once, write result.json.
+const fs = require("node:fs");
+const path = require("node:path");
+
+const root = process.argv[2];
+const spec = JSON.parse(fs.readFileSync(path.join(root, "input.json"), "utf8"));
+const IDENT = /^[A-Za-z_$][\w$]*$/;
+
+function describe(e) {
+  const name = e && typeof e.name === "string" ? e.name : "Error";
+  const message = e && e.message !== undefined ? e.message : String(e);
+  return `${name}: ${message}`.slice(0, 300);
+}
+
+function writeResult(payload) {
+  // Unique name + rename: a symlink planted at result.json is replaced, never followed.
+  const unique = path.join(root, `.result-${process.pid}-${Math.random().toString(36).slice(2)}`);
+  fs.writeFileSync(unique, JSON.stringify(payload), { flag: "wx" });
+  fs.renameSync(unique, path.join(root, "result.json"));
+}
+
+function load(target) {
+  if (!IDENT.test(target)) throw new TypeError("bad entry name");
+  if (spec.typescript) return require(path.join(root, "solution.cts")).__entry;
+  const source = fs.readFileSync(path.join(root, "solution.js"), "utf8");
+  return new Function(`${source}\n;return ${target};`)();
+}
+
+function call() {
+  const entry = load(spec.cls || spec.fn);
+  if (spec.cls) return new entry()[spec.fn](...spec.args);
+  return entry(...spec.args);
+}
+
+const started = process.hrtime.bigint();
+let payload;
+try {
+  const value = call();
+  payload = { ok: true, value: value === undefined ? null : JSON.parse(JSON.stringify(value)) };
+} catch (e) {
+  payload = { ok: false, error: describe(e) };
+}
+payload.ms = Number(process.hrtime.bigint() - started) / 1e6;
+writeResult(payload);

code_runner/code_runner/harness/harness.py

@@ -0,0 +1,69 @@
+"""Runs inside the sandbox: call the solution once, write result.json."""
+
+from __future__ import annotations
+
+import json
+import os
+from pathlib import Path
+import sys
+import tempfile
+import time
+from typing import Any
+
+PRELUDE = (
+    "from typing import *\n"
+    "import bisect, collections, functools, heapq, itertools, math\n"
+    "import operator, re, string\n"
+)
+MAX_MESSAGE = 300
+
+
+def describe(exc: BaseException) -> str:
+    return f"{type(exc).__name__}: {exc}"[:MAX_MESSAGE]
+
+
+def write_result(root: Path, payload: dict[str, object]) -> None:
+    # Unique name + rename: a symlink the solution planted at result.json is
+    # replaced by this file, never followed.
+    handle, name = tempfile.mkstemp(dir=root, prefix=".result-")
+    with os.fdopen(handle, "w", encoding="utf-8") as out:
+        json.dump(payload, out)
+    Path(name).replace(root / "result.json")
+
+
+def call(root: Path, spec: dict[str, Any]) -> object:
+    namespace: dict[str, Any] = {"__name__": "solution"}
+    exec(PRELUDE, namespace)
+    source = (root / "solution.py").read_text(encoding="utf-8")
+    exec(compile(source, "solution.py", "exec"), namespace)
+    owner, name = spec["cls"], str(spec["fn"])
+    if owner is None:
+        target = namespace.get(name)
+    else:
+        cls = namespace.get(str(owner))
+        target = getattr(cls() if cls else None, name, None)
+    if target is None:
+        message = f"name {owner or name!r} is not defined"
+        raise NameError(message)
+    return target(*spec["args"])
+
+
+def main(directory: str) -> None:
+    root = Path(directory)
+    spec = json.loads((root / "input.json").read_text(encoding="utf-8"))
+    started = time.perf_counter()
+    try:
+        value = call(root, spec)
+        payload: dict[str, object] = {
+            "ok": True,
+            "value": json.loads(json.dumps(value)),
+        }
+    except BaseException as exc:
+        payload = {"ok": False, "error": describe(exc)}
+    payload["ms"] = (time.perf_counter() - started) * 1000
+    sys.stdout.flush()
+    write_result(root, payload)
+
+
+if __name__ == "__main__":
+    main(sys.argv[1])

code_runner/code_runner/tests/conftest.py

@@ -0,0 +1,70 @@
+"""Shared fixtures: real-sandbox tests are skipped where bubblewrap cannot run."""
+
+from __future__ import annotations
+
+import shutil
+
+import pytest
+
+from code_runner._model import Example, RunnerError
+from code_runner._sandbox import check_sandbox
+
+
+def _bwrap_works() -> bool:
+    try:
+        check_sandbox()
+    except RunnerError:
+        return False
+    return True
+
+
+needs_sandbox = pytest.mark.skipif(not _bwrap_works(), reason="bwrap/userns unusable")
+needs_node = pytest.mark.skipif(
+    not _bwrap_works() or not shutil.which("node"), reason="node or bwrap missing"
+)
+
+TWO_SUM_STARTER = {
+    "python3": "class Solution:\n    def twoSum(self, nums: list[int], target: int) -> list[int]:\n",
+    "javascript": "var twoSum = function(nums, target) {\n\n};",
+    "typescript": "function twoSum(nums: number[], target: number): number[] {\n};",
+}
+TWO_SUM_CODE = {
+    "python3": (
+        "class Solution:\n"
+        "    def twoSum(self, nums, target):\n"
+        "        seen = {}\n"
+        "        for i, n in enumerate(nums):\n"
+        "            if target - n in seen:\n"
+        "                return [i, seen[target - n]]\n"
+        "            seen[n] = i\n"
+    ),
+    "javascript": (
+        "var twoSum = function(nums, target) {\n"
+        "  const seen = new Map();\n"
+        "  for (let i = 0; i < nums.length; i++) {\n"
+        "    if (seen.has(target - nums[i])) return [i, seen.get(target - nums[i])];\n"
+        "    seen.set(nums[i], i);\n"
+        "  }\n"
+        "};"
+    ),
+    "typescript": (
+        "function twoSum(nums: number[], target: number): number[] {\n"
+        "  const seen = new Map<number, number>();\n"
+        "  for (let i = 0; i < nums.length; i++) {\n"
+        "    const j = seen.get(target - nums[i]);\n"
+        "    if (j !== undefined) return [i, j];\n"
+        "    seen.set(nums[i], i);\n"
+        "  }\n"
+        "  return [];\n"
+        "}"
+    ),
+}
+
+
+@pytest.fixture
+def two_sum_examples() -> list[Example]:
+    return [
+        Example(1, {"nums": [2, 7, 11, 15], "target": 9}, [0, 1], "any"),
+        Example(2, {"nums": [3, 2, 4], "target": 6}, [1, 2], "any"),
+        Example(3, {"nums": [3, 3], "target": 6}, [0, 1], "any"),
+    ]

code_runner/code_runner/tests/test_compare.py

@@ -0,0 +1,65 @@
+import pytest
+
+from code_runner._compare import MAX_DEPTH, TooDeepError, matches, too_deep
+
+
+@pytest.mark.parametrize(
+    ("expected", "actual", "order", "want"),
+    [
+        ([0, 1], [0, 1], "exact", True),
+        ([0, 1], [1, 0], "exact", False),
+        ([0, 1], [1, 0], "any", True),
+        ([0, 1], [1, 0, 2], "any", False),
+        ([[1, 2], [3]], [[3], [1, 2]], "any", True),
+        ([[1, 2], [3]], [[3], [2, 1]], "any", False),
+        ([[1, 2], [3]], [[3], [2, 1]], "deep", True),
+        ([[1, 2], [3]], [[3], [2, 9]], "deep", False),
+        (3, 3.0, "exact", True),
+        (0.1 + 0.2, 0.3, "exact", True),
+        (2.0, 2.001, "exact", False),
+        (True, 1, "exact", False),
+        (1, True, "exact", False),
+        (True, True, "exact", True),
+        (None, None, "exact", True),
+        (None, 0, "exact", False),
+        ("ab", "ab", "exact", True),
+        ("ab", "ba", "exact", False),
+        ("1", 1, "exact", False),
+        ({"a": [1]}, {"a": [1]}, "exact", True),
+        ({"a": 1}, {"b": 1}, "exact", False),
+        ({"a": 1}, {"a": 2}, "exact", False),
+        (5, 5, "any", True),
+        ([1], 1, "any", False),
+        ([1.0, 2.0], [2.0, 1.0000001], "any", True),
+    ],
+)
+def test_matches(expected: object, actual: object, order: str, *, want: bool) -> None:
+    assert matches(expected, actual, order) is want
+
+
+def test_unknown_order_is_refused() -> None:
+    with pytest.raises(ValueError, match="order"):
+        matches(1, 1, "sideways")
+
+
+def _nest(depth: int) -> object:
+    value: object = 1
+    for _ in range(depth):
+        value = [value]
+    return value
+
+
+@pytest.mark.parametrize("order", ["exact", "any", "deep"])
+def test_deep_nesting_is_refused_not_recursed(order: str) -> None:
+    deep = _nest(100_000)  # would raise RecursionError in the recursive walks
+    with pytest.raises(TooDeepError):
+        matches(deep, deep, order)
+    with pytest.raises(TooDeepError):
+        matches(1, deep, order)
+
+
+def test_nesting_up_to_the_limit_still_compares() -> None:
+    assert matches(_nest(MAX_DEPTH), _nest(MAX_DEPTH), "deep") is True
+    assert too_deep(_nest(MAX_DEPTH + 1)) is True
+    assert too_deep({"a": [{"b": _nest(MAX_DEPTH)}]}) is True
+    assert too_deep({"a": [{"b": 1}]}) is False

code_runner/code_runner/tests/test_examples.py

@@ -0,0 +1,106 @@
+from code_runner._examples import parse_examples
+from code_runner._model import Example
+
+TWO_SUM = """# 1. Two Sum (Easy)
+Tags: Array, Hash Table
+
+Return indices.
+
+You can return the answer in any order.
+
+**Example 1:**
+
+**Input:** `nums = [2,7,11,15], target = 9`
+
+**Output:** `[0,1]`
+
+**Explanation:** 2 + 7.
+
+**Example 2:**
+
+**Input:** `nums = [3,2,4], target = 6`
+
+**Output:** `[1,2]`
+
+## Starter code (python3)
+```
+class Solution:
+    def twoSum(self, nums, target):
+```
+"""
+
+FENCED = """# 5. Pal
+
+Output order does not matter.
+
+**Example 1:**
+```
+Input: s = "a,b", flag = true
+Output: "b"
+Explanation: Output: not an example
+```
+**Example 2:**
+```
+Input: s = 'x', flag = false
+Output: None
+```
+"""
+
+THREE_SUM = """# 15. 3Sum
+Notice that the order of the output and the order of the triplets does not matter.
+**Example 1:**
+**Input:** nums = [-1,0,1,2,-1,-4]
+**Output:** [[-1,-1,2],[-1,0,1]]
+"""
+
+SKIPPED = """# 2. Add Two Numbers
+**Example 1:**
+**Input:** l1 = 1->2
+**Output:** [1,2]
+**Example 2:**
+**Input:** a = 1, b = 2
+**Output:** 3.00000
+**Example 3:**
+**Input:** a = 1
+**Output:** 1->2
+**Example 4:**
+**Input:** a
+**Output:** 1
+"""
+
+
+def test_two_sum_examples_are_typed_and_ordered_any() -> None:
+    found = parse_examples(TWO_SUM)
+    assert found == [
+        Example(1, {"nums": [2, 7, 11, 15], "target": 9}, [0, 1], "any"),
+        Example(2, {"nums": [3, 2, 4], "target": 6}, [1, 2], "any"),
+    ]
+
+
+def test_fenced_examples_and_python_literals() -> None:
+    found = parse_examples(FENCED)
+    assert [e.args for e in found] == [
+        {"s": "a,b", "flag": True},
+        {"s": "x", "flag": False},
+    ]
+    assert [e.expected for e in found] == ["b", None]
+    assert {e.order for e in found} == {"exact"}
+
+
+def test_triplets_sentence_means_deep_order() -> None:
+    assert parse_examples(THREE_SUM)[0].order == "deep"
+
+
+def test_unparseable_examples_are_skipped_but_numbering_is_kept() -> None:
+    found = parse_examples(SKIPPED)
+    assert found == [Example(2, {"a": 1, "b": 2}, 3.0, "exact")]
+
+
+def test_output_without_input_and_empty_statement() -> None:
+    assert parse_examples("**Output:** 1\n") == []
+    assert parse_examples("") == []
+
+
+def test_starter_section_is_never_scanned() -> None:
+    text = "# t\n## Starter code (python3)\n```\nInput: a = 1\nOutput: 2\n```\n"
+    assert parse_examples(text) == []

code_runner/code_runner/tests/test_internals.py

@@ -0,0 +1,235 @@
+"""Branches that a healthy sandbox never reaches: crashes, bad results, no bwrap."""
+
+from __future__ import annotations
+
+import dataclasses
+import os
+from pathlib import Path
+import subprocess
+from typing import TYPE_CHECKING
+
+import pytest
+
+from code_runner import Example, Limits, RunnerError, run
+from code_runner._examples import parse_examples
+from code_runner._languages import JAVASCRIPT, PYTHON, get_language
+from code_runner._run import Box, _read_result
+from code_runner._sandbox import BoxResult, bwrap_argv, check_sandbox
+from code_runner._starter import parse_starter
+
+if TYPE_CHECKING:
+    from collections.abc import Iterator
+
+STARTER = "def f(a):\n"
+EXAMPLE = Example(1, {"a": 1}, 1)
+
+
+@pytest.fixture(autouse=True)
+def _fresh_sandbox_check() -> Iterator[None]:
+    check_sandbox.cache_clear()
+    yield
+    check_sandbox.cache_clear()
+
+
+def _fake(result: BoxResult, files: dict[str, str] | None = None) -> Box:
+    def box(
+        command: list[str],
+        scratch: Path,
+        limits: Limits,
+        memory: int,
+        binds: tuple[str, ...],
+    ) -> BoxResult:
+        for name, text in (files or {}).items():
+            (scratch / name).write_text(text)
+        return result
+
+    return box
+
+
+def _box(**kw: object) -> BoxResult:
+    fields = {
+        "returncode": 0,
+        "timed_out": False,
+        "stdout": "",
+        "stderr": "",
+        "elapsed_ms": 7.0,
+    }
+    return BoxResult(**(fields | kw))
+
+
+def _run(box: Box) -> object:
+    return run("python3", "", STARTER, [EXAMPLE], box=box)[0]
+
+
+def test_crash_without_result_names_the_exit_code() -> None:
+    result = _run(_fake(_box(returncode=139)))
+    assert (result.ok, result.actual, result.error, result.ms) == (
+        False,
+        None,
+        "Crashed (exit 139)",
+        7.0,
+    )
+
+
+def test_out_of_memory_crash_is_named() -> None:
+    stderr = "FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory"
+    assert (
+        _run(_fake(_box(returncode=134, stderr=stderr))).error == "MemoryLimitExceeded"
+    )
+
+
+def test_timeout_wins_over_a_result_file() -> None:
+    files = {"result.json": '{"ok": true, "value": 1}'}
+    result = _run(_fake(_box(timed_out=True, elapsed_ms=1000.0), files))
+    assert (result.error, result.ms) == ("TimeLimitExceeded", 1000.0)
+
+
+def test_result_file_variants() -> None:
+    good = _run(
+        _fake(_box(stdout="hi"), {"result.json": '{"ok": true, "value": 1, "ms": 2.5}'})
+    )
+    assert (good.ok, good.ms, good.stdout) == (True, 2.5, "hi")
+    no_ms = _run(_fake(_box(), {"result.json": '{"ok": true, "value": 1}'}))
+    assert no_ms.ms == 7.0
+    failed = _run(_fake(_box(), {"result.json": '{"ok": false}'}))
+    assert failed.error == "Error"
+    named = _run(
+        _fake(_box(), {"result.json": '{"ok": false, "error": "KeyError: x"}'})
+    )
+    assert named.error == "KeyError: x"
+
+
+def test_hostile_result_files_cost_one_example_not_the_run() -> None:
+    hostile = {
+        "bomb": "[" * 200_000,
+        "deep": '{"ok": true, "value": ' + "[" * 5000 + "]" * 5000 + "}",
+        "huge_ms": '{"ok": true, "value": 1, "ms": 1' + "0" * 400 + "}",
+        "bool_ms": '{"ok": true, "value": 1, "ms": true}',
+        "long_error": '{"ok": false, "error": "' + "x" * 5000 + '"}',
+    }
+    results = {
+        name: _run(_fake(_box(), {"result.json": text}))
+        for name, text in hostile.items()
+    }
+    assert results["bomb"].error == "Crashed (exit 0)"
+    assert results["deep"].error == "Crashed (exit 0)"
+    assert (results["huge_ms"].ok, results["huge_ms"].ms) == (True, 7.0)
+    assert results["bool_ms"].ms == 7.0
+    assert len(results["long_error"].error) == 200
+
+
+def test_too_deep_expected_or_actual_is_that_examples_error() -> None:
+    deep: object = 1
+    for _ in range(100):
+        deep = [deep]
+    files = {"result.json": '{"ok": true, "value": 1}'}
+    cases = [Example(1, {"a": 1}, deep), EXAMPLE]
+    results = run("python3", "", STARTER, cases, box=_fake(_box(), files))
+    assert (results[0].ok, results[0].actual) == (False, None)
+    assert results[0].error == "InvalidResult: TooDeepError"
+    assert results[1].ok is True  # the run went on
+
+
+def test_recursion_in_compare_is_that_examples_error(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    def boom(*args: object) -> bool:
+        raise RecursionError
+
+    monkeypatch.setattr("code_runner._run.matches", boom)
+    files = {"result.json": '{"ok": true, "value": 1}'}
+    result = _run(_fake(_box(), files))
+    assert result.error == "InvalidResult: RecursionError"
+
+
+def test_unreadable_result_files_are_a_crash(tmp_path: Path) -> None:
+    secret = tmp_path / "secret"
+    secret.write_text('{"ok": true, "value": "stolen"}')
+    (tmp_path / "result.json").symlink_to(secret)
+    assert _read_result(tmp_path) is None
+    (tmp_path / "result.json").unlink()
+    os.mkfifo(tmp_path / "result.json")
+    assert _read_result(tmp_path) is None
+    (tmp_path / "result.json").unlink()
+    for text in ("not json", "[1]", '{"no": "ok"}'):
+        (tmp_path / "result.json").write_text(text)
+        assert _read_result(tmp_path) is None
+    assert _read_result(tmp_path / "missing") is None
+
+
+def test_binary_resolution(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> None:
+    ghost = dataclasses.replace(
+        PYTHON, binary_default="/nope", program="no-such-program-xyz"
+    )
+    monkeypatch.delenv(ghost.binary_env, raising=False)
+    with pytest.raises(RunnerError, match="no-such-program-xyz not found"):
+        ghost.resolve()
+    tool = tmp_path / "bin" / "python3"
+    tool.parent.mkdir()
+    tool.write_text("#!/bin/sh\n")
+    monkeypatch.setenv(ghost.binary_env, str(tool))
+    assert ghost.resolve() == tool
+    assert ghost.ro_binds(tool) == (str(tool.parent.parent),)
+    assert PYTHON.ro_binds(Path("/usr/bin/python3")) == ()
+    assert get_language("TS").name == "typescript"
+
+
+def _has(argv: list[str], seq: list[str]) -> bool:
+    return any(argv[i : i + len(seq)] == seq for i in range(len(argv)))
+
+
+def test_argv_binds_links_and_limits(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
+) -> None:
+    monkeypatch.setattr("code_runner._sandbox._ROOT_LINKS", ("usr", "no-such-root-dir"))
+    argv = bwrap_argv(["prog"], tmp_path, Limits(timeout_s=1.2), 123, ("/opt/x",))
+    assert _has(argv, ["--ro-bind", "/usr", "/usr"])
+    assert _has(argv, ["--ro-bind", "/opt/x", "/opt/x"])
+    assert _has(argv, ["--bind", str(tmp_path), "/work"])
+    assert "--share-net" not in argv
+    assert {"--unshare-all", "--die-with-parent", "--clearenv", "--remount-ro"} <= set(
+        argv
+    )
+    assert "--as=123" in argv
+    assert "--cpu=3" in argv
+    assert argv[-1] == "prog"
+
+
+def test_sandbox_preflight_failures(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setattr("code_runner._sandbox.PRLIMIT", "/no/such/prlimit")
+    with pytest.raises(RunnerError, match="prlimit"):
+        check_sandbox()
+    monkeypatch.undo()
+
+    def missing(*args: object, **kwargs: object) -> None:
+        message = "bwrap"
+        raise FileNotFoundError(message)
+
+    monkeypatch.setattr(subprocess, "run", missing)
+    with pytest.raises(RunnerError, match="cannot start"):
+        check_sandbox()
+
+    def refused(*args: object, **kwargs: object) -> subprocess.CompletedProcess[bytes]:
+        return subprocess.CompletedProcess(args, 1, b"", b"bwrap: No permissions\n")
+
+    monkeypatch.setattr(subprocess, "run", refused)
+    with pytest.raises(RunnerError, match="No permissions"):
+        check_sandbox()
+
+
+def test_statement_and_starter_edge_cases() -> None:
+    escaped = parse_examples('**Input:** s = "a\\"b,c", t = 1\n**Output:** "x"\n')
+    assert escaped[0].args == {"s": 'a"b,c', "t": 1}
+    assert parse_starter("python3", "def f(a, b").params == ("a", "b")
+    skipped = "class S {\n  if (x) {\n  solve(a) {\n  }\n}\n"
+    assert parse_starter("javascript", skipped).name == "solve"
+
+
+def test_typescript_class_entry_is_exported(tmp_path: Path) -> None:
+    entry = parse_starter(
+        "typescript", "class Solution {\n  solve(a: number): number {\n  }\n}\n"
+    )
+    get_language("typescript").prepare(tmp_path, "class Solution {}", entry, [1])
+    assert "__entry: Solution" in (tmp_path / "solution.cts").read_text()
+    JAVASCRIPT.prepare(tmp_path, "x", entry, [1])
+    assert (tmp_path / "solution.js").read_text() == "x"

code_runner/code_runner/tests/test_isolation.py

@@ -0,0 +1,206 @@
+"""Escape attempts: each one runs hostile code in the real sandbox and must fail."""
+
+from __future__ import annotations
+
+from pathlib import Path
+import socket
+import subprocess
+import time
+from typing import Any, cast
+
+import pytest
+
+from code_runner import CaseResult, Example, Limits, run
+from code_runner.tests.conftest import needs_node, needs_sandbox
+
+pytestmark = needs_sandbox
+STARTER = "def f(a):\n"
+FAST = Limits(timeout_s=1.0)
+
+
+def _one(code: str, lang: str = "python3", limits: Limits = FAST) -> CaseResult:
+    starter = {"python3": STARTER, "javascript": "var f = function(a) {};"}[lang]
+    return run(lang, code, starter, [Example(1, {"a": 0}, 1)], limits=limits)[0]
+
+
+def _pgrep(pattern: str) -> int:
+    done = subprocess.run(
+        ["/usr/bin/pgrep", "-fc", pattern], capture_output=True, text=True, check=False
+    )
+    return int(done.stdout.strip() or 0)
+
+
+def test_no_network_not_even_host_loopback() -> None:
+    with socket.socket() as listener:
+        listener.bind(("127.0.0.1", 0))
+        listener.listen()
+        listener.settimeout(0)
+        port = listener.getsockname()[1]
+        code = (
+            "import socket\n"
+            "def f(a):\n"
+            "    out = []\n"
+            f"    for host, port in [('127.0.0.1', {port}), ('1.1.1.1', 53), ('::1', {port})]:\n"
+            "        try:\n"
+            "            socket.create_connection((host, port), timeout=1).close()\n"
+            "            out.append('connected')\n"
+            "        except OSError:\n"
+            "            out.append('blocked')\n"
+            "    return out\n"
+        )
+        result = _one(code)
+        with pytest.raises(BlockingIOError):
+            listener.accept()
+    assert result.actual == ["blocked", "blocked", "blocked"]
+
+
+@needs_node
+def test_no_network_from_node() -> None:
+    with socket.socket() as listener:
+        listener.bind(("127.0.0.1", 0))
+        listener.listen()
+        port = listener.getsockname()[1]
+        code = (
+            "var f = function(a) {\n"
+            "  const net = process.getBuiltinModule('node:net');\n"
+            "  let seen = 'pending';\n"
+            f"  const s = net.connect({port}, '127.0.0.1');\n"
+            "  s.on('connect', () => { seen = 'connected'; });\n"
+            "  s.on('error', () => {});\n"
+            "  return seen;\n"
+            "};"
+        )
+        result = _one(code, "javascript")
+        listener.settimeout(0.2)
+        with pytest.raises(TimeoutError):
+            listener.accept()
+    assert result.actual == "pending"
+
+
+def test_writes_outside_scratch_never_reach_the_host(tmp_path: Path) -> None:
+    marker = tmp_path / "marker"
+    code = (
+        "def f(a):\n"
+        "    out = []\n"
+        f"    for path in ['/usr/escape', '/etc/escape', '/escape', {str(marker)!r}]:\n"
+        "        try:\n"
+        "            open(path, 'w').write('x')\n"
+        "            out.append('wrote')\n"
+        "        except OSError:\n"
+        "            out.append('denied')\n"
+        "    return out\n"
+    )
+    result = _one(code)
+    assert result.actual == ["denied", "denied", "denied", "denied"]
+    assert not marker.exists()
+    assert not Path("/usr/escape").exists()
+
+
+def test_host_files_and_environment_are_invisible(
+    tmp_path: Path, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    secret = tmp_path / "secret.txt"
+    secret.write_text("hunter2")
+    monkeypatch.setenv("CODE_RUNNER_SECRET", "hunter2")
+    code = (
+        "import os\n"
+        "def f(a):\n"
+        f"    return [os.path.exists({str(secret)!r}), os.path.exists({str(Path.home())!r}),\n"
+        "            'CODE_RUNNER_SECRET' in os.environ, sorted(os.environ), os.getpid() < 20,\n"
+        "            os.path.exists('/proc/1/environ') and 'hunter2' in open('/proc/1/environ').read()]\n"
+    )
+    values = cast("list[Any]", _one(code).actual)
+    exists, home, env_leak, names, own_pidns, proc_leak = values
+    assert (exists, home, env_leak, own_pidns, proc_leak) == (
+        False,
+        False,
+        False,
+        True,
+        False,
+    )
+    assert set(names) <= {"HOME", "LANG", "PATH", "PWD", "PYTHONIOENCODING"}
+
+
+def test_symlink_in_scratch_cannot_exfiltrate_a_host_file(tmp_path: Path) -> None:
+    secret = tmp_path / "secret.txt"
+    secret.write_text("hunter2-secret-contents")
+    code = (
+        "import os\n"
+        "def f(a):\n"
+        f"    os.symlink({str(secret)!r}, '/work/result.json')\n"
+        "    return 7\n"
+    )
+    result = _one(code)
+    assert "hunter2" not in repr(result)
+    assert result.actual == 7  # the harness replaced the link, never followed it
+
+
+def test_fork_bomb_is_contained_and_leaves_nothing_behind() -> None:
+    before = _pgrep("python3 -I /work/harness.py")
+    code = (
+        "import os, subprocess\n"
+        "def f(a):\n"
+        "    subprocess.Popen(['sleep', '97'])\n"
+        "    while True:\n"
+        "        os.fork()\n"
+    )
+    started = time.monotonic()
+    result = _one(code)
+    assert time.monotonic() - started < 6
+    assert result.ok is False
+    assert result.error is not None
+    assert result.error.startswith(("TimeLimitExceeded", "BlockingIOError"))
+    time.sleep(0.3)
+    assert _pgrep("python3 -I /work/harness.py") == before
+    assert _pgrep("^sleep 97$") == 0
+
+
+def test_memory_bomb_is_a_visible_error() -> None:
+    result = _one("def f(a):\n    return len(bytearray(8 * 1024**3))\n")
+    assert result.ok is False
+    assert result.error is not None
+    assert result.error.startswith("MemoryError")
+
+
+@needs_node
+def test_node_heap_bomb_is_a_visible_error() -> None:
+    code = "var f = function(a) { const x = []; for (;;) x.push(new Array(1e6).fill(1)); };"
+    result = _one(code, "javascript", Limits(timeout_s=5.0))
+    assert result.ok is False
+    assert result.error in {"MemoryLimitExceeded", "RangeError: Invalid array length"}
+
+
+def test_output_flood_is_capped() -> None:
+    code = "def f(a):\n    while True:\n        print('x' * 1000)\n"
+    started = time.monotonic()
+    result = _one(code)
+    assert time.monotonic() - started < 6
+    assert result.ok is False
+    assert len(result.stdout) <= 64 * 1024
+
+
+def test_disk_fill_is_capped_by_the_file_size_limit() -> None:
+    code = (
+        "def f(a):\n"
+        "    with open('/work/big', 'wb') as fh:\n"
+        "        while True:\n"
+        "            fh.write(b'x' * 65536)\n"
+    )
+    result = _one(code)
+    assert result.ok is False
+    assert result.error is not None
+    assert "OSError" in result.error
+
+
+def test_timeout_kills_every_process_the_solution_started() -> None:
+    code = (
+        "import subprocess\n"
+        "def f(a):\n"
+        "    subprocess.Popen(['sleep', '97'])\n"
+        "    while True:\n"
+        "        pass\n"
+    )
+    result = _one(code)
+    assert result.error == "TimeLimitExceeded"
+    time.sleep(0.3)
+    assert _pgrep("^sleep 97$") == 0

code_runner/code_runner/tests/test_run.py

@@ -0,0 +1,227 @@
+"""End to end through the real sandbox: every language, every kind of outcome."""
+
+from __future__ import annotations
+
+import pytest
+
+from code_runner import CaseResult, Example, Limits, RunnerError, run
+from code_runner.tests.conftest import (
+    TWO_SUM_CODE,
+    TWO_SUM_STARTER,
+    needs_node,
+    needs_sandbox,
+)
+
+LANGS = [
+    pytest.param("python3", marks=needs_sandbox),
+    pytest.param("javascript", marks=needs_node),
+    pytest.param("typescript", marks=needs_node),
+]
+FAST = Limits(timeout_s=1.0)
+
+
+def _two_sum(lang: str, code: str | None, examples: list[Example]) -> list[CaseResult]:
+    return run(lang, code or TWO_SUM_CODE[lang], TWO_SUM_STARTER[lang], examples)
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_correct_solution_passes_every_example(
+    lang: str, two_sum_examples: list[Example]
+) -> None:
+    results = _two_sum(lang, None, two_sum_examples)
+    assert [r.ok for r in results] == [True, True, True]
+    assert [r.n for r in results] == [1, 2, 3]
+    first = results[0]
+    assert first.input == {"nums": [2, 7, 11, 15], "target": 9}
+    assert first.expected == [0, 1]
+    assert sorted(first.actual) == [0, 1]  # any order: [1, 0] is fine
+    assert first.error is None
+    assert first.ms >= 0
+
+
+WRONG = {
+    "python3": "class Solution:\n    def twoSum(self, nums, target):\n        return [0, 0]\n",
+    "javascript": "var twoSum = function(nums, target) { return [0, 0]; };",
+    "typescript": "function twoSum(nums: number[], target: number): number[] { return [0, 0]; }",
+}
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_wrong_answer_is_reported_with_actual_value(
+    lang: str, two_sum_examples: list[Example]
+) -> None:
+    results = _two_sum(lang, WRONG[lang], two_sum_examples)
+    assert [r.ok for r in results] == [False, False, False]
+    assert results[0].actual == [0, 0]
+    assert results[0].error is None
+
+
+THROWS = {
+    "python3": "class Solution:\n    def twoSum(self, nums, target):\n        return nums[99]\n",
+    "javascript": "var twoSum = function(nums, target) { throw new RangeError('boom'); };",
+    "typescript": "function twoSum(nums: number[], target: number): number[] { throw new RangeError('boom'); }",
+}
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_exception_becomes_an_error_per_case(
+    lang: str, two_sum_examples: list[Example]
+) -> None:
+    results = _two_sum(lang, THROWS[lang], two_sum_examples[:1])
+    only = results[0]
+    assert only.ok is False
+    assert only.actual is None
+    assert only.error is not None
+    assert only.error.split(":")[0] in {"IndexError", "RangeError"}
+
+
+BROKEN = {
+    "python3": "class Solution:\n    def twoSum(self, nums, target) return 1\n",
+    "javascript": "var twoSum = function(nums, target) { return ; ;; }}",
+    "typescript": "function twoSum(nums: number[], target: number): number[] { return [ }",
+}
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_syntax_error_is_reported(lang: str, two_sum_examples: list[Example]) -> None:
+    results = _two_sum(lang, BROKEN[lang], two_sum_examples[:2])
+    assert [r.ok for r in results] == [False, False]
+    assert all(r.error and r.error.startswith("SyntaxError") for r in results)
+
+
+PRINTS = {
+    "python3": "class Solution:\n    def twoSum(self, nums, target):\n        print('hi', target)\n        return [0, 1]\n",
+    "javascript": "var twoSum = function(nums, target) { console.log('hi', target); return [0, 1]; };",
+    "typescript": "function twoSum(nums: number[], target: number): number[] { console.log('hi', target); return [0, 1]; }",
+}
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_prints_land_in_stdout_per_case(
+    lang: str, two_sum_examples: list[Example]
+) -> None:
+    results = _two_sum(lang, PRINTS[lang], two_sum_examples[:2])
+    assert results[0].stdout.strip() == "hi 9"
+    assert results[1].stdout.strip() == "hi 6"
+
+
+@pytest.mark.parametrize("lang", LANGS)
+def test_missing_entry_point_in_code_is_a_case_error(
+    lang: str, two_sum_examples: list[Example]
+) -> None:
+    code = {
+        "python3": "x = 1\n",
+        "javascript": "var other = 1;",
+        "typescript": "const other = 1;",
+    }
+    results = _two_sum(lang, code[lang], two_sum_examples[:1])
+    assert results[0].ok is False
+    assert results[0].error is not None
+    assert "twoSum" in results[0].error or "Solution" in results[0].error
+
+
+def test_class_style_javascript_and_function_style_python() -> None:
+    js = run(
+        "javascript",
+        "class Solution { solve(a, b) { return a + b; } }",
+        "class Solution {\n  solve(a, b) {\n  }\n}",
+        [Example(1, {"a": 1, "b": 2}, 3)],
+    )
+    py = run(
+        "python3",
+        "def add(a, b):\n    return a + b\n",
+        "def add(a: int, b: int) -> int:\n",
+        [Example(1, {"a": 1, "b": 2}, 3)],
+    )
+    assert js[0].ok is True
+    assert py[0].ok is True
+
+
+def test_python_prelude_and_aliases() -> None:
+    code = "class Solution:\n    def f(self, xs: List[int]) -> int:\n        return collections.Counter(xs)[1]\n"
+    starter = "class Solution:\n    def f(self, xs: List[int]) -> int:\n"
+    out = run("python", code, starter, [Example(1, {"xs": [1, 1, 2]}, 2)])
+    assert out[0].ok is True
+
+
+def test_positional_fallback_when_statement_names_differ() -> None:
+    out = run(
+        "python3",
+        "def add(a, b):\n    return a - b\n",
+        "def add(a, b):\n",
+        [Example(1, {"x": 5, "y": 2}, 3)],
+    )
+    assert out[0].ok is True
+
+
+def test_argument_count_mismatch_is_a_case_error_without_running() -> None:
+    out = run(
+        "python3",
+        "def add(a, b):\n    return a\n",
+        "def add(a, b):\n",
+        [Example(1, {"only": 5}, 3)],
+    )
+    assert out[0].ok is False
+    assert out[0].error is not None
+    assert out[0].error.startswith("ArgumentError")
+
+
+def test_non_json_result_is_a_case_error() -> None:
+    out = run(
+        "python3",
+        "def f(a):\n    return {1, 2}\n",
+        "def f(a):\n",
+        [Example(1, {"a": 1}, [1, 2])],
+    )
+    assert out[0].error is not None
+    assert out[0].error.startswith("TypeError")
+
+
+def test_deep_order_and_float_tolerance() -> None:
+    out = run(
+        "python3",
+        "def f(a):\n    return [[2, 1], [0, -1]]\n",
+        "def f(a):\n",
+        [Example(1, {"a": 0}, [[-1, 0], [1, 2]], "deep")],
+    )
+    assert out[0].ok is True
+
+
+def test_timeout_is_per_case() -> None:
+    out = run(
+        "python3",
+        "def f(a):\n    while a:\n        pass\n    return 1\n",
+        "def f(a):\n",
+        [Example(1, {"a": 1}, 1), Example(2, {"a": 0}, 1)],
+        limits=FAST,
+    )
+    assert out[0].error == "TimeLimitExceeded"
+    assert out[0].ok is False
+    assert out[0].ms >= 900
+    assert out[1].ok is True
+
+
+def test_unsupported_language_and_empty_examples() -> None:
+    with pytest.raises(RunnerError, match="language"):
+        run("cobol", "", "", [])
+    assert run("python3", "def f(a): pass\n", "def f(a):\n", []) == []
+
+
+@needs_sandbox
+def test_json_bomb_result_costs_that_example_only() -> None:
+    code = (
+        "import os\n"
+        "def f(a):\n"
+        "    if a:\n"
+        "        open('/work/result.json', 'w').write('[' * 200000)\n"
+        "        os._exit(0)\n"
+        "    return 1\n"
+    )
+    out = run(
+        "python3",
+        code,
+        "def f(a):\n",
+        [Example(1, {"a": 1}, 1), Example(2, {"a": 0}, 1)],
+    )
+    assert (out[0].ok, out[0].error) == (False, "Crashed (exit 0)")
+    assert out[1].ok is True

code_runner/code_runner/tests/test_starter.py

@@ -0,0 +1,78 @@
+import pytest
+
+from code_runner._model import RunnerError
+from code_runner._starter import Entry, extract_starter, parse_starter
+
+PY_CLASS = """class Solution:
+    def twoSum(self, nums: list[int], target: int) -> list[int]:
+"""
+PY_COMMENTED = """# Definition for singly-linked list.
+# class ListNode:
+#     def __init__(self, val=0, next=None):
+class Solution:
+    def solve(self, grid: Dict[str, Tuple[int, int]], k: int = 3, *rest) -> int:
+        pass
+"""
+PY_FUNC = "def add(a: int, b: int = 2) -> int:\n    pass\n"
+JS_VAR = """/**
+ * @param {number[]} nums
+ */
+var twoSum = function(nums, target) {
+
+};"""
+JS_FUNC = "function isPalindrome(x, y = 3) {\n}\n"
+JS_ARROW = "const f = (a, b) => {\n}\n"
+JS_CLASS = "class Solution {\n  constructor() {}\n  solve(nums, ...rest) {\n  }\n}\n"
+TS_FUNC = "function twoSum(nums: number[], target: number): number[] {\n};"
+TS_ARROW = "const f = (a: Map<string, number>, b?: number): number => {\n}\n"
+
+
+@pytest.mark.parametrize(
+    ("lang", "text", "want"),
+    [
+        ("python3", PY_CLASS, Entry("Solution", "twoSum", ("nums", "target"))),
+        ("python3", PY_COMMENTED, Entry("Solution", "solve", ("grid", "k"))),
+        ("python3", PY_FUNC, Entry(None, "add", ("a", "b"))),
+        ("javascript", JS_VAR, Entry(None, "twoSum", ("nums", "target"))),
+        ("javascript", JS_FUNC, Entry(None, "isPalindrome", ("x", "y"))),
+        ("javascript", JS_ARROW, Entry(None, "f", ("a", "b"))),
+        ("javascript", JS_CLASS, Entry("Solution", "solve", ("nums",))),
+        ("typescript", TS_FUNC, Entry(None, "twoSum", ("nums", "target"))),
+        ("typescript", TS_ARROW, Entry(None, "f", ("a", "b"))),
+    ],
+)
+def test_parse_starter(lang: str, text: str, want: Entry) -> None:
+    assert parse_starter(lang, text) == want
+
+
+@pytest.mark.parametrize(
+    ("lang", "text"),
+    [
+        ("python3", "x = 1\n"),
+        ("python3", "class Solution:\n    pass\n"),
+        ("javascript", "// nothing here\n"),
+        ("javascript", "class Solution {\n  constructor() {}\n}\n"),
+        ("typescript", ""),
+    ],
+)
+def test_starter_without_entry_is_refused(lang: str, text: str) -> None:
+    with pytest.raises(RunnerError, match="entry point"):
+        parse_starter(lang, text)
+
+
+def test_extract_starter_takes_the_fenced_block() -> None:
+    statement = (
+        "# 1. Two Sum (Easy)\n\nbody\n\n## Starter code (python3)\n```\n"
+        + PY_CLASS
+        + "```\n\n## Site hints\nnope\n"
+    )
+    assert extract_starter(statement) == PY_CLASS
+
+
+def test_extract_starter_without_section_is_empty() -> None:
+    assert extract_starter("# 1. Two Sum\n\nno starter\n") == ""
+
+
+def test_extract_starter_unfenced_runs_to_next_section() -> None:
+    text = "intro\n## Starter code (javascript)\nvar f = function(a) {};\n## Hints\nx\n"
+    assert extract_starter(text) == "var f = function(a) {};\n"

code_runner/install.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+# Editable user install so /usr/bin/python3 (the tutor server, the CLI) can
+# `import code_runner`. Stdlib only; idempotent.
+set -euo pipefail
+
+REPO_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+readonly REPO_DIR
+readonly SYSTEM_PYTHON="/usr/bin/python3"
+
+"$SYSTEM_PYTHON" -m pip install --user --no-deps --break-system-packages -e "$REPO_DIR"
+cd /
+"$SYSTEM_PYTHON" -c 'import code_runner; print(code_runner.__file__)'

code_runner/pyproject.toml

@@ -3,17 +3,18 @@ requires = ["setuptools>=61"]
 build-backend = "setuptools.build_meta"
 
 [project]
-name = "feedback-inbox"
+name = "code_runner"
-version = "0.1.0"
+version = "0.1.1"
-description = "Turns every game dev-feedback bundle into one todo-app task"
+description = "Run a learner solution (Python, JavaScript, TypeScript) against a problem's examples in a bubblewrap sandbox"
 requires-python = ">=3.14"
+# Stdlib only on purpose: the tutor server imports it from the system python.
 dependencies = []
 
-[project.scripts]
-feedback-inbox = "feedback_inbox._cli:main"
-
 [tool.setuptools.packages.find]
-include = ["feedback_inbox*"]
+include = ["code_runner*"]
+
+[tool.setuptools.package-data]
+code_runner = ["py.typed", "harness/*"]
 
 [tool.ruff]
 target-version = "py314"
@@ -23,23 +24,35 @@ exclude = [".git", ".venv", "__pycache__", "build", "dist", ".eggs"]
 [tool.ruff.lint]
 select = ["ALL"]
 ignore = [
+    "CPY001",
+    "D100",
+    "D101",
+    "D102",
+    "D103",
+    "D104",
+    "D105",
+    "D107",
     "D203",
     "D213",
     "COM812",
     "ISC001",
     "S603",
-    # Stamps are local wall-clock names the games write; naive on purpose.
-    "DTZ001",
-    "DTZ007",
-    # The todo server is loopback http by design (MCP_URL), never user input.
-    "S310",
 ]
 fixable = ["ALL"]
 unfixable = []
 
 [tool.ruff.lint.per-file-ignores]
-"**/tests/**/*.py" = ["ARG", "D", "PLC0415", "N802", "PLR0913", "PLR2004", "S101", "SLF001"]
+"scripts/*.py" = ["INP001"]
-"**/test_*.py" = ["ARG", "D", "PLC0415", "N802", "PLR0913", "PLR2004", "S101", "SLF001"]
+# The sandbox-side harness is a program of its own: it reads its job from files.
+# The sandbox mounts its own private tmpfs at /tmp.
+"**/_sandbox.py" = ["S108"]
+"**/harness/*.py" = ["INP001", "T201", "S102", "BLE001", "PTH123"]
+# Tests embed the learner programs they run as source strings.
+"**/tests/**/*.py" = ["ARG", "D", "E501", "PLC0415", "PLR2004", "S101", "SLF001"]
+"**/test_*.py" = ["ARG", "D", "E501", "PLC0415", "PLR2004", "S101", "SLF001"]
+
+[tool.ruff.lint.pylint]
+max-args = 6
 
 [tool.ruff.lint.pydocstyle]
 convention = "google"
@@ -47,7 +60,7 @@ convention = "google"
 [tool.ruff.lint.isort]
 force-single-line = false
 force-sort-within-sections = true
-known-first-party = ["feedback_inbox"]
+known-first-party = ["code_runner"]
 
 [tool.ruff.lint.flake8-quotes]
 docstring-quotes = "double"
@@ -91,7 +104,7 @@ exclude = [".venv/"]
 [tool.pylint.main]
 # The repo root on the path, so pylint resolves the package's own imports the
 # same way pytest does instead of reporting import-error for them.
-init-hook = "import sys; sys.path.insert(0, 'feedback_inbox')"
+init-hook = "import sys; sys.path.insert(0, 'code_runner')"
 analyse-fallback-blocks = true
 persistent = true
 jobs = 0
@@ -104,14 +117,21 @@ ignore = [".venv", "__pycache__"]
 # to be `ignore-paths`, not `ignore = ["tests"]`: `ignore` only matches the
 # basename of what pylint is *given*, and pre-commit hands it explicit file
 # paths, so a directory name in `ignore` never fires under the hook.
-ignore-paths = ["^(.*/)?tests/.*$"]
+# harness/ runs inside the sandbox and must exec the solution it is given.
+ignore-paths = ["^(.*/)?tests/.*$", "^(.*/)?harness/.*$"]
 ignore-patterns = [".*\\.pyi$"]
 
 [tool.pylint.messages_control]
 enable = "all"
-disable = []
+disable = [
+    "missing-module-docstring",
+    "missing-class-docstring",
+    "missing-function-docstring",
+    "use-implicit-booleaness-not-comparison-to-zero",
+]
 
 [tool.pylint.design]
+max-args = 6
 min-public-methods = 0
 max-module-lines = 1000
 max-attributes = 10
@@ -132,7 +152,8 @@ generated-members = [
 exclude_dirs = ["tests", ".venv"]
 
 [tool.pytest.ini_options]
-testpaths = ["feedback_inbox/tests"]
+markers = ["sandbox: needs bubblewrap with unprivileged user namespaces (skipped otherwise)"]
+testpaths = ["code_runner/tests"]
 python_files = ["test_*.py", "*_test.py"]
 python_classes = ["Test*"]
 python_functions = ["test_*"]
@@ -141,7 +162,7 @@ addopts = [
     "--strict-markers",
     "--strict-config",
     "-ra",
-    "--cov=feedback_inbox",
+    "--cov=code_runner",
     "--cov-branch",
     "--cov-report=term-missing",
     "--cov-report=lcov",
@@ -153,9 +174,11 @@ filterwarnings = [
 ]
 
 [tool.coverage.run]
-source = ["feedback_inbox"]
+source = ["code_runner"]
 branch = true
 omit = [
+    # Runs inside the sandbox in another process; behaviour is tested through run().
+    "*/harness/*",
     "*/__pycache__/*",
     "*/tests/*",
     "*/.venv/*",

code_runner/requirements.txt

@@ -7,4 +7,4 @@ pre-commit==4.7.0
 pylint==4.1.3
 pytest==9.2.0
 pytest-cov==7.1.0
-ruff==0.17.4
+ruff==0.17.3

crdt_sync_dart/lib/src/io_store.dart

@@ -7,11 +7,8 @@ import 'store.dart';
 /// Writes atomically -- to a per-process temp file, then `rename`d over the
 /// real path -- so a concurrent reader (e.g. a background sync isolate writing
 /// while the foreground app also writes) never observes a half-written file.
-/// The pid in the temp name keeps two processes from colliding; within one
+/// The pid in the temp name keeps two writers from colliding. This generalizes
-/// instance, writes are serialized in issue order (last issued wins), so
+/// diet_guard's hand-rolled `LogStorageService` write scheme.
-/// overlapping calls (e.g. per-keystroke saves) never share the temp file
-/// mid-flight. This generalizes diet_guard's hand-rolled `LogStorageService`
-/// write scheme.
 ///
 /// Lives behind the `package:crdt_sync/crdt_sync_io.dart` entrypoint, not the
 /// main barrel, because it imports `dart:io` and so is unavailable on web --
@@ -23,10 +20,6 @@ class FileLogPersistence implements LogPersistence {
 
   final File _file;
 
-  /// Completes when the most recently issued write has settled. Never
-  /// rejects, so one failed write cannot block the ones queued after it.
-  Future<void> _tail = Future<void>.value();
-
   @override
   Future<String?> read() async {
     if (!_file.existsSync()) return null;
@@ -38,14 +31,7 @@ class FileLogPersistence implements LogPersistence {
   }
 
   @override
-  Future<void> write(String text) {
+  Future<void> write(String text) async {
-    final result = _tail.then((_) => _writeNow(text));
-    // The caller still sees its own error via [result]; the chain ignores it.
-    _tail = result.then((_) {}, onError: (Object _) {});
-    return result;
-  }
-
-  Future<void> _writeNow(String text) async {
     await _file.parent.create(recursive: true);
     final tmp = File('${_file.path}.$pid.tmp');
     await tmp.writeAsString(text);

crdt_sync_dart/pubspec.yaml

@@ -2,7 +2,7 @@ name: crdt_sync
 description: >-
   Shared CRDT merge scheme + GitHub-Contents-API sync transport for
   cross-device apps. Pure Dart, no Flutter dependency.
-version: 0.12.1
+version: 0.12.0
 publish_to: 'none'
 
 environment:

crdt_sync_dart/test/default_client_and_io_test.dart

@@ -128,32 +128,5 @@ void main() {
       await store.write('{"hello":"world"}');
       expect(await store.read(), '{"hello":"world"}');
     });
-
-    test('50 overlapping writes all succeed and the last issued wins',
-        () async {
-      // Regression: every call shared one `<file>.<pid>.tmp`, so the first
-      // rename moved it away and a later rename threw PathNotFoundException.
-      final store = FileLogPersistence(File('${dir.path}/log.json'));
-      await Future.wait([
-        for (var i = 0; i < 50; i++) store.write('{"n":$i}'),
-      ]);
-      expect(await store.read(), '{"n":49}');
-      expect(dir.listSync().map((e) => e.uri.pathSegments.last),
-          ['log.json']);
-    });
-
-    test('a failed write rejects its caller but does not block later writes',
-        () async {
-      final target = File('${dir.path}/log.json');
-      // A directory at the target path makes the rename fail.
-      Directory(target.path).createSync();
-      final store = FileLogPersistence(target);
-      await expectLater(
-          store.write('{"n":0}'), throwsA(isA<FileSystemException>()));
-
-      Directory(target.path).deleteSync();
-      await store.write('{"n":1}');
-      expect(await store.read(), '{"n":1}');
-    });
   });
 }

crdt_sync_flutter/pubspec.yaml

@@ -4,7 +4,7 @@ description: >-
   persisted per-install device id, and a one-call bootstrap. Removes the
   ~150 lines of firebase_backend.dart every consuming app was copying.
 publish_to: 'none'
-version: 0.5.1
+version: 0.5.0
 
 environment:
   sdk: ^3.12.2
@@ -13,7 +13,7 @@ dependencies:
   crdt_sync:
     git:
       url: https://github.com/kuhyx/utils
-      ref: crdt_sync_dart-v0.12.1
+      ref: crdt_sync_dart-v0.12.0
       path: crdt_sync_dart
   crypto: 3.0.8
   flutter:

data_transfer_flutter/pubspec.yaml

@@ -4,13 +4,13 @@ description: >-
   Flutter apps: system Save-as on Android/iOS (never cloud-only), a save
   dialog on desktop, a download on web, plus a versioned export envelope.
 publish_to: 'none'
-version: 0.2.2
+version: 0.2.1
 
 environment:
   sdk: ^3.12.2
 
 dependencies:
-  file_picker: 13.3.3
+  file_picker: 13.3.2
   file_selector: 1.1.0
   flutter:
     sdk: flutter

dep_freshness/_tables.py

@@ -91,8 +91,6 @@ NPM_ACCEPT: Final = "application/vnd.npm.install-v1+json"
 # --- Cache ------------------------------------------------------------------
 
 CACHE_PATH_ENV: Final = "DEP_FRESHNESS_CACHE"
-# Seconds; overrides every ecosystem TTL (a repo that wants a once-a-day gate).
-TTL_ENV: Final = "DEP_FRESHNESS_TTL_SECONDS"
 DEFAULT_CACHE_DIR: Final = "~/.cache/dep-freshness"
 CACHE_FILE: Final = "registry.json"
 TTL_SECONDS: Final = 6 * 3600

dep_freshness/cache.py

@@ -21,7 +21,6 @@ from dep_freshness._tables import (
     DEFAULT_CACHE_DIR,
     GITCOMMIT,
     GITTAG,
-    TTL_ENV,
     TTL_GITTAG_SECONDS,
     TTL_SECONDS,
 )
@@ -34,14 +33,7 @@ def cache_dir() -> Path:
 
 
 def ttl_for(ecosystem: str) -> int:
-    """Git refs move rarely and cost a subprocess, so they cache longer.
+    """Git refs move rarely and cost a subprocess, so they cache longer."""
-
-    `DEP_FRESHNESS_TTL_SECONDS` overrides both; garbage is ignored, because a
-    typo in an env var must not break a commit hook.
-    """
-    override = os.environ.get(TTL_ENV, "")
-    if override.isdigit() and int(override) > 0:
-        return int(override)
     return TTL_GITTAG_SECONDS if ecosystem in (GITTAG, GITCOMMIT) else TTL_SECONDS
 
 

dep_freshness/tests/test_resolve.py

@@ -12,7 +12,6 @@ from pathlib import Path
 
 import pytest
 
-from dep_freshness._tables import TTL_SECONDS
 from dep_freshness.cache import Cache, Entry, ttl_for
 from dep_freshness.models import Dep
 from dep_freshness.registries import http
@@ -140,16 +139,3 @@ def _offline(*_args):
 
 def _explode(*_args):
     raise RuntimeError("boom")
-
-
-def test_ttl_env_overrides_the_registry_ttl(monkeypatch) -> None:
-    """A repo can stretch the refresh window (once-a-day pre-push gates)."""
-    monkeypatch.setenv("DEP_FRESHNESS_TTL_SECONDS", "86400")
-    assert ttl_for("npm") == 86400
-    assert ttl_for("gittag") == 86400
-
-
-def test_ttl_env_garbage_falls_back_to_the_default(monkeypatch) -> None:
-    """A malformed value must never break a commit hook."""
-    monkeypatch.setenv("DEP_FRESHNESS_TTL_SECONDS", "soon")
-    assert ttl_for("npm") == TTL_SECONDS

dopamine-ux/DOCS-dopamine-ux-index.md

@@ -0,0 +1,165 @@
+# Dopamine UX — session prompts
+
+Ten pasteable prompts derived from *"The Dopamine Architecture: How UX Designers
+Hold User Attention in 2026"* (london-post.co.uk), applied to kuhy's repos.
+
+Paste **one file's contents** into a fresh Claude session. Each is
+self-contained: it restates the findings it depends on, so no session needs the
+context of another.
+
+## The two tests — every change must pass both
+
+The article's four mechanics (variable rewards, micro-animations, sound design,
+near-miss) are in scope, aimed at **kuhy's own stated goals**. They are gated by
+two questions, not by an allow/deny list:
+
+1. **Is the displayed state true?** A progress ring at 93% is legitimate *iff*
+   you are actually at 93%. A streak is legitimate iff it counts real days.
+   Dopamine may come from framing, timing and salience — **never from a
+   fabricated number**.
+2. **Is the variance real?** A random pick from a real pool (next workout, next
+   task) is legitimate. Injected random bonus points to manufacture uncertainty
+   are not.
+
+### The load-bearing corollary
+
+**Reward the recording, never the number recorded.** In diet-guard the
+satisfying moment fires on *logging a meal*, not on hitting a calorie target —
+otherwise the app rewards eating less rather than recording honestly, and
+corrupts its own data. Same shape in workout_app (reward logging a workout, not
+beating a PB) and todo (reward capture, not completion count).
+
+## Running order
+
+`01` gates everything. `07` gates `08`. Everything else is independent.
+
+| # | File | Repo | Findings |
+|---|---|---|---|
+| 01 | `01-motion-tokens.md` | `~/src/utils` | 3 |
+| 02 | `02-structural-check.md` | `~/src/utils` | 2 |
+| 03 | todo note `85ca329b-9809-412c-a47f-eddf055322cf` | `~/src/diet-guard` | 4 |
+| 04 | todo note `73a20837-6480-4b31-b966-1898a4e7272e` | `~/src/screen-locker` | 3 |
+| 05 | todo note `e50d6a03-ae7d-4da5-bbd3-c33b17f6f074` | `~/src/screen-locker/stronglift_replacement/workout_app` | 4 |
+| 06 | `06-todo.md` | `~/src/todo` | 3 |
+| 07 | `07-dufs-theme.md` | `~/src/dufs-cloud` | 2 |
+| 08 | todo note `e301e090-b2ee-4e26-8387-00f356b63521` | `~/src/dufs-cloud` | 4 |
+| 09 | todo note `9e1e4922-9f29-4bad-bcf4-96826aa8ffa4` | `~/src/wake-alarm` | 2 |
+
+**27 ranked findings total.**
+
+### Dependency graph
+
+```
+01-motion-tokens ──┬─→ 03, 04, 05, 09   (transcribe tokens locally)
+                   ├─→ 06               (bump design_system-v0.2.0)
+                   └─→ 07 ─→ 08         (bump web_ui-v0.3.2)
+02-structural-check  (independent; run after 01 so it has motion tokens to check)
+```
+
+## The top three findings fleet-wide
+
+Ranked by value-over-effort from the survey data:
+
+1. **screen-locker streak render** (prompt 04) — `snapshot.streak` is already
+   populated and no renderer reads it. Zero new state, zero conftest change,
+   zero new dependency.
+2. **diet-guard `_onLogMeal` feedback** (prompt 03) — the flagship case.
+   Dependency-free via `HapticFeedback` from `flutter/services.dart`.
+3. **workout_app finish-workout cue** (prompt 05) — `audioplayers`, `vibration`,
+   the `assets:` block and the VIBRATE permission are **already present**. Only
+   the call is missing.
+
+## Distributed copies, and the completion protocol
+
+Each prompt is **also copied into the repo it acts on**, at
+`<repo>/prompts/dopamine-ux-*.md`, so you can open a repo and say
+"do dopamine-ux-<name>" without going hunting. **This directory stays the source
+of truth** — if a prompt needs correcting, fix it here and re-distribute.
+
+| Prompt | Distributed to |
+|---|---|
+| 01 | `~/src/utils/prompts/dopamine-ux-01-motion-tokens.md` — deleted (done) |
+| 02 | `~/src/utils/prompts/dopamine-ux-02-structural-check.md` |
+| 03 | `~/src/diet-guard/prompts/dopamine-ux-diet-guard.md` |
+| 04 | `~/src/screen-locker/prompts/dopamine-ux-04-screen-locker.md` — deleted (done) |
+| 05 | `~/src/screen-locker/prompts/dopamine-ux-05-workout-app.md` — deleted (done) |
+| 06 | `~/src/todo/prompts/dopamine-ux-todo.md` |
+| 07 | `~/src/dufs-cloud/prompts/dopamine-ux-07-theme.md` |
+| 08 | `~/src/dufs-cloud/prompts/dopamine-ux-08-motion.md` |
+| 09 | `~/src/wake-alarm/prompts/dopamine-ux-wake-alarm.md` |
+
+When a session finishes a prompt it **deletes its repo copy** (in the same
+commit as the implementation) and **appends a row below**. A finished prompt left
+in a repo is indistinguishable from a pending one, and the next session re-runs
+it. Partial completion is never recorded as DONE — the file is edited to state
+what remains instead.
+
+### Completion log
+
+| Prompt | Status | Impl commit | Note |
+|---|---|---|---|
+| 01-motion-tokens.md | DONE 2026-08-22 | `f741896` | `motion.md` frozen; `AppDuration`/`AppCurve` + `--duration-*`/`--ease-*` shipped. Tags: `design_system-v0.2.0`, `web_ui-v0.3.2`. |
+| 02-structural-check.md | not started | — | — |
+| todo note `85ca329b-9809-412c-a47f-eddf055322cf` | not started | — | — |
+| todo note `73a20837-6480-4b31-b966-1898a4e7272e` | DONE 2026-11-01 | `89c65bd` | Status window `_section_streak` + week-transition reward text (kept in `extra_benefits_state.json`); the dead 09:00 early-bird claim is not shown. Per kuhy: no daily-streak state, no shutdown time, no freezes. Commit is on a worktree branch, unpushed. |
+| todo note `e50d6a03-ae7d-4da5-bbd3-c33b17f6f074` | DONE 2026-11-01 | `e938201` | Set/warmup `selectionClick`, finish `mediumImpact` + generated chime (music_theory, pinned tag) + dialog arrival, `AppDuration`/`AppCurve` in theme.dart, streak counts up, reduced motion honoured, one Settings "Sound" switch for chime and break-end (foreground service too). Worktree commits, unpushed; `62c1c2c` covers old export/import gaps. |
+| 06-todo.md | PARTIAL 2026-11-02 (step 1 DONE; step 2 sound REJECTED by kuhy) | `kuhyx/todo-app 2ffae26` | haptic + checkmark on capture, appRoute motion tokens, animated wizard bar; no sound by decision |
+| 07-dufs-theme.md | not started | — | — |
+| todo note `e301e090-b2ee-4e26-8387-00f356b63521` | not started | — | — |
+| todo note `9e1e4922-9f29-4bad-bcf4-96826aa8ffa4` | not started | — | — |
+
+## Keeping these prompts honest
+
+Every path and symbol cited across these files is machine-checked:
+
+```
+python3 ~/src/utils/dopamine-ux/verify_prompts.py
+```
+
+Exit 0 means every cited path resolves and every cited symbol is still greppable
+in the file it is attributed to. **Re-run it after any refactor of the surveyed
+repos** — a prompt pointing at a renamed function is worse than no prompt, and a
+path check alone cannot catch that (which is why it greps symbols too).
+
+Verified 2026-08-16: 146 paths, 30 symbols, all passing.
+
+## Rules that apply to every prompt
+
+1. **Tag cuts and version bumps are explicit steps.** `todo` consumes
+   `design_system` at git ref `design_system-v0.1.0`; `dufs-cloud/web` consumes
+   `@kuhyx/web-ui` at `web_ui-v0.3.1`. Adding tokens to those packages does
+   nothing for consumers until a new tag is cut. Prompt 01 ends by cutting
+   `design_system-v0.2.0` and `web_ui-v0.3.2`; prompts 06 and 08 begin by
+   bumping to those literal refs.
+2. **Anchor on symbols, not line numbers.** Line numbers in these prompts are
+   "as of 2026-08-16" and will drift. The symbol name is the real anchor —
+   grep for it rather than trusting the line.
+3. **Reduce-motion is mandatory and is NOT the sound toggle.** Web:
+   `@media (prefers-reduced-motion: reduce)`. Flutter:
+   `MediaQuery.of(context).disableAnimations`. The OS motion preference is
+   honoured automatically; the sound opt-out is a separate user-facing toggle.
+4. **Motion and haptics first, sound second.** Haptics need no dependency;
+   sound needs a pubspec dep, an `assets:` block, and a settings toggle. Each
+   prompt stages them separately so a short session still lands the valuable
+   half, and sound stays independently revertable.
+5. **Sound ships on (opt-out), never in screen-locker enforcement paths.**
+6. **250-line cap** applies to code and prose alike (`~/src/utils/file_length`).
+
+## Source
+
+Article findings, verbatim from the piece:
+
+- **Variable reward schedules** — unpredictable outcomes produce stronger
+  dopaminergic activity than guaranteed ones. Duolingo streaks, feed ordering,
+  pull-to-refresh.
+- **Micro-animations** — "a button that bounces on press, a like counter that
+  ticks upward with a slight delay". Replaces latency with feedback, so waiting
+  reads as responsiveness rather than frustration.
+- **Sound design** — "operates below the threshold of conscious attention";
+  the article's judgement is that it is more powerful than visual design and
+  gets less critical attention.
+- **Near-miss effects** — a fitness ring at ninety-three per cent creates
+  discomfort that drives re-engagement, and works *even when the user is
+  explicitly told the mechanism*.
+- Core claim: **dopamine is released in anticipation of a reward, not on
+  receipt. The uncertainty is the trigger.**

dopamine-ux/verify_prompts.py

@@ -0,0 +1,209 @@
+#!/usr/bin/env python3
+"""Verify every path and symbol cited in the dopamine-ux prompt files.
+
+A copy-pasteable prompt that points at a file which does not exist is worse than
+no prompt, so this gate adjudicates with an exit code rather than by eye.
+
+Two checks:
+  1. Every cited on-disk path resolves.
+  2. Every cited symbol is greppable in the file it is attributed to. A path
+     check alone cannot catch a stale line number; a symbol grep can.
+"""
+
+from __future__ import annotations
+
+import re
+import sys
+from pathlib import Path
+
+PROMPT_DIR = Path.home() / "src/utils" / "dopamine-ux"
+HOME = Path.home()
+
+PATH_RE = re.compile(r"`(~?/[^`\s]+|[\w.\-]+/[\w./\-]+\.\w+)`")
+
+# Not on-disk paths despite matching the shape above.
+SKIP_SUBSTRINGS = (
+    "://",
+    "$",
+    "*",
+    "<",
+    ">",
+    "…",
+    "devices/",  # Firebase/crdt key path
+    "settings/advanced",  # Firebase key path
+    "android.permission",
+    "android.intent",
+    "/sdcard/",
+    "/etc/",
+    "/dev/input/",
+    "/opt/",
+    "com.kuhy",
+    "app-release.apk",
+    "build/app/outputs",  # build outputs
+    "flutter/services.dart",
+    "flutter/material.dart",  # package imports
+)
+
+# Files a prompt CREATES. Citing them forward is correct, so absence is expected
+# -- but pairing each with its creating prompt still catches a typo.
+CREATED_BY_PROMPT = {
+    "~/src/utils/unified-design-system/motion.md": "01",
+    "~/src/utils/unified-design-system/scripts/structural_check.py": "02",
+}
+
+# Cited precisely to state that they do NOT exist.
+ASSERTED_ABSENT = {
+    "~/gatelock",
+    # Distributed prompt copies are `git rm`d on completion (see the protocol in
+    # 00-INDEX.md), but 00-INDEX.md keeps citing them in its "Distributed copies"
+    # table. Asserting ABSENT rather than dropping the row keeps the check
+    # two-directional: a finished prompt reappearing in a repo is itself a
+    # failure, because the next session would re-run it. Move a row here as each
+    # prompt completes.
+    "~/src/utils/prompts/dopamine-ux-01-motion-tokens.md",
+}
+
+# Generic prose fragments rather than citations of one specific file.
+PROSE_FRAGMENTS = {"lib/ui/theme.dart", "lib/main.dart", "pubspec.yaml"}
+
+# Symbols each prompt claims exist, as (repo-relative file, symbol).
+SYMBOL_CLAIMS = [
+    ("~/src/diet-guard/app/lib/screens/log_meal_screen.dart", "_onLogMeal"),
+    ("~/src/diet-guard/app/lib/widgets/today_progress_card.dart", "TodayProgressCard"),
+    ("~/src/diet-guard/app/lib/widgets/streak_summary_row.dart", "StreakSummaryRow"),
+    ("~/src/diet-guard/app/lib/screens/log_meal_progress.dart", "buildTodayProgress"),
+    ("~/src/diet-guard/app/lib/services/app_settings_service.dart", "_writeToDisk"),
+    ("~/src/screen-locker/screen_locker/_status_types.py", "StatusSnapshot"),
+    ("~/src/screen-locker/screen_locker/_extra_benefits.py", "process_week_transition"),
+    ("~/src/screen-locker/screen_locker/_extra_benefits.py", "current_streak"),
+    ("~/src/screen-locker/screen_locker/_unlock_view.py", "unlock_screen"),
+    ("~/src/screen-locker/screen_locker/_log_mixin.py", "write_signed_entry"),
+    ("~/src/screen-locker/screen_locker/tests/conftest.py", "_ISOLATED_STATE"),
+    ("~/src/screen-locker/screen_locker/_weekly_check.py", "WEEKLY_WORKOUT_MINIMUM"),
+    (
+        "~/src/screen-locker/stronglift_replacement/workout_app/lib/screens/"
+        "workout_screen_session.dart",
+        "_playBreakEndCue",
+    ),
+    (
+        "~/src/screen-locker/stronglift_replacement/workout_app/lib/screens/"
+        "workout_screen_finish.dart",
+        "_persistFinishedWorkout",
+    ),
+    (
+        "~/src/screen-locker/stronglift_replacement/workout_app/lib/widgets/"
+        "workout_summary_dialog.dart",
+        "WorkoutSummaryDialog",
+    ),
+    (
+        "~/src/screen-locker/stronglift_replacement/workout_app/lib/services/"
+        "storage_service.dart",
+        "_getSetting",
+    ),
+    (
+        "~/src/screen-locker/stronglift_replacement/workout_app/lib/widgets/"
+        "exercise_tile_rows.dart",
+        "AnimatedContainer",
+    ),
+    ("~/src/todo/lib/data/app_settings.dart", "withAdvancedMode"),
+    ("~/src/todo/lib/ui/settings_screen.dart", "SwitchListTile"),
+    ("~/src/dufs-cloud/web/src/hooks/use-cloud-index.ts", "walkInto"),
+    ("~/src/dufs-cloud/app/lib/services/cloud_index.dart", "buildCloudIndex"),
+    ("~/src/dufs-cloud/web/src/lib/download.ts", "buildSelectionZip"),
+    ("~/src/dufs-cloud/app/lib/services/download_zip.dart", "buildSelectionZip"),
+    ("~/src/wake-alarm/phone_app/lib/screens/home_screen.dart", "_setPhoneAlarm"),
+    ("~/src/wake-alarm/wake_alarm/_audio.py", "_play_on_all_sinks"),
+    ("~/src/utils/unified-design-system/scripts/palette_check.py", "NON_COLOUR_CSS"),
+    ("~/src/utils/unified-design-system/scripts/palette_map.py", "PALETTE"),
+    ("~/src/utils/design_system/lib/src/feedback.dart", "showToast"),
+    ("~/src/utils/design_system/lib/src/tokens.dart", "AppSpacing"),
+    ("~/src/utils/gatelock/gatelock/_window.py", "LockConfig"),
+]
+
+
+def expand(raw: str) -> Path:
+    return HOME / raw[2:] if raw.startswith("~/") else Path(raw)
+
+
+def candidate_paths(text: str) -> set[str]:
+    return {
+        raw
+        for raw in PATH_RE.findall(text)
+        if not any(s in raw for s in SKIP_SUBSTRINGS)
+        and raw not in PROSE_FRAGMENTS
+        and not re.search(r"\.\w+:\d+$", raw)  # trailing :line suffix
+    }
+
+
+def prompt_roots(text: str) -> list[Path]:
+    roots = []
+    for label in ("Repo", "App", "Flutter app"):
+        for m in re.finditer(rf"{label}: `(~/[^`]+)`", text):
+            roots.append(expand(m.group(1)))
+    return roots
+
+
+def check_paths() -> list[str]:
+    failures = []
+    checked = 0
+    for prompt in sorted(PROMPT_DIR.glob("*.md")):
+        text = prompt.read_text()
+        roots = prompt_roots(text)
+        for raw in sorted(candidate_paths(text)):
+            if raw in ASSERTED_ABSENT:
+                if expand(raw).exists():
+                    failures.append(
+                        f"{prompt.name}: `{raw}` is asserted ABSENT but EXISTS"
+                    )
+                continue
+            if raw in CREATED_BY_PROMPT:
+                owner = CREATED_BY_PROMPT[raw]
+                cites_owner = re.search(rf"[Pp]rompt {owner}", text) is not None
+                if not prompt.name.startswith(owner) and not cites_owner:
+                    failures.append(
+                        f"{prompt.name}: cites `{raw}` without noting prompt "
+                        f"{owner} creates it"
+                    )
+                continue
+
+            checked += 1
+            if raw.startswith(("~/", "/")):
+                if not expand(raw).exists():
+                    failures.append(f"{prompt.name}: `{raw}` MISSING")
+            elif not any((r / raw).exists() for r in roots):
+                failures.append(
+                    f"{prompt.name}: `{raw}` not found under any of "
+                    f"{[str(r) for r in roots]}"
+                )
+    print(f"paths: checked {checked}")
+    return failures
+
+
+def check_symbols() -> list[str]:
+    failures = []
+    for raw_file, symbol in SYMBOL_CLAIMS:
+        path = expand(raw_file)
+        if not path.exists():
+            failures.append(f"symbol source MISSING: {raw_file}")
+            continue
+        if symbol not in path.read_text(errors="replace"):
+            failures.append(f"symbol `{symbol}` NOT FOUND in {raw_file}")
+    print(f"symbols: checked {len(SYMBOL_CLAIMS)}")
+    return failures
+
+
+def main() -> int:
+    n_prompts = len(list(PROMPT_DIR.glob("*.md")))
+    print(f"verifying {n_prompts} prompt files in {PROMPT_DIR}")
+    failures = check_paths() + check_symbols()
+    if failures:
+        print(f"\nFAIL -- {len(failures)} problem(s):")
+        for f in failures:
+            print(f"  {f}")
+        return 1
+    print("\nPASS -- every cited path exists and every cited symbol is present")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

earned_time/README.md

@@ -199,7 +199,6 @@ their `max_gaming_minutes` + pure bonuses again (240/120 min = 2/min; full day =
 from the start day, cutting all of it, 180 with LeetCode's formerly free hour (0.20.0); maturity never delays it. Ledger `~/.local/share/focus_project/ledger.json`:
 signed `credit` rows, `entry_id` `<project>-<YYYYMMDD>-m<n>`, `detail.minutes` (positive int),
 `detail.ended_at` (else `day`). `focus_project` unusable = no project (logged).
-Unbacked timer time pays 0 from 2026-11-08 (focus_project 0.10.0).
 
 ## Adding a gate
 

earned_time/earned_time/tests/test_focus_backing.py

@@ -1,88 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Regression: a timer-only focus day after the anti-cheat cutover pays 0.
-
-Runs the REAL ``focus_project`` (the sibling package of this monorepo):
-timer rows -> harvester -> signed ledger -> ``credit_units``. earned_time
-itself does not change; this pins that the gate reaches pay.
-"""
-
-from __future__ import annotations
-
-from datetime import date, datetime
-import importlib
-import json
-from pathlib import Path
-import sys
-from typing import TYPE_CHECKING
-
-import pytest
-
-from earned_time import credit_units, earners_for
-
-if TYPE_CHECKING:
-    from types import ModuleType
-
-DAY = date(2026, 11, 9)  # after focus_project.BACKING_FROM
-_SIBLING = Path(__file__).resolve().parents[3] / "focus_project"
-
-
-def at(hour: int, minute: int) -> float:
-    return datetime(2026, 11, 9, hour, minute).astimezone().timestamp()
-
-
-@pytest.fixture
-def real_fp(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Path, key_file: Path
-) -> ModuleType:
-    """The real focus_project on redirected state (HOME, data dir, key)."""
-    monkeypatch.setenv("HOME", str(tmp_path / "home"))
-    monkeypatch.setenv("FOCUS_PROJECT_DATA", str(tmp_path / "data"))
-    monkeypatch.setenv("FOCUS_PROJECT_KEY", str(key_file))
-    monkeypatch.setenv("FOCUS_PROJECT_LEETCODE_LEDGER", str(tmp_path / "lc.json"))
-    monkeypatch.delitem(sys.modules, "focus_project")  # the conftest's fake
-    monkeypatch.syspath_prepend(str(_SIBLING))
-    fp = importlib.import_module("focus_project")
-    fp.create_project(
-        "hr",
-        name="HR",
-        start=DAY,
-        end=DAY,
-        replaces=["leetcode"],
-        daily_minutes=120,
-        unlock_minutes=30,
-    )
-    return fp
-
-
-def pay(fp: ModuleType, tmp_path: Path, key_file: Path, *rows: object) -> int | None:
-    log = tmp_path / "data" / "activity.jsonl"
-    log.write_text("".join(json.dumps(r) + "\n" for r in rows), encoding="utf-8")
-    harvest = importlib.import_module("focus_project.harvest")
-    assert harvest.harvest(now=at(23, 0)) == 0
-    (item,) = (e for e in earners_for(DAY) if e.name == "focus")
-    return credit_units(item, tmp_path / "data" / "ledger.json", key_file, DAY)
-
-
-def row(source: str, kind: str, ts: float) -> dict[str, object]:
-    return {"ts": ts, "source": source, "kind": kind, "project": "hr"}
-
-
-TIMER = (
-    row("timer", "start", at(9, 0)),
-    row("timer", "checkin", at(9, 5)),
-    row("timer", "stop", at(9, 10)),
-)
-
-
-def test_timer_only_day_pays_zero(
-    real_fp: ModuleType, tmp_path: Path, key_file: Path
-) -> None:
-    assert date(2026, 11, 8) == real_fp.BACKING_FROM
-    assert pay(real_fp, tmp_path, key_file, *TIMER) == 0
-
-
-def test_backed_timer_day_still_pays(
-    real_fp: ModuleType, tmp_path: Path, key_file: Path
-) -> None:
-    beat = row("tutor", "heartbeat", at(9, 5))
-    assert pay(real_fp, tmp_path, key_file, *TIMER, beat) == 10

feedback_inbox/README.md

@@ -1,70 +0,0 @@
-# feedback_inbox
-
-Turns every game's dev-feedback bundle into one todo-app task, so feedback
-kuhy saves with F12 is never forgotten.
-
-A game saves `~/data/<repo>_binaries/feedback/<stamp>/` (`report.json`,
-screenshots, logs) and appends a ready-to-run Claude prompt to
-`feedback-<date>.md` beside it. This package imports each complete bundle once
-as a task in the todo app for `kuhyx/<repo>`.
-
-## Task shape
-
-- Title: `MANUAL WORK REQUIRED: <repo> feedback: <first line of his note>` (note capped
-  at 120 characters). The prefix makes todoImplementer list it as a human part and never
-  run it. No note: `... feedback: <game> <combo> <HH:MM>`. A `report.json` that does
-  not parse still gets a task, tagged `[corrupt report.json]`; the note then
-  comes from the markdown entry.
-- Body: a `Due: <next day 10:00>` line, then the bundle's markdown entry (the
-  prompt with absolute paths).
-- The task is its own note (`add_task` with `new_note`). The server stores that as
-  `<repo> todos: <title>` plus a `## <title>` section, so right after adding, the inbox
-  rewrites it with `update_task` to the exact title and a section-free body (the entry's
-  own `##` headings become `###`). todoImplementer splits notes at `## ` headings, and a
-  title that does not start with `MANUAL WORK REQUIRED: ` would make a runnable part.
-  The marker records `format: 2` once the rewrite worked; a failed rewrite is retried on
-  the next sweep (never a second task), and markers without it are repaired the same way.
-- Closing the task cancels the feedback. "Act on feedback" in a Claude session
-  means working every open `<repo> feedback:` task.
-
-The todo app has no due-date or reminder field yet, so the due time is only
-text in the body.
-
-## How it runs
-
-| Unit | Does |
-|---|---|
-| `feedback-inbox.path` | `PathChanged` on every existing `*_binaries/feedback` folder; runs `feedback-inbox run` |
-| `feedback-inbox-sync.path` | `PathChanged` on `~/data` and every `*_binaries` folder; runs `sync` then `run` |
-| `feedback-inbox-sync.timer` | every 5 min, the same sweep (events during a run are lost by `.path` units) |
-
-`.path` units take no globs: `feedback-inbox sync` writes the folder list into
-the drop-ins `feedback-inbox{,-sync}.path.d/dirs.conf` and reloads systemd only
-when the set of games changed.
-
-A bundle is complete when `report.json` is at least 2 s old and its markdown
-entry exists (matched by the stamp folder in the entry's paths). Incomplete
-bundles are retried; `run` waits up to 15 s for them.
-
-The import marker is `.inbox-imported.json` inside the bundle (task id, title,
-due). It is written only after the todo server accepted the task, so a server
-outage leaves no marker, exits non-zero, and the next sweep retries.
-
-## Install
-
-```bash
-scripts/run_subproject_tests.sh feedback_inbox   # from the monorepo root
-feedback_inbox/install.sh
-```
-
-`install.sh` builds a stdlib-only venv, links `feedback-inbox` into
-`~/.local/bin`, mints a todo-server token into `~/.config/feedback-inbox/token`
-(0600; revoke by deleting its line in `~/.config/todo-mcp/tokens`) and enables
-the units. Enabling sweeps once, so existing bundles are imported then.
-
-## Tests
-
-`pytest` from this directory (100 % branch coverage enforced). Tests use a temp
-`HOME`, a fake todo client, and a local HTTP server for the MCP client; the
-autouse `conftest.py` redirects the one path resolver and points the todo URL
-at a dead port.

feedback_inbox/feedback_inbox/__init__.py

@@ -1,2 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Turns every game dev-feedback bundle into one todo-app task."""

feedback_inbox/feedback_inbox/__main__.py

@@ -1,15 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""``python -m feedback_inbox``."""
-
-from __future__ import annotations
-
-from feedback_inbox._cli import main
-
-
-def run() -> int:
-    """Run the CLI on ``sys.argv``."""
-    return main()
-
-
-if __name__ == "__main__":
-    raise SystemExit(run())

feedback_inbox/feedback_inbox/_bundle.py

@@ -1,57 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Finding games' feedback folders and bundles, and the stamp arithmetic."""
-
-from __future__ import annotations
-
-from datetime import datetime, timedelta
-import re
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-BINARIES_SUFFIX = "_binaries"
-STAMP_FORMAT = "%Y-%m-%dT%H-%M-%S"
-_STAMP = re.compile(r"^\d{4}-\d{2}-\d{2}T\d{2}-\d{2}-\d{2}-\d{3}$")
-_REMINDER_HOUR = 10
-
-
-def binaries_dirs(data_root: Path) -> list[Path]:
-    """Every ``<game>_binaries`` directory directly under ``data_root``."""
-    if not data_root.is_dir():
-        return []
-    return sorted(
-        p
-        for p in data_root.iterdir()
-        if p.name.endswith(BINARIES_SUFFIX) and p.is_dir()
-    )
-
-
-def feedback_dirs(data_root: Path) -> list[Path]:
-    """Every existing ``<game>_binaries/feedback`` directory."""
-    return [
-        d / "feedback" for d in binaries_dirs(data_root) if (d / "feedback").is_dir()
-    ]
-
-
-def repo_name(feedback_dir: Path) -> str:
-    """The game's repo name: ``cleanup-plate_binaries`` -> ``cleanup-plate``."""
-    return feedback_dir.parent.name.removesuffix(BINARIES_SUFFIX)
-
-
-def bundles_in(feedback_dir: Path) -> list[Path]:
-    """Bundle folders (named by a millisecond stamp), oldest first."""
-    return sorted(
-        p for p in feedback_dir.iterdir() if p.is_dir() and _STAMP.match(p.name)
-    )
-
-
-def parse_stamp(stamp: str) -> datetime:
-    """The local time a bundle was saved, from its folder name."""
-    return datetime.strptime(stamp[:19], STAMP_FORMAT)
-
-
-def due_at(stamp: str) -> datetime:
-    """The reminder time: the day after the bundle, at 10:00 local."""
-    day = parse_stamp(stamp) + timedelta(days=1)
-    return day.replace(hour=_REMINDER_HOUR, minute=0, second=0, microsecond=0)

feedback_inbox/feedback_inbox/_cli.py

@@ -1,80 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""``feedback-inbox run | sync``: what the systemd units execute."""
-
-from __future__ import annotations
-
-import argparse
-import fcntl
-import logging
-import time
-from typing import TYPE_CHECKING
-
-from feedback_inbox._inbox import Importer, Outcome, Waiter, run_settled
-from feedback_inbox._paths import resolve_paths
-from feedback_inbox._todo import McpTodoClient
-from feedback_inbox._units import run_systemctl, sync_units
-
-if TYPE_CHECKING:
-    from collections.abc import Callable, Sequence
-
-    from feedback_inbox._todo import TodoClient
-
-LOG = logging.getLogger("feedback_inbox")
-OWNER = "kuhyx"
-DEFAULT_WAIT = 15.0
-
-
-def _parser() -> argparse.ArgumentParser:
-    parser = argparse.ArgumentParser(prog="feedback-inbox")
-    sub = parser.add_subparsers(dest="command", required=True)
-    run = sub.add_parser("run", help="import new bundles as todo tasks")
-    run.add_argument("--wait", type=float, default=DEFAULT_WAIT)
-    sub.add_parser("sync", help="re-arm the path units for the current games")
-    return parser
-
-
-def _run(wait: float, client: TodoClient | None) -> int:
-    paths = resolve_paths()
-    todo = client or McpTodoClient(paths.mcp_url, paths.token_file)
-    paths.lock_file.parent.mkdir(parents=True, exist_ok=True)
-    # Blocking lock: a second run waits, then finds the markers and skips.
-    with paths.lock_file.open("a", encoding="utf-8") as lock:
-        fcntl.flock(lock, fcntl.LOCK_EX)
-        results = run_settled(
-            paths.data_root,
-            Importer(todo, OWNER),
-            Waiter(wait, time.time, time.sleep),
-        )
-    failed = 0
-    for result in results:
-        if result.outcome is Outcome.FAILED:
-            failed += 1
-            LOG.error("failed %s: %s", result.bundle, result.detail)
-        elif result.outcome is not Outcome.SKIPPED:
-            LOG.info("%s %s %s", result.outcome, result.bundle, result.task_id or "")
-    return 1 if failed else 0
-
-
-def _sync(systemctl: Callable[[Sequence[str]], int]) -> int:
-    paths = resolve_paths()
-    try:
-        changed = sync_units(paths.data_root, paths.unit_dir, systemctl)
-    except RuntimeError:
-        LOG.exception("sync failed")
-        return 1
-    LOG.info("watch list %s", "rewritten" if changed else "unchanged")
-    return 0
-
-
-def main(
-    argv: Sequence[str] | None = None,
-    *,
-    client: TodoClient | None = None,
-    systemctl: Callable[[Sequence[str]], int] = run_systemctl,
-) -> int:
-    """Entry point; ``client`` and ``systemctl`` are test seams."""
-    logging.basicConfig(level=logging.INFO, format="%(levelname)s %(message)s")
-    args = _parser().parse_args(argv)
-    if args.command == "run":
-        return _run(args.wait, client)
-    return _sync(systemctl)

feedback_inbox/feedback_inbox/_entry.py

@@ -1,74 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""A bundle's markdown entry and its report.json."""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-import json
-import re
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-_SECTION = re.compile(r"(?m)^(?=## )")
-_GAME = re.compile(r"Game `([^`]*)`, combo `([^`]*)`")
-_BODY_END = re.compile(r"(?m)^- Screenshot")
-
-
-@dataclass(frozen=True)
-class Report:
-    """The fields of report.json the inbox needs."""
-
-    note: str
-    game: str
-    combo: str
-
-
-def find_entry(feedback_dir: Path, bundle: Path) -> str | None:
-    """The ``feedback-<date>.md`` section whose paths point into ``bundle``.
-
-    Matched by the stamp folder in its paths, not by the heading time (which
-    drops the milliseconds), so a bundle never picks up a neighbour's entry.
-    """
-    needle = f"/{bundle.name}/"
-    for md in sorted(feedback_dir.glob("feedback-*.md")):
-        try:
-            text = md.read_text(encoding="utf-8")
-        except OSError, UnicodeDecodeError:
-            continue
-        for section in _SECTION.split(text):
-            if needle in section:
-                return section.strip()
-    return None
-
-
-def read_report(path: Path) -> Report | None:
-    """Parse report.json; ``None`` when it cannot be read as a JSON object."""
-    try:
-        data = json.loads(path.read_text(encoding="utf-8"))
-    except OSError, UnicodeDecodeError, json.JSONDecodeError:
-        return None
-    if not isinstance(data, dict):
-        return None
-    return Report(
-        note=_text(data.get("note")),
-        game=_text(data.get("game")),
-        combo=_text(data.get("combo")),
-    )
-
-
-def _text(value: object) -> str:
-    return value if isinstance(value, str) else ""
-
-
-def note_from_entry(entry: str) -> str:
-    """His note, as written between the heading and the file list."""
-    _head, _, rest = entry.partition("\n")
-    return _BODY_END.split(rest, maxsplit=1)[0].strip()
-
-
-def game_and_combo(entry: str) -> tuple[str, str]:
-    """Game and combo from the entry's ``Game ..., combo ...`` line."""
-    found = _GAME.search(entry)
-    return (found.group(1), found.group(2)) if found else ("", "")

feedback_inbox/feedback_inbox/_inbox.py

@@ -1,235 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Turn complete feedback bundles into todo tasks, once each."""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-from enum import StrEnum
-import json
-import logging
-import re
-from typing import TYPE_CHECKING
-
-from feedback_inbox._bundle import (
-    bundles_in,
-    due_at,
-    feedback_dirs,
-    parse_stamp,
-    repo_name,
-)
-from feedback_inbox._entry import (
-    Report,
-    find_entry,
-    game_and_combo,
-    note_from_entry,
-    read_report,
-)
-from feedback_inbox._todo import TodoError
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-    from pathlib import Path
-
-    from feedback_inbox._todo import TodoClient
-
-LOG = logging.getLogger("feedback_inbox")
-MARKER = ".inbox-imported.json"
-# Marker format 2: the stored note was rewritten to our exact title and body.
-FORMAT = 2
-SETTLE_SECONDS = 2.0
-TITLE_CAP = 120
-MANUAL_PREFIX = "MANUAL WORK REQUIRED: "
-CORRUPT_TAG = " [corrupt report.json]"
-CORRUPT_LINE = "Warning: report.json is corrupt; the note below comes from the entry."
-
-
-class Outcome(StrEnum):
-    """What happened to one bundle in a run."""
-
-    IMPORTED = "imported"
-    SKIPPED = "skipped"
-    PENDING = "pending"
-    FAILED = "failed"
-
-
-@dataclass(frozen=True)
-class Result:
-    """One bundle's outcome, for the journal and the exit code."""
-
-    bundle: Path
-    outcome: Outcome
-    task_id: str | None = None
-    detail: str = ""
-
-
-@dataclass(frozen=True)
-class TaskDraft:
-    """The task to create: title, body and due time."""
-
-    title: str
-    body: str
-    due: str
-
-
-def _demote(entry: str) -> str:
-    """Turn ``## `` headings into ``### ``: no section may start a todo part.
-
-    The todo app splits a note into runnable parts at every ``## `` heading,
-    so the entry's own headings must not look like sections.
-    """
-    return re.sub(r"(?m)^##(?=\s)", "###", entry)
-
-
-def _first_line(note: str, fallback: str) -> str:
-    first = next((x.strip() for x in note.splitlines() if x.strip()), "")
-    if len(first) > TITLE_CAP:
-        first = first[: TITLE_CAP - 1] + "…"
-    return first or fallback
-
-
-def build_task(repo: str, stamp: str, entry: str, report: Report | None) -> TaskDraft:
-    """Compose the title and body from the entry and (maybe corrupt) report."""
-    found_game, found_combo = game_and_combo(entry)
-    note = note_from_entry(entry) if report is None else report.note
-    game = (report.game if report else "") or found_game
-    combo = (report.combo if report else "") or found_combo
-    clock = parse_stamp(stamp).strftime("%H:%M")
-    fallback = " ".join(x for x in (game, combo, clock) if x)
-    suffix = CORRUPT_TAG if report is None else ""
-    title = f"{MANUAL_PREFIX}{repo} feedback: {_first_line(note, fallback)}{suffix}"
-    due = due_at(stamp)
-    lines = [f"Due: {due:%Y-%m-%d %H:%M}", ""]
-    if report is None:
-        lines += [CORRUPT_LINE, ""]
-    return TaskDraft(title, "\n".join([*lines, _demote(entry)]), due.isoformat())
-
-
-def _write_marker(bundle: Path, payload: dict[str, object]) -> None:
-    tmp = bundle / f"{MARKER}.tmp"
-    tmp.write_text(json.dumps(payload, indent=2) + "\n", encoding="utf-8")
-    tmp.replace(bundle / MARKER)
-
-
-def _read_marker(bundle: Path) -> dict[str, object] | None:
-    """The bundle's marker, or None when it is not a usable marker."""
-    try:
-        data = json.loads((bundle / MARKER).read_text(encoding="utf-8"))
-    except OSError, UnicodeDecodeError, json.JSONDecodeError:
-        return None
-    ok = isinstance(data, dict) and isinstance(data.get("task_id"), str)
-    return data if ok and isinstance(data.get("title"), str) else None
-
-
-@dataclass(frozen=True)
-class Importer:
-    """Who files the tasks (``client``) under which GitHub owner."""
-
-    client: TodoClient
-    owner: str
-    settle: float = SETTLE_SECONDS
-
-    def import_bundle(self, feedback_dir: Path, bundle: Path, now: float) -> Result:
-        """Import one bundle if complete; repair the title of one already filed."""
-        if (bundle / MARKER).exists():
-            return self._repair(feedback_dir, bundle)
-        report_path = bundle / "report.json"
-        entry = find_entry(feedback_dir, bundle)
-        if entry is None or not report_path.is_file():
-            return Result(bundle, Outcome.PENDING, detail="report or entry missing")
-        if now - report_path.stat().st_mtime < self.settle:
-            return Result(bundle, Outcome.PENDING, detail="report.json settling")
-        repo = repo_name(feedback_dir)
-        draft = build_task(repo, bundle.name, entry, read_report(report_path))
-        return self._file(bundle, f"{self.owner}/{repo}", draft)
-
-    def _file(self, bundle: Path, repo: str, draft: TaskDraft) -> Result:
-        """Create the task, write the marker, then fix the title."""
-        try:
-            ref = self.client.add_task(repo, draft.title, draft.body)
-        except TodoError as exc:
-            return Result(bundle, Outcome.FAILED, detail=str(exc))
-        # Marker first: once the task exists a retry must never add a second.
-        # A merged task belongs to another note, which is not ours to rewrite.
-        payload: dict[str, object] = {
-            "task_id": ref.task_id,
-            "title": draft.title,
-            "due": draft.due,
-            "merged": ref.merged,
-        }
-        if ref.merged:
-            payload["format"] = FORMAT
-        try:
-            _write_marker(bundle, payload)
-        except OSError as exc:
-            return Result(
-                bundle,
-                Outcome.FAILED,
-                ref.task_id,
-                f"task {ref.task_id} filed but marker not written: {exc}",
-            )
-        if ref.merged:
-            return Result(
-                bundle, Outcome.IMPORTED, ref.task_id, "merged into an existing note"
-            )
-        return self._finish(bundle, payload, draft)
-
-    def _repair(self, feedback_dir: Path, bundle: Path) -> Result:
-        """Finish a filed task whose rewrite never completed (or predates it)."""
-        data = _read_marker(bundle)
-        if data is None or data.get("merged") or data.get("format") == FORMAT:
-            return Result(bundle, Outcome.SKIPPED)
-        entry = find_entry(feedback_dir, bundle)
-        if entry is None:
-            return Result(bundle, Outcome.PENDING, detail="entry missing for repair")
-        report = read_report(bundle / "report.json")
-        draft = build_task(repo_name(feedback_dir), bundle.name, entry, report)
-        return self._finish(bundle, data, draft)
-
-    def _finish(
-        self, bundle: Path, data: dict[str, object], draft: TaskDraft
-    ) -> Result:
-        """Rewrite the stored note to the exact title/body, then mark it done."""
-        task_id = str(data["task_id"])
-        try:
-            self.client.update_task(task_id, draft.title, draft.body)
-        except TodoError as exc:
-            return Result(bundle, Outcome.FAILED, task_id, f"update: {exc}")
-        marked = {**data, "title": draft.title, "due": draft.due, "format": FORMAT}
-        try:
-            _write_marker(bundle, marked)
-        except OSError as exc:
-            return Result(bundle, Outcome.FAILED, task_id, f"marker not written: {exc}")
-        return Result(bundle, Outcome.IMPORTED, task_id)
-
-
-@dataclass(frozen=True)
-class Waiter:
-    """How long to wait for half-written bundles, and the clock to do it by."""
-
-    wait: float
-    clock: Callable[[], float]
-    sleep: Callable[[float], None]
-
-
-def run_once(data_root: Path, importer: Importer, now: float) -> list[Result]:
-    """Import every complete, not yet imported bundle of every game."""
-    return [
-        importer.import_bundle(folder, bundle, now)
-        for folder in feedback_dirs(data_root)
-        for bundle in bundles_in(folder)
-    ]
-
-
-def run_settled(data_root: Path, importer: Importer, waiter: Waiter) -> list[Result]:
-    """Like :func:`run_once`, retrying up to ``wait`` s while bundles are pending.
-
-    A bundle is written in pieces; the watch fires on the first piece, so a
-    short wait lets the rest land instead of leaving it to the timer.
-    """
-    deadline = waiter.clock() + waiter.wait
-    while True:
-        results = run_once(data_root, importer, waiter.clock())
-        pending = any(r.outcome is Outcome.PENDING for r in results)
-        if not pending or waiter.clock() >= deadline:
-            return results
-        waiter.sleep(1.0)

feedback_inbox/feedback_inbox/_paths.py

@@ -1,51 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Every on-disk and network location, resolved in one place.
-
-Tests redirect :func:`resolve_paths` once; nothing else names a real path.
-"""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-from pathlib import Path
-
-MCP_URL = "http://127.0.0.1:8735/mcp"
-# Port 9 (discard) refuses connections: a test that forgets to inject a client
-# fails closed instead of reaching the live todo server.
-DEAD_URL = "http://127.0.0.1:9/mcp"
-
-
-@dataclass(frozen=True)
-class Paths:
-    """Where the inbox reads games, its token, its lock and its units."""
-
-    data_root: Path
-    token_file: Path
-    lock_file: Path
-    unit_dir: Path
-    mcp_url: str
-
-    @classmethod
-    def under(cls, root: Path) -> Paths:
-        """A throwaway layout below ``root`` that can never reach the live app."""
-        return cls(
-            data_root=root / "data",
-            token_file=root / "token",
-            lock_file=root / "state" / "lock",
-            unit_dir=root / "units",
-            mcp_url=DEAD_URL,
-        )
-
-
-def resolve_paths(given: Paths | None = None) -> Paths:
-    """Return ``given`` or the real layout under the user's home."""
-    if given is not None:
-        return given
-    home = Path.home()
-    return Paths(
-        data_root=home / "data",
-        token_file=home / ".config" / "feedback-inbox" / "token",
-        lock_file=home / ".local" / "share" / "feedback-inbox" / "lock",
-        unit_dir=home / ".config" / "systemd" / "user",
-        mcp_url=MCP_URL,
-    )

feedback_inbox/feedback_inbox/_todo.py

@@ -1,165 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The todo app's MCP server as a minimal ``add_task`` client."""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-import json
-import re
-from typing import TYPE_CHECKING, Protocol
-import urllib.error
-import urllib.request
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-_PROTOCOL = "2025-03-26"
-_TIMEOUT = 20
-_REPLY = re.compile(r"^(added|merged into|appended to) (\S+)")
-
-
-class TodoError(Exception):
-    """The todo server could not take the task."""
-
-
-@dataclass(frozen=True)
-class TaskRef:
-    """The task a call created (or the note it was merged into)."""
-
-    task_id: str
-    merged: bool
-
-
-class TodoClient(Protocol):
-    """What the inbox needs from a todo app."""
-
-    def add_task(self, repo: str, title: str, body: str) -> TaskRef:
-        """Create one task in its own note and return its id."""
-        raise NotImplementedError
-
-    def update_task(self, task_id: str, title: str, body: str) -> None:
-        """Replace the task's stored title and body with exactly these."""
-        raise NotImplementedError
-
-
-class McpTodoClient:
-    """Streamable-HTTP MCP client: initialize, then one ``add_task`` call."""
-
-    def __init__(self, url: str, token_file: Path) -> None:
-        """Remember where the server and the (lazily read) token are."""
-        self._url = url
-        self._token_file = token_file
-
-    def add_task(self, repo: str, title: str, body: str) -> TaskRef:
-        """Create a task with its own note via ``add_task``.
-
-        The server stores a ``new_note`` title as ``<name> todos: <title>``;
-        :meth:`update_task` restores the exact title (and drops the
-        ``## <title>`` section the server adds) afterwards.
-        """
-        arguments = {"repo": repo, "title": title, "body": body, "new_note": True}
-        said = self._call("add_task", arguments)
-        found = _REPLY.match(said)
-        if not found:
-            msg = f"add_task refused: {said[:120]}"
-            raise TodoError(msg)
-        return TaskRef(task_id=found.group(2), merged=found.group(1) != "added")
-
-    def update_task(self, task_id: str, title: str, body: str) -> None:
-        """Rewrite via ``update_task``: exact title, and body with no sections."""
-        said = self._call("update_task", {"id": task_id, "title": title, "body": body})
-        if not said.startswith("updated"):
-            msg = f"update_task refused: {said[:120]}"
-            raise TodoError(msg)
-
-    def _call(self, tool: str, arguments: dict[str, object]) -> str:
-        token = self._token()
-        session = self._initialize(token)
-        payload = {
-            "jsonrpc": "2.0",
-            "id": 2,
-            "method": "tools/call",
-            "params": {"name": tool, "arguments": arguments},
-        }
-        return _tool_text(self._post(token, payload, session)[1])
-
-    def _token(self) -> str:
-        try:
-            return self._token_file.read_text(encoding="utf-8").strip()
-        except OSError as exc:
-            msg = f"token file unreadable: {self._token_file}"
-            raise TodoError(msg) from exc
-
-    def _initialize(self, token: str) -> str:
-        init = {
-            "jsonrpc": "2.0",
-            "id": 1,
-            "method": "initialize",
-            "params": {
-                "protocolVersion": _PROTOCOL,
-                "capabilities": {},
-                "clientInfo": {"name": "feedback-inbox", "version": "0.1.0"},
-            },
-        }
-        headers, _body = self._post(token, init, None)
-        session = headers.get("mcp-session-id")
-        if not session:
-            msg = "todo server sent no session id"
-            raise TodoError(msg)
-        self._post(
-            token,
-            {"jsonrpc": "2.0", "method": "notifications/initialized"},
-            session,
-        )
-        return session
-
-    def _post(
-        self,
-        token: str,
-        payload: dict[str, object],
-        session: str | None,
-    ) -> tuple[dict[str, str], str]:
-        headers = {
-            "Authorization": f"Bearer {token}",
-            "Content-Type": "application/json",
-            "Accept": "application/json, text/event-stream",
-        }
-        if session:
-            headers["mcp-session-id"] = session
-        request = urllib.request.Request(
-            self._url,
-            data=json.dumps(payload).encode(),
-            headers=headers,
-        )
-        try:
-            with urllib.request.urlopen(request, timeout=_TIMEOUT) as resp:
-                return (
-                    {k.lower(): v for k, v in resp.headers.items()},
-                    resp.read().decode(),
-                )
-        except urllib.error.HTTPError as exc:
-            exc.close()
-            msg = f"todo server answered HTTP {exc.code}"
-            raise TodoError(msg) from exc
-        except OSError as exc:
-            msg = f"todo server unreachable: {exc}"
-            raise TodoError(msg) from exc
-
-
-def _tool_text(body: str) -> str:
-    """The text a tool call answered; an MCP-level error becomes TodoError."""
-    text = body.strip()
-    if not text.startswith("{"):
-        lines = [x[5:].strip() for x in text.splitlines() if x.startswith("data:")]
-        text = next((x for x in lines if x.startswith("{")), "")
-    try:
-        result = json.loads(text)["result"]
-        said = str(result["content"][0]["text"])
-        failed = bool(result.get("isError"))
-    except (ValueError, KeyError, IndexError, TypeError) as exc:
-        msg = f"unparsable MCP reply: {body[:80]!r}"
-        raise TodoError(msg) from exc
-    if failed:
-        msg = f"tool refused: {said[:120]}"
-        raise TodoError(msg)
-    return said

feedback_inbox/feedback_inbox/_units.py

@@ -1,67 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Keep the path units' watch lists equal to the games that exist.
-
-systemd ``.path`` units take no globs, so each ``*_binaries`` folder is listed
-in a generated drop-in. The set changes only when a game is added or removed;
-the drop-in is rewritten, and systemd reloaded, only then (a reload per tick
-would be the daemon-reload loop the fleet rules forbid).
-"""
-
-from __future__ import annotations
-
-import subprocess
-from typing import TYPE_CHECKING
-
-from feedback_inbox._bundle import binaries_dirs, feedback_dirs
-
-if TYPE_CHECKING:
-    from collections.abc import Callable, Sequence
-    from pathlib import Path
-
-WATCH_PATH = "feedback-inbox.path"
-SYNC_PATH = "feedback-inbox-sync.path"
-SYSTEMCTL = "/usr/bin/systemctl"
-_HEADER = "# Generated by `feedback-inbox sync`; do not edit.\n[Path]\n"
-
-
-def _dropin(unit_dir: Path, unit: str) -> Path:
-    return unit_dir / f"{unit}.d" / "dirs.conf"
-
-
-def _render(folders: list[Path]) -> str:
-    return _HEADER + "".join(f"PathChanged={d}\n" for d in folders)
-
-
-def desired(data_root: Path, unit_dir: Path) -> dict[Path, str]:
-    """The drop-ins that should exist: feedback folders, and game folders."""
-    return {
-        _dropin(unit_dir, WATCH_PATH): _render(feedback_dirs(data_root)),
-        _dropin(unit_dir, SYNC_PATH): _render(binaries_dirs(data_root)),
-    }
-
-
-def run_systemctl(args: Sequence[str]) -> int:
-    """Run ``systemctl --user <args>`` and return its exit code."""
-    return subprocess.run([SYSTEMCTL, "--user", *args], check=False).returncode
-
-
-def sync_units(
-    data_root: Path,
-    unit_dir: Path,
-    systemctl: Callable[[Sequence[str]], int],
-) -> bool:
-    """Rewrite drifted drop-ins, then reload and re-arm once; True if changed."""
-    changed = False
-    for path, text in desired(data_root, unit_dir).items():
-        if path.is_file() and path.read_text(encoding="utf-8") == text:
-            continue
-        path.parent.mkdir(parents=True, exist_ok=True)
-        path.write_text(text, encoding="utf-8")
-        changed = True
-    if not changed:
-        return False
-    for args in (["daemon-reload"], ["restart", WATCH_PATH, SYNC_PATH]):
-        if systemctl(args):
-            msg = f"systemctl --user {' '.join(args)} failed"
-            raise RuntimeError(msg)
-    return True

feedback_inbox/feedback_inbox/tests/__init__.py

@@ -1,2 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Tests for feedback_inbox."""

feedback_inbox/feedback_inbox/tests/conftest.py

@@ -1,31 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Redirect the one path resolver, so no test can touch real user state.
-
-Every writer resolves its paths through ``feedback_inbox._paths.resolve_paths``;
-redirecting that one function moves all of them (kuhy's real ``~/data``,
-token file, lock and systemd unit dir included).
-"""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-import pytest
-
-from feedback_inbox._paths import Paths
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-
-@pytest.fixture(autouse=True)
-def paths(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> Paths:
-    """Point every default path at ``tmp_path`` before any test runs."""
-    redirected = Paths.under(tmp_path / "home")
-
-    def _resolve(given: Paths | None = None) -> Paths:
-        return given if given is not None else redirected
-
-    monkeypatch.setattr("feedback_inbox._cli.resolve_paths", _resolve)
-    redirected.data_root.mkdir(parents=True)
-    return redirected

feedback_inbox/feedback_inbox/tests/test_bundle.py

@@ -1,42 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Discovery, stamps and due dates."""
-
-from __future__ import annotations
-
-from datetime import datetime
-from typing import TYPE_CHECKING
-
-from feedback_inbox._bundle import bundles_in, due_at, feedback_dirs, repo_name
-from feedback_inbox.tests.test_support import make_bundle
-
-if TYPE_CHECKING:
-    from feedback_inbox._paths import Paths
-
-
-def test_feedback_dirs_covers_every_game(paths: Paths) -> None:
-    make_bundle(paths.data_root, game="cleanup-plate")
-    make_bundle(paths.data_root, game="dread-grid")
-    (paths.data_root / "other" / "feedback").mkdir(parents=True)
-    (paths.data_root / "empty_binaries").mkdir()
-    (paths.data_root / "file_binaries").write_text("x", encoding="utf-8")
-    found = feedback_dirs(paths.data_root)
-    assert [repo_name(d) for d in found] == ["cleanup-plate", "dread-grid"]
-
-
-def test_feedback_dirs_without_data_root(paths: Paths) -> None:
-    assert feedback_dirs(paths.data_root / "missing") == []
-
-
-def test_bundles_in_keeps_only_stamp_folders(paths: Paths) -> None:
-    first = make_bundle(paths.data_root, stamp="2026-11-11T13-53-19-124")
-    make_bundle(paths.data_root, stamp="2026-11-11T13-52-33-162")
-    feedback = first.parent
-    (feedback / "notes").mkdir()
-    (feedback / "2026-11-11T13-52-33-999").write_text("x", encoding="utf-8")
-    names = [b.name for b in bundles_in(feedback)]
-    assert names == ["2026-11-11T13-52-33-162", "2026-11-11T13-53-19-124"]
-
-
-def test_due_is_next_day_ten(paths: Paths) -> None:
-    assert due_at("2026-11-11T13-52-33-162") == datetime(2026, 11, 12, 10, 0)
-    assert due_at("2026-12-31T23-59-59-001") == datetime(2027, 1, 1, 10, 0)

feedback_inbox/feedback_inbox/tests/test_cli.py

@@ -1,103 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The command line: run, sync, exit codes, lock."""
-
-from __future__ import annotations
-
-import logging
-from typing import TYPE_CHECKING
-
-import pytest
-
-from feedback_inbox import __main__ as entry
-from feedback_inbox._cli import main
-from feedback_inbox._inbox import MARKER
-from feedback_inbox._paths import Paths, resolve_paths
-from feedback_inbox.tests.test_support import FakeTodo, make_bundle
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-    from pathlib import Path
-
-
-def test_run_imports_and_exits_zero(
-    paths: Paths, caplog: pytest.LogCaptureFixture
-) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo()
-    with caplog.at_level(logging.INFO):
-        assert main(["run", "--wait", "0"], client=todo) == 0
-    assert (bundle / MARKER).exists()
-    assert len(todo.calls) == 1
-    assert "imported" in caplog.text
-    assert main(["run", "--wait", "0"], client=todo) == 0
-    assert len(todo.calls) == 1
-
-
-def test_run_fails_loudly_when_todo_is_down(
-    paths: Paths, caplog: pytest.LogCaptureFixture
-) -> None:
-    bundle = make_bundle(paths.data_root)
-    assert main(["run", "--wait", "0"], client=FakeTodo(fail=True)) == 1
-    assert not (bundle / MARKER).exists()
-    assert "todo down" in caplog.text
-
-
-def test_run_pending_is_not_an_error(paths: Paths) -> None:
-    make_bundle(paths.data_root, entry=False)
-    assert main(["run", "--wait", "0"], client=FakeTodo()) == 0
-
-
-def test_run_builds_the_real_client_from_paths(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    assert main(["run", "--wait", "0"]) == 1
-
-
-def test_run_with_nothing_to_do_needs_no_token(paths: Paths) -> None:
-    assert main(["run", "--wait", "0"]) == 0
-
-
-def test_run_takes_the_lock_file(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    assert main(["run", "--wait", "0"], client=FakeTodo()) == 0
-    assert paths.lock_file.exists()
-
-
-def test_sync_command(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    calls: list[list[str]] = []
-
-    def fake(args: Sequence[str]) -> int:
-        calls.append(list(args))
-        return 0
-
-    assert main(["sync"], systemctl=fake) == 0
-    assert calls[0] == ["daemon-reload"]
-
-
-def test_sync_failure_exit_code(paths: Paths, caplog: pytest.LogCaptureFixture) -> None:
-    make_bundle(paths.data_root)
-    assert main(["sync"], systemctl=lambda _a: 1) == 1
-    assert "daemon-reload" in caplog.text
-
-
-def test_usage_error() -> None:
-    with pytest.raises(SystemExit):
-        main([])
-
-
-def test_real_paths_are_under_home(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
-) -> None:
-    monkeypatch.setattr("feedback_inbox._paths.Path.home", lambda: tmp_path)
-    real = resolve_paths()
-    assert real.data_root == tmp_path / "data"
-    assert real.token_file == tmp_path / ".config" / "feedback-inbox" / "token"
-    assert real.unit_dir == tmp_path / ".config" / "systemd" / "user"
-    assert real.mcp_url == "http://127.0.0.1:8735/mcp"
-    given = Paths.under(tmp_path)
-    assert resolve_paths(given) is given
-
-
-def test_module_entry_point(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr("sys.argv", ["feedback-inbox", "run", "--wait", "0"])
-    assert entry.run() == 0

feedback_inbox/feedback_inbox/tests/test_entry.py

@@ -1,68 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Finding a bundle's markdown entry and reading its report."""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-from feedback_inbox._entry import (
-    find_entry,
-    game_and_combo,
-    note_from_entry,
-    read_report,
-)
-from feedback_inbox.tests.test_support import make_bundle
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-    from feedback_inbox._paths import Paths
-
-
-def test_find_entry_picks_the_matching_section(paths: Paths) -> None:
-    first = make_bundle(paths.data_root, stamp="2026-11-11T13-52-33-162", note="alpha")
-    second = make_bundle(paths.data_root, stamp="2026-11-11T13-53-19-124", note="beta")
-    entry = find_entry(first.parent, second)
-    assert entry is not None
-    assert "beta" in entry
-    assert "alpha" not in entry
-    assert find_entry(first.parent, first) is not None
-
-
-def test_find_entry_missing(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root, entry=False)
-    assert find_entry(bundle.parent, bundle) is None
-    (bundle.parent / "feedback-2026-11-11.md").write_bytes(b"\xff\xfe")
-    assert find_entry(bundle.parent, bundle) is None
-
-
-def test_read_report(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root, note="  hello  \nworld")
-    report = read_report(bundle / "report.json")
-    assert report is not None
-    assert report.note == "  hello  \nworld"
-    assert (report.game, report.combo) == ("museum", "loop=P2")
-
-
-def test_read_report_corrupt_forms(tmp_path: Path) -> None:
-    path = tmp_path / "report.json"
-    for bad in ("{broken", "[1]", '{"note": 3, "game": null}'):
-        path.write_text(bad, encoding="utf-8")
-        report = read_report(path)
-        if bad == '{"note": 3, "game": null}':
-            assert report is not None
-            assert (report.note, report.game, report.combo) == ("", "", "")
-        else:
-            assert report is None
-    path.write_bytes(b"\xff\xfe")
-    assert read_report(path) is None
-
-
-def test_entry_fallbacks(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root, note="line one\nline two")
-    entry = find_entry(bundle.parent, bundle)
-    assert entry is not None
-    assert note_from_entry(entry) == "line one\nline two"
-    assert game_and_combo(entry) == ("museum", "loop=P2,carry=d,id=c")
-    assert note_from_entry("## head only") == ""
-    assert game_and_combo("nothing here") == ("", "")

feedback_inbox/feedback_inbox/tests/test_inbox.py

@@ -1,194 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Import rules: complete, incomplete, idempotent, titles, due date, failures."""
-
-from __future__ import annotations
-
-import json
-from typing import TYPE_CHECKING
-
-from feedback_inbox._entry import Report
-from feedback_inbox._inbox import (
-    MARKER,
-    Importer,
-    Outcome,
-    Waiter,
-    build_task,
-    run_once,
-    run_settled,
-)
-from feedback_inbox.tests.test_support import STAMP, FakeTodo, make_bundle
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-    from feedback_inbox._paths import Paths
-
-NOW = 1_000_000.0
-OWNER = "kuhyx"
-MANUAL = "MANUAL WORK REQUIRED: "
-
-
-def _run(paths: Paths, todo: FakeTodo) -> list[Outcome]:
-    return [r.outcome for r in run_once(paths.data_root, Importer(todo, OWNER), NOW)]
-
-
-def test_complete_bundle_becomes_one_task(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    ((repo, title, body),) = todo.calls
-    assert repo == "kuhyx/cleanup-plate"
-    assert (
-        title
-        == MANUAL + "cleanup-plate feedback: I would like to see the painting IN FULL"
-    )
-    assert body.startswith("Due: 2026-11-12 10:00\n\n### 2026-11-11T13:52:33")
-    assert f"{bundle}/report.json" in body
-    marker = json.loads((bundle / MARKER).read_text(encoding="utf-8"))
-    assert marker["task_id"] == "id0001" + "0" * 28
-    assert marker["due"] == "2026-11-12T10:00:00"
-
-
-def test_rerun_never_duplicates(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    assert _run(paths, todo) == [Outcome.SKIPPED]
-    assert len(todo.calls) == 1
-
-
-def test_incomplete_bundles_wait(paths: Paths) -> None:
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-00-001", report=None)
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-01-001", entry=False)
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-02-001", age=0.5)
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.PENDING] * 3
-    assert todo.calls == []
-    later = run_once(paths.data_root, Importer(todo, OWNER), NOW + 10)
-    assert [r.outcome for r in later] == [
-        Outcome.PENDING,
-        Outcome.PENDING,
-        Outcome.IMPORTED,
-    ]
-
-
-def test_multiple_games(paths: Paths) -> None:
-    make_bundle(paths.data_root, game="cleanup-plate", note="a")
-    make_bundle(paths.data_root, game="dread-grid", note="b")
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED] * 2
-    assert [(c[0], c[1]) for c in todo.calls] == [
-        ("kuhyx/cleanup-plate", "MANUAL WORK REQUIRED: cleanup-plate feedback: a"),
-        ("kuhyx/dread-grid", "MANUAL WORK REQUIRED: dread-grid feedback: b"),
-    ]
-
-
-def test_corrupt_report_still_a_task(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root, report="{broken", note="keep me")
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    ((_, title, body),) = todo.calls
-    assert (
-        title
-        == "MANUAL WORK REQUIRED: cleanup-plate feedback: keep me [corrupt report.json]"
-    )
-    assert "report.json is corrupt" in body
-    assert (bundle / MARKER).exists()
-
-
-def test_corrupt_report_without_note(paths: Paths) -> None:
-    make_bundle(paths.data_root, report="[]", note="")
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    title = todo.calls[0][1]
-    game_part = "museum loop=P2,carry=d,id=c 13:52"
-    assert title == f"{MANUAL}cleanup-plate feedback: {game_part} [corrupt report.json]"
-
-
-def test_no_note_title_names_game_combo_time(paths: Paths) -> None:
-    make_bundle(paths.data_root, note="")
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    assert (
-        todo.calls[0][1]
-        == "MANUAL WORK REQUIRED: cleanup-plate feedback: museum loop=P2 13:52"
-    )
-
-
-def test_title_uses_first_line_and_caps_length() -> None:
-    report = Report(note="\n  " + "x" * 300 + "\nsecond", game="g", combo="c")
-    draft = build_task("repo", STAMP, "## e", report)
-    assert draft.title == "MANUAL WORK REQUIRED: repo feedback: " + "x" * 119 + "…"
-    bare = build_task("repo", STAMP, "## e", Report(note="", game="", combo=""))
-    assert bare.title == "MANUAL WORK REQUIRED: repo feedback: 13:52"
-
-
-def test_todo_failure_leaves_no_marker(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo(fail=True)
-    results = run_once(paths.data_root, Importer(todo, OWNER), NOW)
-    assert [r.outcome for r in results] == [Outcome.FAILED]
-    assert "todo down" in results[0].detail
-    assert not (bundle / MARKER).exists()
-    assert _run(paths, FakeTodo()) == [Outcome.IMPORTED]
-
-
-def test_merge_is_flagged_but_marked(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    results = run_once(paths.data_root, Importer(FakeTodo(merged=True), OWNER), NOW)
-    assert results[0].outcome is Outcome.IMPORTED
-    assert "merged" in results[0].detail
-
-
-def test_marker_write_failure_is_reported(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    (bundle / MARKER).mkdir()
-    results = run_once(paths.data_root, Importer(FakeTodo(), OWNER), NOW)
-    assert results[0].outcome is Outcome.SKIPPED
-    (bundle / MARKER).rmdir()
-    bundle.chmod(0o500)
-    try:
-        results = run_once(paths.data_root, Importer(FakeTodo(), OWNER), NOW)
-    finally:
-        bundle.chmod(0o700)
-    assert results[0].outcome is Outcome.FAILED
-    assert "marker" in results[0].detail
-
-
-def test_run_settled_waits_then_gives_up(paths: Paths, tmp_path: Path) -> None:
-    make_bundle(paths.data_root, age=0.0, now=NOW)
-    clock = [NOW]
-    slept: list[float] = []
-
-    def sleep(seconds: float) -> None:
-        slept.append(seconds)
-        clock[0] += seconds
-
-    todo = FakeTodo()
-    results = run_settled(
-        paths.data_root, Importer(todo, OWNER), Waiter(5.0, lambda: clock[0], sleep)
-    )
-    assert [r.outcome for r in results] == [Outcome.IMPORTED]
-    assert slept == [1.0, 1.0]
-
-    other = tmp_path / "other"
-    other.mkdir()
-    make_bundle(other, report=None)
-    slept.clear()
-    results = run_settled(
-        other, Importer(todo, OWNER), Waiter(3.0, lambda: clock[0], sleep)
-    )
-    assert [r.outcome for r in results] == [Outcome.PENDING]
-    assert slept == [1.0, 1.0, 1.0]
-
-
-def test_every_title_starts_with_the_manual_prefix(paths: Paths) -> None:
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-00-001", note="a")
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-01-001", note="")
-    make_bundle(paths.data_root, stamp="2026-11-11T10-00-02-001", report="{")
-    make_bundle(paths.data_root, game="dread-grid", note="b")
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED] * 4
-    titles = [call[1] for call in todo.calls]
-    assert len(titles) == 4
-    assert all(t.startswith("MANUAL WORK REQUIRED: ") for t in titles)

feedback_inbox/feedback_inbox/tests/test_inbox_repair.py

@@ -1,145 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Exact titles, section-free bodies, and repair of half-finished tasks."""
-
-from __future__ import annotations
-
-import json
-from typing import TYPE_CHECKING
-
-from feedback_inbox._entry import Report
-from feedback_inbox._inbox import (
-    FORMAT,
-    MARKER,
-    Importer,
-    Outcome,
-    build_task,
-    run_once,
-)
-from feedback_inbox.tests.test_support import STAMP, FakeTodo, make_bundle
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-    from feedback_inbox._paths import Paths
-
-NOW = 1_000_000.0
-OWNER = "kuhyx"
-MANUAL = "MANUAL WORK REQUIRED: "
-
-
-def _run(paths: Paths, todo: FakeTodo) -> list[Outcome]:
-    return [r.outcome for r in run_once(paths.data_root, Importer(todo, OWNER), NOW)]
-
-
-def _marker(bundle: Path) -> dict[str, object]:
-    data: dict[str, object] = json.loads((bundle / MARKER).read_text(encoding="utf-8"))
-    return data
-
-
-def _no_extra_sections(text: str) -> bool:
-    return not any(line.startswith("## ") for line in text.splitlines())
-
-
-def test_stored_note_is_one_manual_part(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    (stored,) = todo.stored.values()
-    assert stored.startswith(MANUAL)
-    assert _no_extra_sections(stored)
-    assert "### 2026-11-11T13:52:33" in stored
-    assert _marker(bundle)["format"] == FORMAT
-
-
-def test_entry_headings_never_become_sections() -> None:
-    entry = "## a heading\ntext\n## another\n### kept"
-    draft = build_task("r", STAMP, entry, Report(note="n", game="", combo=""))
-    assert _no_extra_sections(draft.body)
-    assert "### a heading\ntext\n### another\n### kept" in draft.body
-
-
-def test_failed_update_keeps_marker_and_retries_without_a_second_task(
-    paths: Paths,
-) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo(fail_update=True)
-    results = run_once(paths.data_root, Importer(todo, OWNER), NOW)
-    assert [r.outcome for r in results] == [Outcome.FAILED]
-    assert "update refused" in results[0].detail
-    assert "format" not in _marker(bundle)
-    todo.fail_update = False
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    assert len(todo.calls) == 1
-    (stored,) = todo.stored.values()
-    assert stored.startswith(MANUAL)
-    assert _no_extra_sections(stored)
-    assert _run(paths, todo) == [Outcome.SKIPPED]
-
-
-def test_old_marker_is_repaired_from_the_bundle(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    task_id = "abc" + "0" * 33
-    (bundle / MARKER).write_text(
-        json.dumps({"task_id": task_id, "title": "old", "title_fixed": True}),
-        encoding="utf-8",
-    )
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    assert todo.calls == []
-    assert todo.stored[task_id].startswith(MANUAL)
-    assert _no_extra_sections(todo.stored[task_id])
-    assert _marker(bundle)["format"] == FORMAT
-
-
-def test_old_marker_whose_entry_vanished_waits(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root, entry=False)
-    (bundle / MARKER).write_text(
-        json.dumps({"task_id": "t" * 36, "title": "old"}), encoding="utf-8"
-    )
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.PENDING]
-    assert todo.updated == []
-
-
-def test_merged_task_is_not_rewritten(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    todo = FakeTodo(merged=True)
-    assert _run(paths, todo) == [Outcome.IMPORTED]
-    assert todo.updated == []
-    assert _marker(bundle)["format"] == FORMAT
-    assert _run(paths, todo) == [Outcome.SKIPPED]
-
-
-def test_old_merged_marker_is_left_alone(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    (bundle / MARKER).write_text(
-        json.dumps({"task_id": "t" * 36, "title": "x", "merged": True}),
-        encoding="utf-8",
-    )
-    todo = FakeTodo()
-    assert _run(paths, todo) == [Outcome.SKIPPED]
-    assert todo.updated == []
-
-
-def test_unreadable_marker_is_left_alone(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    for junk in ("{broken", "[1]", '{"task_id": 5}'):
-        (bundle / MARKER).write_text(junk, encoding="utf-8")
-        todo = FakeTodo()
-        assert _run(paths, todo) == [Outcome.SKIPPED]
-        assert todo.calls == []
-        assert todo.updated == []
-
-
-def test_repair_reports_an_unwritable_marker(paths: Paths) -> None:
-    bundle = make_bundle(paths.data_root)
-    (bundle / MARKER).write_text(
-        json.dumps({"task_id": "t" * 36, "title": "T"}), encoding="utf-8"
-    )
-    bundle.chmod(0o500)
-    try:
-        results = run_once(paths.data_root, Importer(FakeTodo(), OWNER), NOW)
-    finally:
-        bundle.chmod(0o700)
-    assert results[0].outcome is Outcome.FAILED
-    assert "marker not written" in results[0].detail

feedback_inbox/feedback_inbox/tests/test_support.py

@@ -1,95 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Builders for fake feedback bundles and a fake todo client."""
-
-from __future__ import annotations
-
-from dataclasses import dataclass, field
-import json
-import os
-from typing import TYPE_CHECKING
-
-from feedback_inbox._todo import TaskRef, TodoError
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-STAMP = "2026-11-11T13-52-33-162"
-NOTE = "I would like to see the painting IN FULL\n\nsecond paragraph"
-
-
-def entry_text(bundle: Path, note: str = NOTE, game: str = "museum") -> str:
-    """The markdown entry a game appends for ``bundle``."""
-    head = note.splitlines()[0][:70] if note else ""
-    parts = [f"## 2026-11-11T13:52:33: {head}", ""]
-    if note:
-        parts += [note, ""]
-    parts += [
-        f"- Screenshot: `{bundle}/annotated.png`",
-        f"- Report: `{bundle}/report.json`",
-        f"- Game `{game}`, combo `loop=P2,carry=d,id=c`, build abc, session `s`",
-        "",
-        "Task: Read the annotated screenshot and fix it.",
-        "",
-    ]
-    return "\n".join(parts)
-
-
-def make_bundle(
-    data_root: Path,
-    *,
-    game: str = "cleanup-plate",
-    stamp: str = STAMP,
-    note: str = NOTE,
-    report: str | None = "ok",
-    entry: bool = True,
-    age: float = 60.0,
-    now: float = 1_000_000.0,
-) -> Path:
-    """Create ``<game>_binaries/feedback/<stamp>/`` and its md entry."""
-    feedback = data_root / f"{game}_binaries" / "feedback"
-    bundle = feedback / stamp
-    bundle.mkdir(parents=True, exist_ok=True)
-    if report is not None:
-        path = bundle / "report.json"
-        if report == "ok":
-            body = json.dumps({"note": note, "game": "museum", "combo": "loop=P2"})
-        else:
-            body = report
-        path.write_text(body, encoding="utf-8")
-        os.utime(path, (now - age, now - age))
-    if entry:
-        md = feedback / f"feedback-{stamp[:10]}.md"
-        with md.open("a", encoding="utf-8") as handle:
-            handle.write(entry_text(bundle, note))
-    return bundle
-
-
-@dataclass
-class FakeTodo:
-    """A todo server stand-in that stores notes the way the real one does."""
-
-    calls: list[tuple[str, str, str]] = field(default_factory=list)
-    stored: dict[str, str] = field(default_factory=dict)
-    fail: bool = False
-    fail_update: bool = False
-    merged: bool = False
-    updated: list[str] = field(default_factory=list)
-
-    def add_task(self, repo: str, title: str, body: str) -> TaskRef:
-        if self.fail:
-            msg = "todo down"
-            raise TodoError(msg)
-        self.calls.append((repo, title, body))
-        task_id = f"id{len(self.calls):04d}" + "0" * 28
-        # A new_note task: title "<name> todos: <title>", then a "## <title>"
-        # section holding the body.
-        name = repo.rsplit("/", maxsplit=1)[-1]
-        self.stored[task_id] = f"{name} todos: {title}\n\n## {title}\n{body}"
-        return TaskRef(task_id=task_id, merged=self.merged)
-
-    def update_task(self, task_id: str, title: str, body: str) -> None:
-        if self.fail_update:
-            msg = "update refused"
-            raise TodoError(msg)
-        self.updated.append(task_id)
-        self.stored[task_id] = f"{title}\n\n{body}"

feedback_inbox/feedback_inbox/tests/test_todo.py

@@ -1,200 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The MCP client against a real local HTTP server (never the live todo app)."""
-
-from __future__ import annotations
-
-from http.server import BaseHTTPRequestHandler, HTTPServer
-import json
-import threading
-from typing import TYPE_CHECKING, ClassVar
-
-import pytest
-
-from feedback_inbox._todo import McpTodoClient, TodoError
-
-if TYPE_CHECKING:
-    from collections.abc import Iterator
-    from pathlib import Path
-
-FULL_ID = "a1b2c3d4" + "e" * 28
-
-
-class _Handler(BaseHTTPRequestHandler):
-    seen: ClassVar[list[dict[str, object]]] = []
-    headers_seen: ClassVar[list[dict[str, str]]] = []
-    tool_text: ClassVar[str] = f"added {FULL_ID}\nline"
-    is_error: ClassVar[bool] = False
-    status: ClassVar[int] = 200
-    sse: ClassVar[bool] = True
-
-    def do_POST(self) -> None:
-        length = int(self.headers["Content-Length"])
-        body = json.loads(self.rfile.read(length))
-        type(self).seen.append(body)
-        type(self).headers_seen.append({k.lower(): v for k, v in self.headers.items()})
-        if type(self).status != 200:
-            self.send_response(type(self).status)
-            self.end_headers()
-            return
-        self.send_response(200)
-        self.send_header("mcp-session-id", "sess-1")
-        self.end_headers()
-        method = body.get("method")
-        if method == "initialize":
-            self.wfile.write(
-                b'event: message\ndata: {"jsonrpc":"2.0","id":1,"result":{}}\n\n'
-            )
-        elif method == "tools/call":
-            result: dict[str, object] = {
-                "content": [{"type": "text", "text": type(self).tool_text}],
-            }
-            if type(self).is_error:
-                result["isError"] = True
-            reply = json.dumps({"jsonrpc": "2.0", "id": 2, "result": result})
-            out = f"event: message\ndata: {reply}\n\n" if type(self).sse else reply
-            self.wfile.write(out.encode())
-
-    def log_message(self, *_args: object) -> None:
-        return
-
-
-@pytest.fixture
-def server() -> Iterator[str]:
-    _Handler.seen = []
-    _Handler.headers_seen = []
-    _Handler.tool_text = f"added {FULL_ID}\nline"
-    _Handler.is_error = False
-    _Handler.status = 200
-    _Handler.sse = True
-    httpd = HTTPServer(("127.0.0.1", 0), _Handler)
-    thread = threading.Thread(target=httpd.serve_forever, daemon=True)
-    thread.start()
-    yield f"http://127.0.0.1:{httpd.server_port}/mcp"
-    httpd.shutdown()
-    httpd.server_close()
-    thread.join()
-
-
-def _client(url: str, tmp_path: Path) -> McpTodoClient:
-    token = tmp_path / "token"
-    token.write_text("secret-token\n", encoding="utf-8")
-    return McpTodoClient(url, token)
-
-
-def test_add_task_round_trip(server: str, tmp_path: Path) -> None:
-    ref = _client(server, tmp_path).add_task("kuhyx/cleanup-plate", "t", "b")
-    assert (ref.task_id, ref.merged) == (FULL_ID, False)
-    methods = [m.get("method") for m in _Handler.seen]
-    assert methods == ["initialize", "notifications/initialized", "tools/call"]
-    call = _Handler.seen[2]["params"]
-    assert call == {
-        "name": "add_task",
-        "arguments": {
-            "repo": "kuhyx/cleanup-plate",
-            "title": "t",
-            "body": "b",
-            "new_note": True,
-        },
-    }
-    assert _Handler.headers_seen[0]["authorization"] == "Bearer secret-token"
-    assert _Handler.headers_seen[2]["mcp-session-id"] == "sess-1"
-
-
-def test_plain_json_reply_and_merge(server: str, tmp_path: Path) -> None:
-    _Handler.sse = False
-    _Handler.tool_text = f"merged into {FULL_ID}\nstart: skipped"
-    ref = _client(server, tmp_path).add_task("r/x", "t", "b")
-    assert (ref.task_id, ref.merged) == (FULL_ID, True)
-
-
-def test_tool_error_is_reported(server: str, tmp_path: Path) -> None:
-    _Handler.is_error = True
-    _Handler.tool_text = "boom"
-    with pytest.raises(TodoError, match="boom"):
-        _client(server, tmp_path).add_task("r/x", "t", "b")
-
-
-def test_unexpected_reply_text(server: str, tmp_path: Path) -> None:
-    _Handler.tool_text = "who knows"
-    with pytest.raises(TodoError, match="who knows"):
-        _client(server, tmp_path).add_task("r/x", "t", "b")
-
-
-def test_http_error(server: str, tmp_path: Path) -> None:
-    _Handler.status = 403
-    with pytest.raises(TodoError, match="403"):
-        _client(server, tmp_path).add_task("r/x", "t", "b")
-
-
-def test_unreachable_server(tmp_path: Path) -> None:
-    with pytest.raises(TodoError, match="unreachable"):
-        _client("http://127.0.0.1:9/mcp", tmp_path).add_task("r/x", "t", "b")
-
-
-def test_missing_token(tmp_path: Path) -> None:
-    client = McpTodoClient("http://127.0.0.1:9/mcp", tmp_path / "none")
-    with pytest.raises(TodoError, match="token"):
-        client.add_task("r/x", "t", "b")
-
-
-def test_missing_session_header(server: str, tmp_path: Path) -> None:
-    class _NoSession(_Handler):
-        def do_POST(self) -> None:
-            self.rfile.read(int(self.headers["Content-Length"]))
-            self.send_response(200)
-            self.end_headers()
-            self.wfile.write(b"{}")
-
-    httpd = HTTPServer(("127.0.0.1", 0), _NoSession)
-    thread = threading.Thread(target=httpd.serve_forever, daemon=True)
-    thread.start()
-    try:
-        client = _client(f"http://127.0.0.1:{httpd.server_port}/mcp", tmp_path)
-        with pytest.raises(TodoError, match="session"):
-            client.add_task("r/x", "t", "b")
-    finally:
-        httpd.shutdown()
-        httpd.server_close()
-        thread.join()
-
-
-def test_garbage_json(server: str, tmp_path: Path) -> None:
-    _Handler.sse = False
-    _Handler.tool_text = "x"
-
-    class _Bad(_Handler):
-        def do_POST(self) -> None:
-            body = json.loads(self.rfile.read(int(self.headers["Content-Length"])))
-            self.send_response(200)
-            self.send_header("mcp-session-id", "s")
-            self.end_headers()
-            if body.get("method") == "tools/call":
-                self.wfile.write(b"not json")
-
-    httpd = HTTPServer(("127.0.0.1", 0), _Bad)
-    thread = threading.Thread(target=httpd.serve_forever, daemon=True)
-    thread.start()
-    try:
-        client = _client(f"http://127.0.0.1:{httpd.server_port}/mcp", tmp_path)
-        with pytest.raises(TodoError, match="unparsable"):
-            client.add_task("r/x", "t", "b")
-    finally:
-        httpd.shutdown()
-        httpd.server_close()
-        thread.join()
-
-
-def test_update_task_replaces_title_and_body(server: str, tmp_path: Path) -> None:
-    _Handler.tool_text = f"updated {FULL_ID}\nline"
-    _client(server, tmp_path).update_task(FULL_ID, "MANUAL WORK REQUIRED: x", "b")
-    call = _Handler.seen[2]["params"]
-    assert call == {
-        "name": "update_task",
-        "arguments": {"id": FULL_ID, "title": "MANUAL WORK REQUIRED: x", "body": "b"},
-    }
-
-
-def test_update_task_refused(server: str, tmp_path: Path) -> None:
-    _Handler.tool_text = "no such task"
-    with pytest.raises(TodoError, match="no such task"):
-        _client(server, tmp_path).update_task(FULL_ID, "t", "b")

feedback_inbox/feedback_inbox/tests/test_units.py

@@ -1,97 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The path-unit drop-ins follow the set of game folders, and only then reload."""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-import pytest
-
-from feedback_inbox._units import (
-    SYNC_PATH,
-    WATCH_PATH,
-    desired,
-    run_systemctl,
-    sync_units,
-)
-from feedback_inbox.tests.test_support import make_bundle
-
-if TYPE_CHECKING:
-    from collections.abc import Sequence
-    from pathlib import Path
-
-    from feedback_inbox._paths import Paths
-
-
-class _Systemctl:
-    def __init__(self, code: int = 0) -> None:
-        self.calls: list[list[str]] = []
-        self.code = code
-
-    def __call__(self, args: Sequence[str]) -> int:
-        self.calls.append(list(args))
-        return self.code
-
-
-def test_desired_lists_feedback_and_binaries_dirs(paths: Paths) -> None:
-    make_bundle(paths.data_root, game="cleanup-plate")
-    (paths.data_root / "dread-grid_binaries").mkdir()
-    wanted = desired(paths.data_root, paths.unit_dir)
-    watch = wanted[paths.unit_dir / f"{WATCH_PATH}.d" / "dirs.conf"]
-    sync = wanted[paths.unit_dir / f"{SYNC_PATH}.d" / "dirs.conf"]
-    feedback = paths.data_root / "cleanup-plate_binaries" / "feedback"
-    assert f"PathChanged={feedback}\n" in watch
-    assert "dread-grid" not in watch
-    assert f"PathChanged={paths.data_root}/cleanup-plate_binaries\n" in sync
-    assert f"PathChanged={paths.data_root}/dread-grid_binaries\n" in sync
-
-
-def test_sync_writes_reloads_and_restarts_once(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    ctl = _Systemctl()
-    assert sync_units(paths.data_root, paths.unit_dir, ctl) is True
-    assert ctl.calls == [
-        ["daemon-reload"],
-        ["restart", WATCH_PATH, SYNC_PATH],
-    ]
-    again = _Systemctl()
-    assert sync_units(paths.data_root, paths.unit_dir, again) is False
-    assert again.calls == []
-
-
-def test_new_game_triggers_a_rewrite(paths: Paths) -> None:
-    make_bundle(paths.data_root, game="cleanup-plate")
-    sync_units(paths.data_root, paths.unit_dir, _Systemctl())
-    make_bundle(paths.data_root, game="dread-grid")
-    ctl = _Systemctl()
-    assert sync_units(paths.data_root, paths.unit_dir, ctl) is True
-    text = (paths.unit_dir / f"{WATCH_PATH}.d" / "dirs.conf").read_text(
-        encoding="utf-8"
-    )
-    assert "dread-grid_binaries/feedback" in text
-
-
-def test_systemctl_failure_is_an_error(paths: Paths) -> None:
-    make_bundle(paths.data_root)
-    ctl = _Systemctl(code=1)
-    with pytest.raises(RuntimeError, match="daemon-reload"):
-        sync_units(paths.data_root, paths.unit_dir, ctl)
-    assert ctl.calls == [["daemon-reload"]]
-
-
-def test_run_systemctl_uses_the_user_manager(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Path
-) -> None:
-    seen: list[list[str]] = []
-
-    class _Done:
-        returncode = 3
-
-    def fake_run(cmd: list[str], *, check: bool) -> _Done:
-        assert check is False
-        seen.append(cmd)
-        return _Done()
-
-    monkeypatch.setattr("feedback_inbox._units.subprocess.run", fake_run)
-    assert run_systemctl(["restart", "x"]) == 3
-    assert seen == [["/usr/bin/systemctl", "--user", "restart", "x"]]

feedback_inbox/install.sh

@@ -1,87 +0,0 @@
-#!/bin/bash
-
-# ============================================================================
-# Install feedback-inbox for the current user: its own venv, the CLI on PATH,
-# a todo-server token, and the systemd user units that watch every game's
-# ~/data/<game>_binaries/feedback/ folder. Idempotent; rerun after any change.
-#
-# Enabling the units runs the inbox once (the sweep timer fires at once), so
-# every bundle already on disk becomes a task: that is the backfill.
-# Run the test suite first (scripts/run_subproject_tests.sh feedback_inbox).
-# ============================================================================
-
-set -euo pipefail
-
-REPO_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-readonly REPO_DIR
-readonly STATE_DIR="${HOME}/.local/share/feedback-inbox"
-readonly VENV_DIR="${STATE_DIR}/venv"
-readonly BIN_DIR="${HOME}/.local/bin"
-readonly UNIT_DIR="${HOME}/.config/systemd/user"
-readonly CONFIG_DIR="${HOME}/.config/feedback-inbox"
-readonly TOKEN_FILE="${CONFIG_DIR}/token"
-readonly MINT="${HOME}/src/todo/mcp/mint_token.sh"
-readonly SYSTEM_PYTHON="/usr/bin/python3"
-readonly UNITS=(
-    feedback-inbox.service feedback-inbox.path
-    feedback-inbox-sync.service feedback-inbox-sync.path feedback-inbox-sync.timer
-)
-
-log() { printf 'install: %s\n' "$1" >&2; }
-fail() { printf 'install: FAILED -- %s\n' "$1" >&2; exit 1; }
-
-build_venv() {
-    log "building the venv at $VENV_DIR"
-    mkdir -p "$STATE_DIR"
-    [[ -x "$VENV_DIR/bin/python" ]] || "$SYSTEM_PYTHON" -m venv "$VENV_DIR" \
-        || fail "could not create the venv"
-    "$VENV_DIR/bin/pip" install --quiet --no-deps -e "$REPO_DIR" \
-        || fail "feedback-inbox itself (stdlib only, no other packages)"
-    mkdir -p "$BIN_DIR"
-    ln -sf "$VENV_DIR/bin/feedback-inbox" "$BIN_DIR/feedback-inbox"
-}
-
-ensure_token() {
-    # The service has no shell, so TODO_MCP_TOKEN from ~/.zshenv is invisible
-    # to it. It gets its own revocable token (delete its line in
-    # ~/.config/todo-mcp/tokens); the value goes straight to a 0600 file.
-    [[ -s "$TOKEN_FILE" ]] && return 0
-    [[ -x "$MINT" ]] || fail "$MINT is missing (the todo repo)"
-    log "minting a todo-server token for feedback-inbox"
-    mkdir -p "$CONFIG_DIR"
-    (umask 077 && "$MINT" feedback-inbox >"$TOKEN_FILE") || fail "token mint"
-    [[ -s "$TOKEN_FILE" ]] || fail "token file is empty"
-}
-
-install_units() {
-    log "installing and enabling the units"
-    mkdir -p "$UNIT_DIR"
-    local unit
-    for unit in "${UNITS[@]}"; do
-        install -m 644 "$REPO_DIR/systemd/$unit" "$UNIT_DIR/"
-    done
-    systemctl --user daemon-reload || fail "daemon-reload"
-    # Drop-ins first, so the paths are armed with every current game folder.
-    "$VENV_DIR/bin/feedback-inbox" sync || fail "writing the watch lists"
-    systemctl --user enable --now feedback-inbox.path feedback-inbox-sync.path \
-        feedback-inbox-sync.timer || fail "enabling the units"
-}
-
-verify() {
-    local unit
-    for unit in feedback-inbox.path feedback-inbox-sync.path feedback-inbox-sync.timer; do
-        systemctl --user is-active --quiet "$unit" || fail "$unit is not active"
-    done
-    local folders
-    folders="$(systemctl --user show feedback-inbox.path -p Paths | grep -c '/feedback (PathChanged)' || true)"
-    log "ok: watching ${folders} feedback folder(s)"
-}
-
-main() {
-    build_venv
-    ensure_token
-    install_units
-    verify
-}
-
-main "$@"

feedback_inbox/systemd/feedback-inbox-sync.path

@@ -1,12 +0,0 @@
-[Unit]
-Description=feedback-inbox: a game folder appeared or went away
-
-[Path]
-# ~/data fires when a *_binaries folder is created; the generated drop-in
-# feedback-inbox-sync.path.d/dirs.conf adds each existing *_binaries folder, so
-# a game's later `feedback/` subfolder is seen too.
-PathChanged=%h/data
-Unit=feedback-inbox-sync.service
-
-[Install]
-WantedBy=default.target

feedback_inbox/systemd/feedback-inbox-sync.service

@@ -1,12 +0,0 @@
-[Unit]
-Description=feedback-inbox: re-arm the watches for new games, then import
-Documentation=file:///home/kuhy/src/utils/feedback_inbox/README.md
-StartLimitIntervalSec=0
-
-[Service]
-Type=oneshot
-WorkingDirectory=/
-ExecStart=%h/.local/share/feedback-inbox/venv/bin/feedback-inbox sync
-ExecStart=%h/.local/share/feedback-inbox/venv/bin/feedback-inbox run --wait 0
-Nice=10
-TimeoutStartSec=2min

feedback_inbox/systemd/feedback-inbox-sync.timer

@@ -1,12 +0,0 @@
-[Unit]
-Description=feedback-inbox: fallback sweep for events the path units missed
-
-[Timer]
-# Events that land while a run is active are lost by .path units; this catches
-# them. No Persistent=: a missed sweep is harmless, the next one runs soon.
-OnBootSec=2min
-OnUnitActiveSec=5min
-Unit=feedback-inbox-sync.service
-
-[Install]
-WantedBy=timers.target

feedback_inbox/systemd/feedback-inbox.path

@@ -1,12 +0,0 @@
-[Unit]
-Description=feedback-inbox: watch every game's feedback folder
-
-[Path]
-# systemd paths take no globs. The folders themselves are listed in the
-# generated drop-in feedback-inbox.path.d/dirs.conf (`feedback-inbox sync`);
-# this baseline is the token file, which exists once installed and never churns.
-PathChanged=%h/.config/feedback-inbox/token
-Unit=feedback-inbox.service
-
-[Install]
-WantedBy=default.target

feedback_inbox/systemd/feedback-inbox.service

@@ -1,15 +0,0 @@
-[Unit]
-Description=feedback-inbox: turn new game feedback bundles into todo tasks
-Documentation=file:///home/kuhy/src/utils/feedback_inbox/README.md
-# The .path unit fires on every write in a feedback folder; bursts (a bundle is
-# written in pieces) must not trip the default start limit.
-StartLimitIntervalSec=0
-
-[Service]
-Type=oneshot
-WorkingDirectory=/
-# Waits up to 15 s for a half-written bundle to finish before leaving it to the
-# timer. Exits non-zero when the todo server refuses, so the failure is visible.
-ExecStart=%h/.local/share/feedback-inbox/venv/bin/feedback-inbox run --wait 15
-Nice=10
-TimeoutStartSec=2min

focus_project/README.md

@@ -36,18 +36,5 @@ uncovers today. A NEW project starting today may replace gates and set any targe
 new obligation, not a reduction (but see the floor).
 Floor (0.7.1, coordinator decision for kuhy): `daily_minutes >= 30 x replaced gates` (min 30; `book` = `reading`),
 enforced on create and update (`daily_floor(replaces)`), so a day cannot be made cheap.
-Backed timer (0.10.0, days from `BACKING_FROM` = 2026-11-08): a timer span earns only if a
-same-project `tutor`/`sql_drills`/`leetcode` beat (or accepted submission) is in it, 5 min
-before it, or bridges it (beats <= 20 min apart). `unbacked_minutes_for(day)` /
-`status()["unbacked_minutes"]` show the rest. A beat backs only the spans it touches (the tutor chain makes a graded card back the whole stretch). Known gap: any coach turn backs.
 Open the work: `python3 -m focus_project.open [--no-site]` (tutor focus screen, `work_url`).
 Install: `./install.sh` (stdlib only, editable user install for `/usr/bin/python3`).
-
-Interview prep (`kind="interview"`, 0.9): `create_project(..., kind="interview", test_at=, invite_url=, trips=, weekend_answers=)`;
-`end` must be the test day. Budget: 30 min weekdays / 120 Sat-Sun (`daily_on`), exempt from `daily_floor`, unlock clamped.
-Milestones are generated, never stored: `milestones(project, now)` -> `rules`, `trip_ask:<sat>`, `phone:<trip start>`, `mock`, `slot`
-with states pending/done/missed/late_done/unknown. `owed_lock(now)` = earliest overdue uncompleted one (never on test day; wire it
-after focus-guard's sick/free-day stand-down). `status(day)` gains `milestones` and `missed`.
-Evidence is append-only (`evidence/<id>.jsonl` + copied files): `record_evidence(id, milestone, kind, value, file=, now=)`
-(`file`, `tap`, `score`, `slot`, `slot_confirm`, tutor's `phone_check`). `add_trip` / `answer_weekend` only get stricter (ValueError).
-`run_checks(id)` runs the `phone` source (reads the latest `phone_check` row) and latches changes; `register_source` swaps it.

focus_project/focus_project/__init__.py

@@ -23,44 +23,26 @@ from focus_project._api import (
     stands_down,
 )
 from focus_project._errors import RegistryError
-from focus_project._milestones import Milestone, milestones, owed_lock
 from focus_project._model import Limits, Project
-from focus_project._mutate import add_trip, answer_weekend, record_evidence
-from focus_project._trips import Trip
 from focus_project._update import Update, apply_update, update_project
 from focus_project.activity import record_activity
-from focus_project.checks import register_source, run_checks
+from focus_project.minutes import minutes_for
-from focus_project.evidence import Evidence
-from focus_project.minutes import minutes_for, unbacked_minutes_for
-from focus_project.sources.activity import BACKING_FROM
 from focus_project.status import status
 
 __all__ = [
-    "BACKING_FROM",
-    "Evidence",
     "Limits",
-    "Milestone",
     "Project",
     "RegistryError",
-    "Trip",
     "Update",
     "active",
-    "add_trip",
-    "answer_weekend",
     "apply_update",
     "create_project",
     "daily_floor",
     "end_project",
     "list_projects",
-    "milestones",
     "minutes_for",
-    "owed_lock",
     "record_activity",
-    "record_evidence",
-    "register_source",
-    "run_checks",
     "stands_down",
     "status",
-    "unbacked_minutes_for",
     "update_project",
 ]

focus_project/focus_project/_api.py

@@ -4,24 +4,19 @@
 from __future__ import annotations
 
 from dataclasses import replace
+from datetime import date  # noqa: TC003
 import logging
 import re
 import time
-from typing import TYPE_CHECKING
+from urllib.parse import urlsplit
 
-from focus_project import _clock, _interview, _store
+from focus_project import _clock, _store
 from focus_project._errors import RegistryError
-from focus_project._kind import FOCUS
 from focus_project._model import GATES, Project, normalize_gate
 
-if TYPE_CHECKING:
-    from datetime import date, datetime
-
-    from focus_project._trips import Trip
-
 _LOG = logging.getLogger(__name__)
 _ID = re.compile(r"[a-z0-9][a-z0-9-]*")
-MAX_WORK_URL = _interview.MAX_URL
+MAX_WORK_URL = 2000
 # A project must ask real work: the focus earner pays per minute of ``daily``.
 MIN_DAILY_PER_GATE = 30
 
@@ -58,7 +53,17 @@ def stands_down(gate: str, day: date) -> bool:
 
 def _validate_work_url(url: str) -> None:
     """Empty, or a plain http(s) URL with a host: it goes to ``xdg-open``."""
-    _interview.check_http_url(url, "work_url")
+    if not url:
+        return
+    parts = urlsplit(url) if url.isprintable() and " " not in url else None
+    if (
+        parts is None
+        or len(url) > MAX_WORK_URL
+        or parts.scheme not in {"http", "https"}
+        or not parts.netloc
+    ):
+        msg = "work_url must be an http(s) URL without spaces (or empty)"
+        raise ValueError(msg)
 
 
 def daily_floor(replaces: tuple[str, ...]) -> int:
@@ -80,9 +85,8 @@ def _validate(project: Project) -> None:
     if not project.replaces or not set(project.replaces) <= GATES:
         msg = f"replaces must be a non-empty subset of gates {sorted(GATES)}"
         raise ValueError(msg)
-    _interview.validate(project)
     floor = daily_floor(project.replaces)
-    if project.kind == FOCUS and project.daily_minutes < floor:
+    if project.daily_minutes < floor:
         msg = (
             f"daily_minutes must be at least {floor} "
             f"({MIN_DAILY_PER_GATE} per replaced gate)"
@@ -105,19 +109,8 @@ def create_project(  # noqa: PLR0913
     subjects: list[str] | None = None,
     site_patterns: list[str] | None = None,
     work_url: str = "",
-    kind: str = FOCUS,
-    test_at: datetime | None = None,
-    invite_url: str = "",
-    trips: list[Trip] | None = None,
-    weekend_answers: dict[date, bool] | None = None,
 ) -> Project:
-    """Add a project; ValueError on bad input, a duplicate id or an overlap.
+    """Add a project; ValueError on bad input, a duplicate id or an overlap."""
-
-    ``kind="interview"`` takes ``test_at`` (naive local; its date must be
-    ``end``), ``invite_url``, ``trips`` and ``weekend_answers``; an interview
-    is exempt from :func:`daily_floor` and runs the 30/120 budget.
-    """
-    created_ms = time.time_ns() // 1_000_000
     project = Project(
         id=project_id,
         name=name,
@@ -129,15 +122,7 @@ def create_project(  # noqa: PLR0913
         subjects=tuple(subjects or ()),
         site_patterns=tuple(site_patterns or ()),
         work_url=work_url,
-        created_ms=created_ms,
+        created_ms=time.time_ns() // 1_000_000,
-        kind=kind,
-        test_at=test_at,
-        invite_url=invite_url,
-        trips=tuple(
-            t if t.entered_ms else replace(t, entered_ms=created_ms)
-            for t in trips or ()
-        ),
-        weekend_answers=tuple(sorted((weekend_answers or {}).items())),
     )
     _validate(project)
     with _store.locked():

focus_project/focus_project/_interview.py

@@ -1,89 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Validation of the interview-only fields."""
-
-from __future__ import annotations
-
-from datetime import date, timedelta
-from typing import TYPE_CHECKING
-from urllib.parse import urlsplit
-
-from focus_project._kind import FOCUS, KINDS, WEEKDAY_MINUTES
-
-if TYPE_CHECKING:
-    from focus_project._model import Project
-
-MAX_URL = 2000
-SATURDAY = 5
-
-
-def check_http_url(url: str, label: str) -> None:
-    """Empty, or a plain http(s) URL with a host: it goes to ``xdg-open``."""
-    if not url:
-        return
-    parts = urlsplit(url) if url.isprintable() and " " not in url else None
-    if (
-        parts is None
-        or len(url) > MAX_URL
-        or parts.scheme not in {"http", "https"}
-        or not parts.netloc
-    ):
-        msg = f"{label} must be an http(s) URL without spaces (or empty)"
-        raise ValueError(msg)
-
-
-def saturdays(project: Project) -> list[date]:
-    """The Saturdays whose weekend lies inside D1..T-1."""
-    if project.test_at is None:
-        return []
-    last = project.test_at.date() - timedelta(days=1)
-    day = project.start + timedelta(days=(SATURDAY - project.start.weekday()) % 7)
-    found: list[date] = []
-    while day <= last:
-        found.append(day)
-        day += timedelta(days=7)
-    return found
-
-
-def _has_interview_fields(project: Project) -> bool:
-    return bool(
-        project.test_at
-        or project.invite_url
-        or project.trips
-        or project.weekend_answers
-    )
-
-
-def _validate_interview(project: Project) -> None:
-    if project.test_at is None or project.test_at.tzinfo is not None:
-        msg = "an interview needs test_at as a naive local datetime"
-        raise ValueError(msg)
-    test_day = project.test_at.date()
-    if project.end != test_day:
-        msg = f"end must be the test day {test_day}"
-        raise ValueError(msg)
-    if project.daily_minutes < WEEKDAY_MINUTES:
-        msg = f"daily_minutes must be at least {WEEKDAY_MINUTES}"
-        raise ValueError(msg)
-    for trip in project.trips:
-        if trip.end < trip.start or trip.start > test_day:
-            msg = f"trip {trip.start}..{trip.end} must start by the test day"
-            raise ValueError(msg)
-    weekends = saturdays(project)
-    for day, _ in project.weekend_answers:
-        if day not in weekends:
-            msg = f"{day} is not a weekend inside the project before the test day"
-            raise ValueError(msg)
-
-
-def validate(project: Project) -> None:
-    """ValueError unless the kind and its interview fields are coherent."""
-    if project.kind not in KINDS:
-        msg = f"kind must be one of {sorted(KINDS)}"
-        raise ValueError(msg)
-    check_http_url(project.invite_url, "invite_url")
-    if project.kind == FOCUS:
-        if _has_interview_fields(project):
-            msg = "test_at, invite_url, trips, weekend_answers need kind='interview'"
-            raise ValueError(msg)
-        return
-    _validate_interview(project)

focus_project/focus_project/_kind.py

@@ -1,23 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Project kinds and the interview budget (the one place the numbers live)."""
-
-from __future__ import annotations
-
-from typing import TYPE_CHECKING
-
-if TYPE_CHECKING:
-    from datetime import date
-
-FOCUS = "focus"
-INTERVIEW = "interview"
-KINDS = frozenset({FOCUS, INTERVIEW})
-
-# kuhy, 2026-11-10: near-zero weekday energy, weekends are for learning.
-WEEKDAY_MINUTES = 30
-WEEKEND_MINUTES = 120
-_FIRST_WEEKEND_DAY = 5  # Saturday
-
-
-def interview_minutes(day: date) -> int:
-    """The daily target of an interview project on ``day`` (Sat/Sun = 120)."""
-    return WEEKEND_MINUTES if day.weekday() >= _FIRST_WEEKEND_DAY else WEEKDAY_MINUTES

focus_project/focus_project/_milestones.py

@@ -1,235 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Interview milestones: generated from the project, judged from evidence.
-
-Nothing here is stored or editable. ``due`` and the state are recomputed from
-the project (kind, test_at, trips, weekend answers) and the append-only
-evidence rows, so a milestone cannot be moved or ticked off by editing a row.
-
-Failure modes (R8): an unanswered trip question reads as a trip (fail
-closed); a check that could not run is ``unknown`` and counts as not done
-(fail closed); an unreadable registry or evidence file arms no lock (fail
-open, logged), matching focus-guard's "unreadable registry" rule.
-"""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-from datetime import date, datetime, time, timedelta
-import logging
-from typing import TYPE_CHECKING, Any, Literal
-
-from focus_project import _api, _interview, evidence
-from focus_project._errors import RegistryError
-from focus_project._kind import INTERVIEW
-from focus_project.sources import activity
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-
-    from focus_project._model import Project
-
-_LOG = logging.getLogger(__name__)
-State = Literal["pending", "done", "missed", "late_done", "unknown"]
-PHONE_LEAD = timedelta(hours=12)
-MOCK_SESSION_S = 90 * 60
-_RED = frozenset({"missed", "unknown"})
-
-
-@dataclass(frozen=True)
-class Milestone:
-    """One generated obligation of an interview project."""
-
-    id: str
-    due: datetime
-    state: State
-    completed: bool
-
-    @property
-    def red(self) -> bool:
-        """Past due and not done (also: the check could not run)."""
-        return self.state in _RED
-
-    def to_dict(self) -> dict[str, Any]:
-        """The status() row: id, due ISO, state."""
-        return {"id": self.id, "due": self.due.isoformat(), "state": self.state}
-
-
-def _at(day: date, hour: int, minute: int = 0) -> datetime:
-    return datetime.combine(day, time(hour, minute))
-
-
-def _trip_due(trip_start: date, entered_ms: int) -> datetime:
-    """Start - 12 h, or the moment it was entered if that left < 12 h."""
-    nominal = _at(trip_start, 0) - PHONE_LEAD
-    if entered_ms > 0:
-        entered = evidence.from_ms(entered_ms)
-        if nominal - entered < PHONE_LEAD:
-            return entered
-    return nominal
-
-
-def _is_trip(
-    *,
-    answered: bool | None,
-    answer_ms: int | None,
-    ask_due: datetime,
-    phone_due: datetime,
-    now: datetime,
-) -> bool:
-    """Whether the weekend counts as a trip (fail closed).
-
-    Unanswered once the question is due = a trip. A no-trip answer given
-    only after the phone prep was already due cannot erase that overdue
-    milestone: the answer row's own timestamp decides, not a stored flag.
-    """
-    if answered is None:
-        return now >= ask_due
-    if answered:
-        return True
-    return answer_ms is not None and answer_ms >= evidence.to_ms(phone_due)
-
-
-def plan(
-    project: Project, test_day: date, now: datetime, rows: list[evidence.Evidence]
-) -> dict[str, datetime]:
-    """Milestone id -> due; ``project`` must be a validated interview."""
-    dues = {
-        "rules": _at(project.start, 23, 59),
-        "mock": _at(test_day - timedelta(days=2), 23, 59),
-        "slot": _at(test_day - timedelta(days=1), 12),
-    }
-    answers = dict(project.weekend_answers)
-    for sat in _interview.saturdays(project):
-        ask_due = _at(sat - timedelta(days=2), 23, 59)
-        phone_due = _at(sat, 0) - PHONE_LEAD
-        dues[f"trip_ask:{sat}"] = ask_due
-        covered = any(t.covers(sat) for t in project.trips)
-        row = evidence.first(rows, f"trip_ask:{sat}", "answer")
-        answer_ms = None if row is None else row.ms
-        if not covered and _is_trip(
-            answered=answers.get(sat),
-            answer_ms=answer_ms,
-            ask_due=ask_due,
-            phone_due=phone_due,
-            now=now,
-        ):
-            dues[f"phone:{sat}"] = phone_due
-    for trip_row in project.trips:
-        dues[f"phone:{trip_row.start}"] = _trip_due(trip_row.start, trip_row.entered_ms)
-    return dues
-
-
-def _when(row: evidence.Evidence | None) -> datetime | None:
-    return None if row is None else row.at
-
-
-Rows = list[evidence.Evidence]
-Completion = tuple[datetime | None, bool]
-
-
-def _by_evidence(kind: str) -> Callable[[str, Rows, Project], Completion]:
-    """Done at the first evidence row of ``kind`` (rules file, mock score)."""
-
-    def judge(mid: str, rows: Rows, _project: Project) -> Completion:
-        return _when(evidence.first(rows, mid, kind)), False
-
-    return judge
-
-
-def _mock(mid: str, rows: Rows, project: Project) -> Completion:
-    """Done at the first score row that follows a >= 90 min focus session.
-
-    The session (contiguous beats, :func:`activity.sessions`) must END at or
-    before the score row; a score with no such session completes nothing.
-    """
-    ends = [hi for lo, hi in activity.sessions(project.id) if hi - lo >= MOCK_SESSION_S]
-    for row in rows:
-        if (row.milestone, row.kind) == (mid, "score") and any(
-            end <= row.ms / 1000 for end in ends
-        ):
-            return row.at, False
-    return None, False
-
-
-def _trip_ask(mid: str, rows: Rows, project: Project) -> Completion:
-    """Done at the answer row; an answer with no row was given at creation."""
-    if date.fromisoformat(mid.partition(":")[2]) not in dict(project.weekend_answers):
-        return None, False
-    return _when(evidence.first(rows, mid, "answer")) or evidence.from_ms(0), False
-
-
-def _slot(mid: str, rows: Rows, _project: Project) -> Completion:
-    """Done once the latest slot entry has a confirm tap at or after it."""
-    slot = evidence.last(rows, mid, "slot")
-    if slot is None:
-        return None, False
-    confirm = evidence.first(rows, mid, "slot_confirm", since_ms=slot.ms)
-    return (None if confirm is None else max(slot.at, confirm.at)), False
-
-
-def _phone(mid: str, rows: Rows, _project: Project) -> Completion:
-    """Done when the offline tap and a passing auto check both exist.
-
-    A tap with no verdict at all (or a dark source) is ``unknown``.
-    """
-    tap = evidence.first(rows, mid, "tap")
-    passed, latest, _ = evidence.check_summary(rows, mid)
-    if tap is not None and passed is not None:
-        return max(tap.at, evidence.from_ms(passed)), False
-    return None, tap is not None and latest is None
-
-
-_JUDGES = {
-    "rules": _by_evidence("file"),
-    "mock": _mock,
-    "trip_ask": _trip_ask,
-    "slot": _slot,
-    "phone": _phone,
-}
-
-
-def _judge(
-    mid: str, due: datetime, rows: Rows, project: Project, now: datetime
-) -> Milestone:
-    done_at, unknown = _JUDGES[mid.partition(":")[0]](mid, rows, project)
-    if done_at is not None:
-        state: State = "done" if done_at <= due else "late_done"
-        return Milestone(mid, due, state, completed=True)
-    if now < due:
-        return Milestone(mid, due, "pending", completed=False)
-    return Milestone(mid, due, "unknown" if unknown else "missed", completed=False)
-
-
-def milestones(project: Project, now: datetime) -> list[Milestone]:
-    """The project's milestones at ``now`` (naive local), earliest due first.
-
-    Empty for a focus project. May raise OSError reading the evidence file.
-    """
-    if project.kind != INTERVIEW or project.test_at is None:
-        return []
-    rows = evidence.read(project.id)
-    dues = plan(project, project.test_at.date(), now, rows)
-    found = [_judge(mid, due, rows, project, now) for mid, due in dues.items()]
-    return sorted(found, key=lambda m: (m.due, m.id))
-
-
-def owed_lock(now: datetime) -> Milestone | None:
-    """The earliest overdue, uncompleted milestone that locks, or None.
-
-    Locks (missed or unknown) from the moment it is due, every day of the
-    project except the test day itself, where nothing locks. It does not
-    consult sickdays/freedays: focus-guard's ``_off(today)`` precondition
-    already stands the lock down on those days, so wire this after it.
-    Unreadable state arms nothing (fail open, logged).
-    """
-    try:
-        project = _api.active(now.date())
-        if project is None or project.test_at is None:
-            return None
-        if now.date() >= project.test_at.date():
-            return None
-        owed = [m for m in milestones(project, now) if not m.completed and now >= m.due]
-    except (RegistryError, OSError) as exc:
-        _LOG.warning("interview state unreadable, no milestone lock: %s", exc)
-        return None
-    return owed[0] if owed else None

focus_project/focus_project/_model.py

@@ -4,12 +4,10 @@
 from __future__ import annotations
 
 from dataclasses import dataclass
-from datetime import date, datetime
+from datetime import date
 from typing import Any
 
 from focus_project._errors import RegistryError
-from focus_project._kind import FOCUS, INTERVIEW, interview_minutes
-from focus_project._trips import Trip
 
 GATES = frozenset({"leetcode", "reading", "tutor"})
 _ALIASES = {"book": "reading"}
@@ -20,17 +18,6 @@ def normalize_gate(gate: str) -> str:
     return _ALIASES.get(gate, gate)
 
 
-def _answers(raw: dict[str, Any] | None) -> tuple[tuple[date, bool], ...]:
-    """Parse the saturday -> bool map; a non-bool answer is a bad row."""
-    pairs = []
-    for key, value in (raw or {}).items():
-        if not isinstance(value, bool):
-            msg = f"weekend answer for {key} is not true/false"
-            raise TypeError(msg)
-        pairs.append((date.fromisoformat(key), value))
-    return tuple(sorted(pairs))
-
-
 @dataclass(frozen=True)
 class Limits:
     """The minute targets that held up to and including ``until``."""
@@ -83,30 +70,19 @@ class Project:
     work_url: str = ""
     # Earlier targets, oldest first: an edit applies from its day, never back.
     history: tuple[Limits, ...] = ()
-    # "focus" (every pre-interview row) or "interview"; the rest is interview-only.
-    kind: str = FOCUS
-    # Local naive datetime of the test; its date is the project's last day.
-    test_at: datetime | None = None
-    invite_url: str = ""
-    trips: tuple[Trip, ...] = ()
-    # Saturday -> True (trip that weekend) / False (no trip); sorted pairs.
-    weekend_answers: tuple[tuple[date, bool], ...] = ()
 
     def _limits_on(self, day: date) -> Limits | None:
         return next((h for h in self.history if day <= h.until), None)
 
     def daily_on(self, day: date) -> int:
         """The daily target that held on ``day``."""
-        if self.kind == INTERVIEW:
-            return interview_minutes(day)
         past = self._limits_on(day)
         return self.daily_minutes if past is None else past.daily_minutes
 
     def unlock_on(self, day: date) -> int:
-        """The unlock target that held on ``day`` (clamped for an interview)."""
+        """The unlock target that held on ``day``."""
         past = self._limits_on(day)
-        held = self.unlock_minutes if past is None else past.unlock_minutes
+        return self.unlock_minutes if past is None else past.unlock_minutes
-        return min(held, interview_minutes(day)) if self.kind == INTERVIEW else held
 
     def replaces_on(self, day: date) -> tuple[str, ...]:
         """The gates this project switched off on ``day``."""
@@ -128,7 +104,7 @@ class Project:
 
     def to_dict(self) -> dict[str, Any]:
         """The projects.json row."""
-        row: dict[str, Any] = {
+        return {
             "id": self.id,
             "name": self.name,
             "start": self.start.isoformat(),
@@ -142,23 +118,13 @@ class Project:
             "created_ms": self.created_ms,
             "work_url": self.work_url,
             "history": [h.to_dict() for h in self.history],
-            "kind": self.kind,
         }
-        if self.kind == INTERVIEW:
-            row["test_at"] = self.test_at.isoformat() if self.test_at else None
-            row["invite_url"] = self.invite_url
-            row["trips"] = [t.to_dict() for t in self.trips]
-            row["weekend_answers"] = {
-                day.isoformat(): trip for day, trip in self.weekend_answers
-            }
-        return row
 
     @classmethod
     def from_dict(cls, row: Any) -> Project:  # noqa: ANN401
         """Parse a registry row, or raise :class:`RegistryError`."""
         try:
             ended = row.get("ended_on")
-            test_at = row.get("test_at")
             return cls(
                 id=str(row["id"]),
                 name=str(row["name"]),
@@ -173,11 +139,6 @@ class Project:
                 created_ms=int(row.get("created_ms", 0)),
                 work_url=str(row.get("work_url") or ""),
                 history=tuple(Limits.from_dict(h) for h in row.get("history", ())),
-                kind=str(row.get("kind") or FOCUS),
-                test_at=datetime.fromisoformat(test_at) if test_at else None,
-                invite_url=str(row.get("invite_url") or ""),
-                trips=tuple(Trip.from_dict(t) for t in row.get("trips", ())),
-                weekend_answers=_answers(row.get("weekend_answers")),
             )
         except (KeyError, TypeError, ValueError, AttributeError) as exc:
             msg = f"bad project row {row!r}: {exc}"

focus_project/focus_project/_mutate.py

@@ -1,178 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Interview mutations: trips, weekend answers, evidence. Stricter only."""
-
-from __future__ import annotations
-
-from dataclasses import replace
-from datetime import date, datetime
-import math
-from typing import TYPE_CHECKING
-
-from focus_project import _interview, _milestones, _store, evidence
-from focus_project._kind import INTERVIEW
-from focus_project._trips import Trip, add
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-    from focus_project._model import Project
-
-# Which evidence kinds each milestone family accepts from a caller.
-# ``check`` and ``answer`` are internal: written by run_checks/answer_weekend.
-_ACCEPTS = {
-    "rules": {"file"},
-    "mock": {"score"},
-    "slot": {"slot", "slot_confirm"},
-    "phone": {"tap"},
-}
-# tutor's worker reports phone readiness under the bare id "phone".
-_PHONE_REPORT = "phone"
-
-
-def local_now(now: datetime | None) -> datetime:
-    """``now`` or the clock's local time."""
-    return evidence.clock_now() if now is None else now
-
-
-def find(existing: list[Project], project_id: str) -> tuple[int, Project]:
-    """Index and interview project; KeyError unknown, ValueError focus kind."""
-    for i, project in enumerate(existing):
-        if project.id == project_id:
-            if project.kind != INTERVIEW:
-                msg = f"{project_id!r} is not an interview project (no milestones)"
-                raise ValueError(msg)
-            return i, project
-    raise KeyError(project_id)
-
-
-def add_trip(
-    project_id: str, start: date, end: date, *, now: datetime | None = None
-) -> Project:
-    """Announce a trip; it may only grow (see :func:`focus_project._trips.add`)."""
-    stamp = local_now(now)
-    with _store.locked():
-        existing = _store.read_all()
-        i, project = find(existing, project_id)
-        trips = add(project.trips, Trip(start, end, evidence.to_ms(stamp)))
-        updated = replace(project, trips=trips)
-        _interview.validate(updated)
-        existing[i] = updated
-        _store.write_all(existing)
-    return updated
-
-
-def answer_weekend(
-    project_id: str, saturday: date, *, trip: bool, now: datetime | None = None
-) -> Project:
-    """Answer the trip question of a weekend; True never flips back to False."""
-    stamp = local_now(now)
-    with _store.locked():
-        existing = _store.read_all()
-        i, project = find(existing, project_id)
-        if saturday not in _interview.saturdays(project):
-            msg = f"{saturday} is not a weekend of {project_id!r} before its test day"
-            raise ValueError(msg)
-        answers = dict(project.weekend_answers)
-        current = answers.get(saturday)
-        if current is trip:
-            return project
-        if current is True:
-            msg = "a trip answer may only get stricter: it stays True"
-            raise ValueError(msg)
-        if not trip and any(t.covers(saturday) for t in project.trips):
-            msg = f"an announced trip covers {saturday}: it cannot be a no-trip weekend"
-            raise ValueError(msg)
-        updated = replace(
-            project, weekend_answers=tuple(sorted({**answers, saturday: trip}.items()))
-        )
-        # Evidence first: a crash between the two leaves the question open
-        # (fail closed), never an answer with no time.
-        evidence.append(project_id, f"trip_ask:{saturday}", "answer", trip, stamp)
-        existing[i] = updated
-        _store.write_all(existing)
-    return updated
-
-
-def _number(value: object) -> float:
-    if (
-        isinstance(value, bool)
-        or not isinstance(value, int | float)
-        or not math.isfinite(value)
-        or value < 0
-    ):
-        msg = "a score must be a finite, non-negative number"
-        raise ValueError(msg)
-    return value
-
-
-def _count(value: object, label: str) -> int:
-    if isinstance(value, bool) or not isinstance(value, int) or value < 0:
-        msg = f"{label} must be a non-negative integer"
-        raise ValueError(msg)
-    return value
-
-
-def _clean(
-    project_id: str, kind: str, value: object, file: Path | None, stamp: datetime
-) -> object:
-    """The value to store for ``kind``; ValueError if it is not acceptable."""
-    if kind == "file":
-        if file is None or not file.is_file() or not file.stat().st_size:
-            msg = "a non-empty file is required"
-            raise ValueError(msg)
-        return evidence.store_file(project_id, file, stamp)
-    if kind == "score":
-        return _number(value)
-    if kind == "slot":
-        moment = datetime.fromisoformat(str(value)) if isinstance(value, str) else None
-        if moment is None or moment.tzinfo is not None:
-            msg = "slot must be a naive local ISO datetime"
-            raise ValueError(msg)
-        return moment.isoformat()
-    if kind == "phone_check":
-        if not isinstance(value, dict) or not isinstance(value.get("plan"), bool):
-            msg = "phone_check needs {plan: bool, due: int, seen_ms: int}"
-            raise ValueError(msg)
-        return {
-            "plan": value["plan"],
-            "due": _count(value.get("due"), "due"),
-            "seen_ms": _count(value.get("seen_ms"), "seen_ms"),
-        }
-    if kind == "slot_confirm" and value not in {True, None}:
-        msg = "slot_confirm is a tap: value True"
-        raise ValueError(msg)
-    return True  # tap, slot_confirm
-
-
-def record_evidence(  # noqa: PLR0913
-    project_id: str,
-    milestone: str,
-    kind: str,
-    value: object = None,
-    *,
-    file: Path | None = None,
-    now: datetime | None = None,
-) -> evidence.Evidence:
-    """Append evidence for a milestone; ValueError if it does not fit.
-
-    Kinds: ``file`` (rules, copied from ``file``), ``tap`` (phone:*),
-    ``score`` (mock), ``slot`` (ISO datetime) and ``slot_confirm`` (slot),
-    ``phone_check`` ({plan, due, seen_ms}, milestone "phone", from tutor).
-    There is no way to delete a row.
-    """
-    stamp = local_now(now)
-    _, project = find(_store.read_all(), project_id)
-    if milestone == _PHONE_REPORT:
-        accepted = {"phone_check"}
-    else:
-        known = {m.id for m in _milestones.milestones(project, stamp)}
-        accepted = _ACCEPTS.get(milestone.partition(":")[0], set())
-        if milestone not in known:
-            msg = f"{milestone!r} is not a milestone of {project_id!r}"
-            raise ValueError(msg)
-    if kind not in accepted:
-        msg = f"{milestone!r} takes evidence kinds {sorted(accepted)}, not {kind!r}"
-        raise ValueError(msg)
-    return evidence.append(
-        project_id, milestone, kind, _clean(project_id, kind, value, file, stamp), stamp
-    )

focus_project/focus_project/_paths.py

@@ -30,18 +30,3 @@ def activity_path() -> Path:
 def lock_path() -> Path:
     """The registry's writer lock."""
     return data_dir() / "projects.lock"
-
-
-def evidence_dir() -> Path:
-    """Interview-prep evidence: one jsonl per project plus copied files."""
-    return data_dir() / "evidence"
-
-
-def evidence_path(project_id: str) -> Path:
-    """The append-only evidence rows of one project."""
-    return evidence_dir() / f"{project_id}.jsonl"
-
-
-def evidence_files_dir(project_id: str) -> Path:
-    """Where evidence files (screenshots) of one project are copied."""
-    return evidence_dir() / project_id

focus_project/focus_project/_trips.py

@@ -1,61 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Trips: days away from the PC. They can only grow (ratchet)."""
-
-from __future__ import annotations
-
-from dataclasses import dataclass, replace
-from datetime import date
-from typing import Any
-
-
-@dataclass(frozen=True)
-class Trip:
-    """An inclusive away range; ``entered_ms`` is when it was announced."""
-
-    start: date
-    end: date
-    entered_ms: int = 0
-
-    def covers(self, day: date) -> bool:
-        """Whether ``day`` is inside the trip."""
-        return self.start <= day <= self.end
-
-    def to_dict(self) -> dict[str, Any]:
-        """The projects.json row."""
-        return {
-            "start": self.start.isoformat(),
-            "end": self.end.isoformat(),
-            "entered_ms": self.entered_ms,
-        }
-
-    @classmethod
-    def from_dict(cls, row: dict[str, Any]) -> Trip:
-        """Parse a trip row (errors surface as RegistryError upstream)."""
-        return cls(
-            start=date.fromisoformat(row["start"]),
-            end=date.fromisoformat(row["end"]),
-            entered_ms=int(row.get("entered_ms", 0)),
-        )
-
-
-def add(trips: tuple[Trip, ...], new: Trip) -> tuple[Trip, ...]:
-    """``trips`` plus ``new``; ValueError if that would loosen an old trip.
-
-    A new range touching an old trip must cover it: it replaces the old one
-    (keeping its entry time when the start did not move earlier). Anything
-    that starts later or ends earlier than what is on file is refused.
-    """
-    kept: list[Trip] = []
-    for old in trips:
-        if old.start > new.end or new.start > old.end:
-            kept.append(old)
-        elif new.start <= old.start and new.end >= old.end:
-            if new.start == old.start:
-                new = replace(new, entered_ms=old.entered_ms)
-        else:
-            msg = (
-                f"a trip may only get stricter: pass a range that covers "
-                f"{old.start}..{old.end}"
-            )
-            raise ValueError(msg)
-    return tuple(sorted([*kept, new], key=lambda t: t.start))

focus_project/focus_project/checks.py

@@ -1,106 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Pluggable check sources for the auto-checked milestones, and latching.
-
-A source is ``fn(project, now) -> True | False | None``; ``None`` means "could
-not check" (logged) and counts as not done. Only ``phone`` is auto-checked
-(the calendar slot check was dropped, 2026-11-10: ``slot`` is two taps).
-
-:func:`run_checks` is the only place sources run: status()/milestones() stay
-cheap and read the latched ``check`` rows it appends (only on a change).
-"""
-
-from __future__ import annotations
-
-from collections.abc import Callable
-from datetime import datetime, timedelta
-import logging
-
-from focus_project import _milestones, _mutate, _store, evidence
-from focus_project._model import Project
-
-_LOG = logging.getLogger(__name__)
-CheckSource = Callable[[Project, datetime], bool | None]
-DAY_MS = 24 * 3600 * 1000
-# A phone check starts 24 h before the milestone is due.
-WINDOW = timedelta(hours=24)
-
-
-def _report_ok(value: object, report_ms: int) -> bool:
-    """A plan exists, cards are due, the phone was seen <= 24 h before the report."""
-    if not isinstance(value, dict):
-        return False
-    try:
-        return bool(
-            value["plan"] is True
-            and int(value["due"]) > 0
-            and report_ms - int(value["seen_ms"]) <= DAY_MS
-        )
-    except KeyError, TypeError, ValueError:
-        return False
-
-
-def phone_source(project: Project, now: datetime) -> bool | None:
-    """Read tutor's latest ``phone_check`` report (<= 24 h old).
-
-    No report, a stale one or one stamped ahead of the clock = ``None``
-    (the worker is not reporting). A fresh one passes only if the plan
-    exists, cards are due and the phone was seen within 24 h of the report.
-    """
-    report = evidence.last(evidence.read(project.id), "phone", "phone_check")
-    if report is None:
-        _LOG.warning("phone check: no tutor report for %s", project.id)
-        return None
-    if not 0 <= evidence.to_ms(now) - report.ms <= DAY_MS:
-        _LOG.warning("phone check: tutor report for %s is stale or ahead", project.id)
-        return None
-    return _report_ok(report.value, report.ms)
-
-
-SOURCES: dict[str, CheckSource] = {"phone": phone_source}
-
-
-def register_source(name: str, source: CheckSource) -> None:
-    """Replace the callable behind an auto check (``phone``)."""
-    if name not in SOURCES:
-        msg = f"unknown check source {name!r}; known: {sorted(SOURCES)}"
-        raise ValueError(msg)
-    SOURCES[name] = source
-
-
-def _run(project: Project, now: datetime) -> bool | None:
-    try:
-        return SOURCES["phone"](project, now)
-    except Exception as exc:  # noqa: BLE001 (pluggable code: any failure = unknown)
-        _LOG.warning("phone check source failed: %s", exc)
-        return None
-
-
-def run_checks(
-    project_id: str, *, now: datetime | None = None
-) -> dict[str, bool | None]:
-    """Run the auto checks of open phone milestones; latch what changed.
-
-    Returns milestone id -> verdict for the ones checked (inside their
-    24 h window, not yet done). KeyError for an unknown project; a focus
-    project has nothing to check.
-    """
-    stamp = _mutate.local_now(now)
-    project = next((p for p in _store.read_all() if p.id == project_id), None)
-    if project is None:
-        raise KeyError(project_id)
-    open_ids = [
-        m.id
-        for m in _milestones.milestones(project, stamp)
-        if m.id.startswith("phone:")
-        and m.state not in {"done", "late_done"}
-        and stamp >= m.due - WINDOW
-    ]
-    if not open_ids:
-        return {}
-    verdict = _run(project, stamp)
-    rows = evidence.read(project_id)
-    for mid in open_ids:
-        _, latest, seen = evidence.check_summary(rows, mid)
-        if verdict != latest and not (verdict is None and not seen):
-            evidence.append(project_id, mid, "check", verdict, stamp)
-    return dict.fromkeys(open_ids, verdict)

focus_project/focus_project/evidence.py

@@ -1,159 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The interview evidence store: append-only rows plus copied files.
-
-One jsonl per project (``{ms, milestone, kind, value}``) under the
-:mod:`focus_project._paths` seam. There is deliberately no delete or rewrite
-function: a milestone once evidenced stays evidenced.
-"""
-
-from __future__ import annotations
-
-from dataclasses import dataclass
-from datetime import UTC, datetime
-import itertools
-import json
-import logging
-import os
-import re
-import shutil
-from typing import TYPE_CHECKING
-
-from focus_project import _clock, _paths
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-_LOG = logging.getLogger(__name__)
-
-
-@dataclass(frozen=True)
-class Evidence:
-    """One evidence row."""
-
-    ms: int
-    milestone: str
-    kind: str
-    value: object
-
-    @property
-    def at(self) -> datetime:
-        """When it was recorded, as a naive local datetime."""
-        return from_ms(self.ms)
-
-
-def to_ms(moment: datetime) -> int:
-    """Unix milliseconds of a naive local datetime."""
-    return int(moment.timestamp() * 1000)
-
-
-def from_ms(ms: int) -> datetime:
-    """Naive local datetime of unix milliseconds."""
-    return datetime.fromtimestamp(ms / 1000, tz=UTC).astimezone().replace(tzinfo=None)
-
-
-def clock_now() -> datetime:
-    """The clock seam (:mod:`focus_project._clock`) as a naive local datetime."""
-    return from_ms(int(_clock.now() * 1000))
-
-
-def append(
-    project_id: str, milestone: str, kind: str, value: object, now: datetime
-) -> Evidence:
-    """Append one row (fsynced) and return it."""
-    row = Evidence(to_ms(now), milestone, kind, value)
-    path = _paths.evidence_path(project_id)
-    path.parent.mkdir(parents=True, exist_ok=True)
-    line = json.dumps(
-        {"ms": row.ms, "milestone": milestone, "kind": kind, "value": value}
-    )
-    with path.open("a", encoding="utf-8") as handle:
-        handle.write(line + "\n")
-        handle.flush()
-        os.fsync(handle.fileno())
-    return row
-
-
-def store_file(project_id: str, source: Path, now: datetime) -> str:
-    """Copy ``source`` beside the rows under a name that never collides."""
-    target = _paths.evidence_files_dir(project_id)
-    target.mkdir(parents=True, exist_ok=True)
-    safe = re.sub(r"[^A-Za-z0-9._-]", "_", source.name)[:80]
-    stamp = to_ms(now)
-    for attempt in itertools.count():
-        name = f"{stamp}-{safe}" if not attempt else f"{stamp}-{attempt}-{safe}"
-        try:
-            with (target / name).open("xb") as out, source.open("rb") as src:
-                shutil.copyfileobj(src, out)
-                out.flush()
-                os.fsync(out.fileno())
-        except FileExistsError:
-            continue
-        return name
-    raise AssertionError  # pragma: no cover (itertools.count never ends)
-
-
-def _parse(line: str) -> Evidence | None:
-    try:
-        row = json.loads(line)
-        if not isinstance(row, dict) or isinstance(row["ms"], bool):
-            return None
-        return Evidence(
-            int(row["ms"]), str(row["milestone"]), str(row["kind"]), row["value"]
-        )
-    except ValueError, KeyError, TypeError:
-        return None
-
-
-def read(project_id: str) -> list[Evidence]:
-    """Every row in file order; a missing file is empty, bad lines are skipped.
-
-    A torn last line (crash mid-append) must not wipe out the whole history,
-    so malformed lines are logged and ignored; an unreadable file raises.
-    """
-    try:
-        path = _paths.evidence_path(project_id)
-        text = path.read_text(encoding="utf-8", errors="replace")
-    except FileNotFoundError:
-        return []
-    rows: list[Evidence] = []
-    for number, line in enumerate(text.splitlines(), start=1):
-        parsed = _parse(line)
-        if parsed is None:
-            _LOG.warning(
-                "evidence %s line %d: skipping malformed row", project_id, number
-            )
-        else:
-            rows.append(parsed)
-    return rows
-
-
-def first(
-    rows: list[Evidence], milestone: str, kind: str, since_ms: int = 0
-) -> Evidence | None:
-    """The earliest matching row at or after ``since_ms``."""
-    return next(
-        (
-            r
-            for r in rows
-            if (r.milestone, r.kind) == (milestone, kind) and r.ms >= since_ms
-        ),
-        None,
-    )
-
-
-def last(rows: list[Evidence], milestone: str, kind: str) -> Evidence | None:
-    """The latest matching row."""
-    return next(
-        (r for r in reversed(rows) if (r.milestone, r.kind) == (milestone, kind)),
-        None,
-    )
-
-
-def check_summary(
-    rows: list[Evidence], milestone: str
-) -> tuple[int | None, bool | None, bool]:
-    """``(first passing ms, latest verdict, any check row)`` for a milestone."""
-    checks = [r for r in rows if (r.milestone, r.kind) == (milestone, "check")]
-    passed = next((r.ms for r in checks if r.value is True), None)
-    latest = checks[-1].value if checks else None
-    return passed, (latest if isinstance(latest, bool) else None), bool(checks)

focus_project/focus_project/minutes.py

@@ -8,10 +8,6 @@ minutes (floor). Time not yet lived is never credited: every interval is
 clipped at "now", so an event stamped ahead of the clock (a fast phone clock)
 earns only once the real clock reaches it. An unreadable registry reads as 0
 (logged): fail closed.
-
-Unbacked minutes (0.10.0): timer time the anti-cheat gate refused
-(:mod:`focus_project.sources.activity`) and no other source covers, on the
-local day up to "now"; carry and the cap play no part. Shown, never paid.
 """
 
 from __future__ import annotations
@@ -19,10 +15,9 @@ from __future__ import annotations
 from datetime import date  # noqa: TC003
 import logging
 
-from focus_project import _api, _clock, carry, sources
+from focus_project import _api, _clock, carry
 from focus_project._errors import RegistryError
 from focus_project._own import union_seconds
-from focus_project.sources import activity
 
 _LOG = logging.getLogger(__name__)
 
@@ -43,26 +38,4 @@ def minutes_for(day: date, now: float | None = None) -> int:
     return carry.balance(day, moment).counted(project.daily_on(day))
 
 
-def unbacked_minutes_for(day: date, now: float | None = None) -> int:
+__all__ = ["minutes_for", "union_seconds"]
-    """Whole minutes of timer time on ``day`` that the gate did not count.
-
-    0 with no project, before :data:`~focus_project.sources.activity.BACKING_FROM`,
-    or with an unreadable registry (logged).
-    """
-    try:
-        project = _api.active(day)
-        if project is None:
-            return 0
-        lo, hi = sources.day_bounds(day)
-        bounds = (lo, min(hi, _clock.now() if now is None else now))
-        unbacked = sources.clip(activity.split_timer(day, project)[1], bounds)
-        counted = sources.clip(sources.collect(day, project), bounds)
-    except (RegistryError, OSError) as exc:
-        _LOG.warning("focus_project registry unreadable, 0 unbacked: %s", exc)
-        return 0
-    # |unbacked minus counted| = |unbacked union counted| - |counted|.
-    gap = union_seconds(unbacked + counted) - union_seconds(counted)
-    return int(gap // 60)
-
-
-__all__ = ["minutes_for", "unbacked_minutes_for", "union_seconds"]

focus_project/focus_project/sources/activity.py

@@ -13,36 +13,23 @@ heartbeat covering the five minutes before it. Only events recorded for
 the open segment, like a ``checkin``: within :data:`MAX_GAP_S` of the last
 beat it extends the segment; later it ends the segment at that last beat. It
 never opens a segment. Other heartbeat sources never chain the timer.
-
-0.10.0 anti-cheat (days from :data:`BACKING_FROM`): a timer span earns only
-if real activity of the same project backs it, whole span or nothing. A
-backing event is a ``tutor``/``sql_drills``/``leetcode`` heartbeat or an
-accepted LeetCode submission (leetcode projects only); ``browser`` never
-backs. A span ``(a, b)`` is backed if a backing event lies in
-``[a - 300, b]``, or between two consecutive backing events at most
-:data:`BRIDGE_S` apart. Check-ins alone earn 0. Earlier days keep the old
-rule (every span counts), so paid days never change.
 """
 
 from __future__ import annotations
 
-from datetime import date
-import importlib
-from itertools import pairwise
 import json
 import logging
-from typing import TYPE_CHECKING, Any, Final
+from typing import TYPE_CHECKING, Any
 
 from focus_project._paths import activity_path
-from focus_project.sources import HEARTBEAT_WINDOW_S, Interval, heartbeat_interval
+from focus_project.sources import Interval, heartbeat_interval
 
 if TYPE_CHECKING:
+    from datetime import date
+
     from focus_project._model import Project
 
 MAX_GAP_S = 600.0
-BACKING_FROM: Final = date(2026, 11, 8)
-BACKING_SOURCES: Final = frozenset({"tutor", "sql_drills", "leetcode"})
-BRIDGE_S: Final = 1200.0
 _LOG = logging.getLogger(__name__)
 
 
@@ -97,68 +84,12 @@ def _is_beat(event: dict[str, Any]) -> bool:
     return event.get("source") == "tutor" and event.get("kind") == "heartbeat"
 
 
-def sessions(project_id: str) -> list[Interval]:
+def intervals(day: date, project: Project) -> list[Interval]:  # noqa: ARG001
-    """Contiguous focus sessions of ``project_id``, time-ordered, as (start, end).
+    """Timer segments plus heartbeats of ``project`` (the registry clips)."""
-
-    Built from the same timer/tutor-beat segments (and so the same
-    :data:`MAX_GAP_S` rule) that credit minutes; spans that touch are one
-    session, which ends at its last beat.
-    """
-    events = read_events(project_id)
-    timer = [e for e in events if e.get("source") == "timer" or _is_beat(e)]
-    merged: list[Interval] = []
-    for lo, hi in timer_intervals(timer):
-        if merged and lo <= merged[-1][1]:
-            merged[-1] = (merged[-1][0], max(hi, merged[-1][1]))
-        else:
-            merged.append((lo, hi))
-    return merged
-
-
-def backing_times(events: list[dict[str, Any]], project: Project) -> list[float]:
-    """Sorted times of the events that may back a timer span.
-
-    The LeetCode ledger is read in its own guard: if it fails, only the
-    submissions are lost, never the heartbeats (fail closed, narrow).
-    """
-    times = [float(e["ts"]) for e in events if e.get("source") in BACKING_SOURCES]
-    try:
-        # Imported lazily: it pulls in earned_time, and the timer must keep
-        # working (heartbeat backing only) where that is not installed.
-        leetcode = importlib.import_module("focus_project.sources.leetcode")
-        times += leetcode.submission_times(project)
-    except Exception:  # noqa: BLE001 (lose the submissions, keep the beats)
-        _LOG.exception("leetcode backing unavailable, heartbeats only")
-    return sorted(times)
-
-
-def is_backed(span: Interval, times: list[float]) -> bool:
-    """Whether a backing event lies in or just before ``span``, or bridges it."""
-    lo, hi = span
-    if any(lo - HEARTBEAT_WINDOW_S <= t <= hi for t in times):
-        return True
-    return any(a <= lo and hi <= b and b - a <= BRIDGE_S for a, b in pairwise(times))
-
-
-def split_timer(
-    day: date, project: Project, events: list[dict[str, Any]] | None = None
-) -> tuple[list[Interval], list[Interval]]:
-    """``(backed, unbacked)`` timer spans of ``project``, judged for ``day``."""
-    found = read_events(project.id) if events is None else events
-    chain = [e for e in found if e.get("source") == "timer" or _is_beat(e)]
-    spans = timer_intervals(chain)
-    if day < BACKING_FROM:
-        return spans, []
-    times = backing_times(found, project)
-    backed = [s for s in spans if is_backed(s, times)]
-    return backed, [s for s in spans if s not in backed]
-
-
-def intervals(day: date, project: Project) -> list[Interval]:
-    """Backed timer segments plus heartbeats of ``project`` (the registry clips)."""
     events = read_events(project.id)
+    timer = [e for e in events if e.get("source") == "timer" or _is_beat(e)]
     beats = [e for e in events if e.get("source") != "timer"]
     return [
-        *split_timer(day, project, events)[0],
+        *timer_intervals(timer),
         *(heartbeat_interval(float(e["ts"])) for e in beats),
     ]

focus_project/focus_project/sources/leetcode.py

@@ -55,8 +55,8 @@ def _submitted_at(row: object, key: bytes) -> float | None:
         return None
 
 
-def submission_times(project: Project) -> list[float]:
+def intervals(day: date, project: Project) -> list[Interval]:  # noqa: ARG001
-    """Times of verified accepted submissions; [] for a non-leetcode project."""
+    """One heartbeat per accepted submission (the registry clips to the day)."""
     if not any("leetcode" in pattern.lower() for pattern in project.site_patterns):
         return []
     key = ledger.read_key()
@@ -64,9 +64,4 @@ def submission_times(project: Project) -> list[float]:
         _LOG.warning("leetcode source: no signing key, contributes 0")
         return []
     stamps = (_submitted_at(row, key) for row in ledger.read_rows(ledger_file()))
-    return [ts for ts in stamps if ts is not None]
+    return [heartbeat_interval(ts) for ts in stamps if ts is not None]
-
-
-def intervals(day: date, project: Project) -> list[Interval]:  # noqa: ARG001
-    """One heartbeat per accepted submission (the registry clips to the day)."""
-    return [heartbeat_interval(ts) for ts in submission_times(project)]

focus_project/focus_project/status.py

@@ -3,32 +3,10 @@
 
 from __future__ import annotations
 
-import logging
+from datetime import date  # noqa: TC003
-from typing import TYPE_CHECKING, Any
+from typing import Any
 
-from focus_project import _api, _milestones, carry, evidence, minutes
+from focus_project import _api, carry
-
-if TYPE_CHECKING:
-    from datetime import date
-
-    from focus_project._model import Project
-
-_LOG = logging.getLogger(__name__)
-
-
-def _milestone_block(project: Project) -> tuple[list[dict[str, Any]], bool]:
-    """Milestone rows and whether any is red; unreadable evidence shows none.
-
-    Display path, so it fails open (logged): the lock path, ``owed_lock``,
-    has its own handling.
-    """
-    now = evidence.clock_now()
-    try:
-        items = _milestones.milestones(project, now)
-    except OSError as exc:
-        _LOG.warning("interview evidence unreadable, no milestones shown: %s", exc)
-        return [], False
-    return [m.to_dict() for m in items], any(m.red for m in items)
 
 
 def status(day: date) -> dict[str, Any]:
@@ -36,8 +14,7 @@ def status(day: date) -> dict[str, Any]:
 
     ``minutes`` is what was really worked plus what the day took from the
     bank, so it may exceed ``daily_minutes`` (121/120); ``counted_minutes``
-    is the capped figure that pays. ``unbacked_minutes`` is timer time the
+    is the capped figure that pays.
-    anti-cheat gate refused (shown as "not counted", never paid).
     """
     project = _api.active(day)
     if project is None:
@@ -53,13 +30,9 @@ def status(day: date) -> dict[str, Any]:
             "unlock_minutes": None,
             "ends": None,
             "replaces": [],
-            "unbacked_minutes": 0,
-            "milestones": [],
-            "missed": False,
         }
     daily = project.daily_on(day)
     bank = carry.balance(day)
-    rows, missed = _milestone_block(project)
     return {
         "project": project.id,
         "name": project.name,
@@ -74,8 +47,4 @@ def status(day: date) -> dict[str, Any]:
         "unlock_minutes": project.unlock_on(day),
         "ends": project.last_day.isoformat(),
         "replaces": list(project.replaces_on(day)),
-        "unbacked_minutes": minutes.unbacked_minutes_for(day),
-        # Interview projects: [{id, due ISO, state}] and whether any is red.
-        "milestones": rows,
-        "missed": missed,
     }

focus_project/focus_project/tests/conftest.py

@@ -11,7 +11,7 @@ from typing import TYPE_CHECKING
 
 import pytest
 
-from focus_project import _clock, _paths, checks
+from focus_project import _clock
 
 if TYPE_CHECKING:
     from pathlib import Path
@@ -40,16 +40,3 @@ def _far_future_clock(monkeypatch: pytest.MonkeyPatch) -> None:
 def data(tmp_path: Path) -> Path:
     """The redirected data directory (may not exist yet)."""
     return tmp_path / "data"
-
-
-@pytest.fixture(autouse=True)
-def _interview_isolation(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Path, _redirect_state: None
-) -> None:
-    """Evidence lives under the redirected data dir; sources are a private copy.
-
-    Fails closed (R3): if the path seam ever resolves outside ``tmp_path`` the
-    test errors before it can write a production evidence row.
-    """
-    assert _paths.evidence_dir().is_relative_to(tmp_path)
-    monkeypatch.setattr(checks, "SOURCES", dict(checks.SOURCES))

focus_project/focus_project/tests/helpers.py

@@ -1,65 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Shared fixtures-as-functions for the interview-prep tests."""
-
-from __future__ import annotations
-
-from datetime import date, datetime, time, timedelta
-import json
-from typing import TYPE_CHECKING
-
-import focus_project as fp
-from focus_project._paths import activity_path
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-START = date(2026, 11, 10)  # Tuesday: D1
-TEST_DAY = date(2026, 11, 27)  # Friday: T
-TEST_AT = datetime.combine(TEST_DAY, time(10, 0))
-SAT1 = date(2026, 11, 14)
-SAT2 = date(2026, 11, 21)
-
-
-def at(day: int, hour: int = 0, minute: int = 0) -> datetime:
-    """A November 2026 local datetime."""
-    return datetime.combine(date(2026, 11, day), time(hour, minute))
-
-
-def make(pid: str = "iv", /, **kw: object) -> fp.Project:
-    """Create an interview project D1 = Nov 10, T = Nov 27."""
-    args: dict[str, object] = {
-        "name": "Interview",
-        "start": START,
-        "end": TEST_DAY,
-        "replaces": ["leetcode", "reading", "tutor"],
-        "daily_minutes": 120,
-        "unlock_minutes": 30,
-        "kind": "interview",
-        "test_at": TEST_AT,
-        "invite_url": "https://app.codility.com/c/run/abc",
-    }
-    args.update(kw)
-    return fp.create_project(pid, **args)  # type: ignore[arg-type]
-
-
-def shot(tmp_path: Path, name: str = "rules.png") -> Path:
-    """A non-empty screenshot stand-in."""
-    path = tmp_path / name
-    path.write_bytes(b"\x89PNG rules")
-    return path
-
-
-def states(project: fp.Project, now: datetime) -> dict[str, str]:
-    """Milestone id -> state."""
-    return {m.id: m.state for m in fp.milestones(project, now)}
-
-
-def session(start: datetime, minutes: int, *, step: int = 5, pid: str = "iv") -> None:
-    """A timer session start -> check-in every ``step`` min -> stop."""
-    activity_path().parent.mkdir(parents=True, exist_ok=True)
-    with activity_path().open("a", encoding="utf-8") as handle:
-        for offset in range(0, minutes + 1, step):
-            kind = "start" if offset == 0 else "checkin"
-            when = start + timedelta(minutes=offset)
-            row = {"ts": when.timestamp(), "source": "timer", "kind": kind}
-            handle.write(json.dumps({**row, "project": pid, "dev": "pc"}) + "\n")

focus_project/focus_project/tests/test_activity_status.py

@@ -137,9 +137,6 @@ def test_status_active_reports_the_worked_minutes() -> None:
         "unlock_minutes": 30,
         "ends": "2026-11-05",
         "replaces": ["leetcode"],
-        "unbacked_minutes": 0,
-        "milestones": [],
-        "missed": False,
     }
 
 
@@ -161,9 +158,6 @@ def test_status_inactive() -> None:
         "unlock_minutes",
         "ends",
         "replaces",
-        "unbacked_minutes",
-        "milestones",
-        "missed",
     }
 
 

focus_project/focus_project/tests/test_backing.py

@@ -1,181 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""0.10.0: a timer span earns only if a same-project beat backs it."""
-
-from __future__ import annotations
-
-from datetime import date
-from typing import TYPE_CHECKING
-
-import focus_project as fp
-from focus_project import _api, minutes
-from focus_project.sources import activity
-from focus_project.tests import test_minutes as tm
-
-if TYPE_CHECKING:
-    import pytest
-
-DAY = date(2026, 11, 8)  # BACKING_FROM
-
-
-def at(hour: int, minute: int = 0) -> float:
-    return tm.at(hour, minute, day=DAY)
-
-
-def proj() -> None:
-    fp.create_project(
-        "hr",
-        name="HR prep",
-        start=DAY,
-        end=date(2026, 11, 12),
-        replaces=["leetcode"],
-        daily_minutes=120,
-        unlock_minutes=30,
-        site_patterns=["leetcode.com"],
-    )
-
-
-def rows(*items: tuple[str, str, float]) -> None:
-    tm.events(*items)
-
-
-def timer_only() -> None:
-    rows(
-        ("timer", "start", at(9)),
-        ("timer", "checkin", at(9, 5)),
-        ("timer", "stop", at(9, 10)),
-    )
-
-
-def test_timer_only_stretch_pays_zero_and_shows_unbacked() -> None:
-    proj()
-    timer_only()
-    assert minutes.minutes_for(DAY) == 0
-    assert minutes.unbacked_minutes_for(DAY) == 10
-    assert fp.status(DAY)["unbacked_minutes"] == 10
-
-
-def test_a_beat_in_the_span_backs_all_of_it() -> None:
-    proj()
-    timer_only()
-    rows(("sql_drills", "heartbeat", at(9, 5)))
-    assert minutes.unbacked_minutes_for(DAY) == 0
-    assert minutes.minutes_for(DAY) == 10
-
-
-def test_a_beat_backs_only_the_spans_it_touches() -> None:
-    proj()
-    timer_only()
-    rows(("sql_drills", "heartbeat", at(9, 7)))
-    # 9:00-9:05 stays unbacked, but the beat itself pays 9:02-9:07.
-    assert minutes.unbacked_minutes_for(DAY) == 2
-
-
-def test_a_beat_at_the_end_of_the_span_backs_it() -> None:
-    proj()
-    rows(
-        ("timer", "start", at(9)),
-        ("timer", "stop", at(9, 10)),
-        ("sql_drills", "heartbeat", at(9, 10)),
-    )
-    assert minutes.unbacked_minutes_for(DAY) == 0
-    assert minutes.minutes_for(DAY) == 10
-
-
-def test_a_beat_just_before_the_span_backs_it() -> None:
-    proj()
-    rows(
-        ("sql_drills", "heartbeat", at(8, 56)),  # within 300 s before 9:00
-        ("timer", "start", at(9)),
-        ("timer", "stop", at(9, 5)),
-    )
-    assert minutes.unbacked_minutes_for(DAY) == 0
-
-
-def test_a_late_beat_does_not_back_it() -> None:
-    proj()
-    timer_only()
-    rows(("sql_drills", "heartbeat", at(9, 40)))
-    assert minutes.unbacked_minutes_for(DAY) == 10
-
-
-def test_beats_up_to_twenty_minutes_apart_bridge_the_span() -> None:
-    proj()
-    rows(
-        ("tutor", "heartbeat", at(8, 50)),
-        ("timer", "start", at(9)),
-        ("timer", "checkin", at(9, 5)),
-        ("timer", "stop", at(9, 8)),
-        ("tutor", "heartbeat", at(9, 10)),
-    )
-    assert minutes.unbacked_minutes_for(DAY) == 0
-
-
-def test_beats_over_twenty_minutes_apart_do_not_bridge() -> None:
-    proj()
-    rows(
-        ("tutor", "heartbeat", at(8, 30)),
-        ("timer", "start", at(9)),
-        ("timer", "stop", at(9, 8)),
-        ("tutor", "heartbeat", at(9, 40)),
-    )
-    assert minutes.unbacked_minutes_for(DAY) == 8
-
-
-def test_browser_never_backs() -> None:
-    proj()
-    timer_only()
-    rows(("browser", "heartbeat", at(9, 7)))
-    # The beat pays its own 9:02-9:07 but backs no timer span.
-    assert minutes.unbacked_minutes_for(DAY) == 5
-    assert minutes.minutes_for(DAY) == 5
-
-
-def test_days_before_backing_from_keep_the_old_rule() -> None:
-    fp.create_project(
-        "hr",
-        name="Old",
-        start=date(2026, 11, 2),
-        end=date(2026, 11, 5),
-        replaces=["leetcode"],
-        daily_minutes=120,
-        unlock_minutes=30,
-        site_patterns=["leetcode.com"],
-    )
-    old = date(2026, 11, 2)
-    rows(("timer", "start", tm.at(9)), ("timer", "stop", tm.at(9, 10)))
-    assert minutes.minutes_for(old) == 10
-    assert minutes.unbacked_minutes_for(old) == 0
-
-
-def test_no_project_is_zero_unbacked() -> None:
-    assert minutes.unbacked_minutes_for(DAY) == 0
-    assert fp.status(DAY)["unbacked_minutes"] == 0
-
-
-def test_unbacked_is_clipped_at_now() -> None:
-    proj()
-    timer_only()
-    assert minutes.unbacked_minutes_for(DAY, now=at(9, 4)) == 4
-
-
-def test_split_timer_returns_backed_and_unbacked() -> None:
-    proj()
-    timer_only()
-    project = _api.active(DAY)
-    assert project is not None
-    backed, unbacked = activity.split_timer(DAY, project)
-    assert backed == []
-    assert unbacked == [(at(9), at(9, 5)), (at(9, 5), at(9, 10))]
-
-
-def test_backing_from_is_exported() -> None:
-    assert date(2026, 11, 8) == fp.BACKING_FROM
-    assert fp.unbacked_minutes_for is minutes.unbacked_minutes_for
-
-
-def test_registry_error_reads_zero(monkeypatch: pytest.MonkeyPatch) -> None:
-    def boom(_day: date) -> None:
-        raise OSError
-
-    monkeypatch.setattr("focus_project._api.active", boom)
-    assert minutes.unbacked_minutes_for(DAY) == 0

focus_project/focus_project/tests/test_backing_edges.py

@@ -1,131 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""More 0.10.0 backing cases: sources, projects, the cutover midnight, chains.
-
-The core rule lives in ``test_backing.py``; these pin the edges the spec
-named (C5-C7) with leetcode-ledger and other-project fixtures.
-"""
-
-from __future__ import annotations
-
-from datetime import date
-import json
-from typing import TYPE_CHECKING
-
-import pytest
-
-import focus_project as fp
-from focus_project import minutes
-from focus_project._paths import activity_path
-from focus_project.tests.test_leetcode_source import ac_row, write
-from focus_project.tests.test_minutes import at
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-DAY = date(2026, 11, 9)  # after the cutover
-PRE = date(2026, 11, 7)  # before it
-
-
-@pytest.fixture(autouse=True)
-def _home(monkeypatch: pytest.MonkeyPatch, tmp_path: Path) -> None:
-    """Nothing may fall back to the real home directory."""
-    monkeypatch.setenv("HOME", str(tmp_path / "home"))
-
-
-def make(pid: str = "hr", patterns: tuple[str, ...] = ("leetcode.com",)) -> None:
-    fp.create_project(
-        pid,
-        name=pid,
-        start=date(2026, 11, 6),
-        end=date(2026, 11, 12),
-        replaces=["leetcode"],
-        daily_minutes=240,
-        unlock_minutes=30,
-        site_patterns=list(patterns),
-    )
-
-
-def rows(*items: tuple[str, str, float] | tuple[str, str, float, str]) -> None:
-    """Append ``(source, kind, ts[, project])`` events (project default hr)."""
-    activity_path().parent.mkdir(parents=True, exist_ok=True)
-    with activity_path().open("a", encoding="utf-8") as handle:
-        for item in items:
-            source, kind, ts = item[:3]
-            pid = item[3] if len(item) == 4 else "hr"
-            row = {"ts": ts, "source": source, "kind": kind, "project": pid}
-            handle.write(json.dumps(row) + "\n")
-
-
-def timer(day: date, *marks: tuple[int, int]) -> None:
-    """A segment: start at the first mark, check-ins, stop at the last."""
-    kinds = ["start", *["checkin"] * (len(marks) - 2), "stop"]
-    rows(
-        *(
-            ("timer", k, at(h, m, day=day))
-            for k, (h, m) in zip(kinds, marks, strict=True)
-        )
-    )
-
-
-def beat(source: str, h: int, m: int, day: date = DAY, pid: str = "hr") -> None:
-    rows((source, "heartbeat", at(h, m, day=day), pid))
-
-
-@pytest.mark.parametrize("source", ["sql_drills", "leetcode"])
-def test_sql_drills_and_leetcode_heartbeats_back_too(source: str) -> None:
-    make()
-    timer(DAY, (9, 0), (9, 5), (9, 10))
-    beat(source, 9, 5)
-    assert minutes.minutes_for(DAY) == 10
-
-
-def test_browser_and_other_project_beats_do_not_back() -> None:
-    make()  # "other" needs no registry entry: rows are filtered by id
-    timer(DAY, (9, 0), (9, 5), (9, 10))
-    beat("browser", 9, 4)
-    beat("tutor", 9, 4, pid="other")
-    assert minutes.minutes_for(DAY) == 5  # the browser beat's own window only
-    assert minutes.unbacked_minutes_for(DAY) == 6  # 9:04-9:10
-
-
-def test_accepted_submission_backs_a_leetcode_project(tmp_path: Path) -> None:
-    make()
-    timer(DAY, (9, 0), (9, 5), (9, 10))
-    write(tmp_path, ac_row(at(9, 4, day=DAY)))
-    assert minutes.minutes_for(DAY) == 11  # 8:59-9:10
-
-
-def test_submission_does_not_back_a_non_leetcode_project(tmp_path: Path) -> None:
-    make("hr", ("sqlzoo.net",))
-    timer(DAY, (9, 0), (9, 5), (9, 10))
-    write(tmp_path, ac_row(at(9, 4, day=DAY)))
-    assert minutes.minutes_for(DAY) == 0
-
-
-def test_span_across_the_cutover_midnight_follows_the_day() -> None:
-    make()
-    rows(
-        ("timer", "start", at(23, 58, day=date(2026, 11, 7))),
-        ("timer", "checkin", at(0, 3, day=date(2026, 11, 8))),
-        ("timer", "stop", at(0, 5, day=date(2026, 11, 8))),
-    )
-    assert minutes.minutes_for(date(2026, 11, 7)) == 2
-    assert minutes.minutes_for(date(2026, 11, 8)) == 0
-    assert minutes.unbacked_minutes_for(date(2026, 11, 8)) == 5
-
-
-def test_tutor_chained_span_is_backed_but_earlier_reading_is_not() -> None:
-    """A tutor beat chains the segment (0.8.0) and backs the span it ends.
-
-    start 9:00, checkin 9:08, tutor beat 9:16: the span 9:08-9:16 ends at a
-    backing beat and earns; 9:00-9:08 has no beat in or 5 min before it and
-    nothing bridges it, so it shows as unbacked.
-    """
-    make()
-    rows(
-        ("timer", "start", at(9, 0, day=DAY)),
-        ("timer", "checkin", at(9, 8, day=DAY)),
-    )
-    beat("tutor", 9, 16)
-    assert minutes.minutes_for(DAY) == 8
-    assert minutes.unbacked_minutes_for(DAY) == 8

focus_project/focus_project/tests/test_checks.py

@@ -1,193 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Pluggable check sources, the phone source, and latching."""
-
-from __future__ import annotations
-
-import json
-from typing import TYPE_CHECKING
-
-import pytest
-
-import focus_project as fp
-from focus_project import _paths, checks, evidence
-from focus_project.tests.helpers import SAT1, at, make
-
-if TYPE_CHECKING:
-    from collections.abc import Callable
-    from datetime import datetime
-
-MID = "phone:2026-11-14"
-
-
-def _ms(moment: datetime) -> int:
-    return int(moment.timestamp() * 1000)
-
-
-def _report(
-    moment: datetime,
-    *,
-    plan: bool = True,
-    due: int = 5,
-    seen: datetime | None = None,
-) -> None:
-    value = {"plan": plan, "due": due, "seen_ms": _ms(seen or moment)}
-    fp.record_evidence("iv", "phone", "phone_check", value, now=moment)
-
-
-def _phone(now: datetime) -> bool | None:
-    return checks.phone_source(fp.list_projects()[0], now)
-
-
-def test_phone_source_without_a_report_is_unknown(
-    caplog: pytest.LogCaptureFixture,
-) -> None:
-    make()
-    with caplog.at_level("WARNING"):
-        assert _phone(at(13, 9)) is None
-    assert "phone" in caplog.text
-
-
-def test_phone_source_passes_on_a_fresh_good_report() -> None:
-    make()
-    _report(at(13, 8), seen=at(12, 20))
-    assert _phone(at(13, 9)) is True
-
-
-@pytest.mark.parametrize(
-    "bad",
-    [{"plan": False}, {"due": 0}, {"seen": at(11, 7)}],
-)
-def test_phone_source_fails_on_any_bad_part(bad: dict[str, object]) -> None:
-    make()
-    _report(at(13, 8), **bad)  # type: ignore[arg-type]
-    assert _phone(at(13, 9)) is False
-
-
-def test_phone_source_stale_or_future_report_is_unknown() -> None:
-    make()
-    _report(at(13, 8))
-    assert _phone(at(14, 8, 1)) is None  # the worker stopped reporting
-    assert _phone(at(13, 7)) is None  # stamped ahead of the clock
-
-
-def test_phone_source_uses_the_latest_report() -> None:
-    make()
-    _report(at(13, 8), plan=False)
-    _report(at(13, 9))
-    assert _phone(at(13, 10)) is True
-
-
-def test_phone_source_hand_edited_row_is_a_failure() -> None:
-    make()
-    _report(at(13, 8))
-    path = _paths.evidence_path("iv")
-    row = {"ms": _ms(at(13, 9)), "milestone": "phone", "kind": "phone_check"}
-    path.write_text(path.read_text() + json.dumps({**row, "value": "ok"}) + "\n")
-    assert _phone(at(13, 10)) is False
-
-
-def _patch(
-    monkeypatch: pytest.MonkeyPatch, fn: Callable[[fp.Project, datetime], bool | None]
-) -> None:
-    monkeypatch.setitem(checks.SOURCES, "phone", fn)
-
-
-def _trip_project() -> None:
-    make()
-    fp.answer_weekend("iv", SAT1, trip=True, now=at(11, 9))
-
-
-def test_run_checks_latches_a_pass(monkeypatch: pytest.MonkeyPatch) -> None:
-    _trip_project()
-    _patch(monkeypatch, lambda p, n: True)
-    assert fp.run_checks("iv", now=at(13, 9)) == {MID: True}
-    assert fp.run_checks("iv", now=at(13, 10)) == {MID: True}
-    rows = [r for r in evidence.read("iv") if r.kind == "check"]
-    assert [(r.milestone, r.value) for r in rows] == [(MID, True)]
-
-
-def test_run_checks_only_inside_the_window(monkeypatch: pytest.MonkeyPatch) -> None:
-    _trip_project()
-    _patch(monkeypatch, lambda p, n: True)
-    assert fp.run_checks("iv", now=at(12, 11, 59)) == {}  # >24 h before due
-    assert fp.run_checks("iv", now=at(12, 12)) == {MID: True}
-
-
-def test_run_checks_records_changes_not_every_tick(
-    monkeypatch: pytest.MonkeyPatch,
-) -> None:
-    _trip_project()
-    answers: list[bool | None] = [None, False, False, None, None]
-    _patch(monkeypatch, lambda p, n: answers.pop(0))
-    for minute in range(5):
-        fp.run_checks("iv", now=at(13, 9, minute))
-    rows = [r.value for r in evidence.read("iv") if r.kind == "check"]
-    assert rows == [False, None]  # leading None writes nothing
-
-
-def test_run_checks_source_crash_is_unknown_and_loud(
-    monkeypatch: pytest.MonkeyPatch, caplog: pytest.LogCaptureFixture
-) -> None:
-    _trip_project()
-
-    def boom(p: fp.Project, n: datetime) -> bool:
-        msg = "tutor down"
-        raise OSError(msg)
-
-    _patch(monkeypatch, boom)
-    with caplog.at_level("WARNING"):
-        assert fp.run_checks("iv", now=at(13, 9)) == {MID: None}
-    assert "tutor down" in caplog.text
-
-
-def test_run_checks_skips_finished_milestones(monkeypatch: pytest.MonkeyPatch) -> None:
-    _trip_project()
-    fp.record_evidence("iv", MID, "tap", value=True, now=at(12, 9))
-    evidence.append("iv", MID, "check", True, at(12, 10))
-    _patch(monkeypatch, lambda p, n: pytest.fail("already done"))
-    assert fp.run_checks("iv", now=at(13, 9)) == {}
-
-
-def test_run_checks_ignores_focus_projects() -> None:
-    fp.create_project(
-        "f",
-        name="F",
-        start=at(10).date(),
-        end=at(12).date(),
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    assert fp.run_checks("f", now=at(11, 9)) == {}
-    with pytest.raises(KeyError):
-        fp.run_checks("ghost", now=at(11, 9))
-
-
-def test_register_source_swaps_the_callable(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setattr(checks, "SOURCES", dict(checks.SOURCES))
-
-    def fn(p: fp.Project, n: datetime) -> bool:
-        return True
-
-    fp.register_source("phone", fn)
-    assert checks.SOURCES["phone"] is fn
-    with pytest.raises(ValueError, match="source"):
-        fp.register_source("calendar", fn)
-
-
-def test_end_to_end_phone_milestone_goes_done() -> None:
-    _trip_project()
-    fp.record_evidence("iv", MID, "tap", value=True, now=at(12, 9))
-    _report(at(13, 9))
-    assert fp.run_checks("iv", now=at(13, 9, 30)) == {MID: True}
-    state = {m.id: m.state for m in fp.milestones(fp.list_projects()[0], at(13, 10))}
-    assert state[MID] == "done"
-
-
-def test_report_missing_a_field_fails() -> None:
-    make()
-    path = _paths.evidence_path("iv")
-    path.parent.mkdir(parents=True, exist_ok=True)
-    row = {"ms": _ms(at(13, 8)), "milestone": "phone", "kind": "phone_check"}
-    path.write_text(json.dumps({**row, "value": {"plan": True}}) + "\n")
-    assert _phone(at(13, 9)) is False

focus_project/focus_project/tests/test_evidence.py

@@ -1,126 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The append-only evidence store and record_evidence validation."""
-
-from __future__ import annotations
-
-import json
-from typing import TYPE_CHECKING
-
-import pytest
-
-import focus_project as fp
-from focus_project import _paths, evidence
-from focus_project.tests.helpers import at, make, shot
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-
-def test_paths_are_isolated(tmp_path: Path) -> None:
-    assert _paths.evidence_dir().is_relative_to(tmp_path)
-    assert _paths.evidence_path("iv").name == "iv.jsonl"
-    assert _paths.evidence_files_dir("iv").name == "iv"
-
-
-def test_file_is_copied_and_row_appended(tmp_path: Path) -> None:
-    make()
-    src = shot(tmp_path)
-    row = fp.record_evidence("iv", "rules", "file", file=src, now=at(10, 9))
-    assert row.kind == "file"
-    stored = _paths.evidence_files_dir("iv") / str(row.value)
-    assert stored.read_bytes() == src.read_bytes()
-    line = json.loads(_paths.evidence_path("iv").read_text().splitlines()[0])
-    assert set(line) == {"ms", "milestone", "kind", "value"}
-    assert line["milestone"] == "rules"
-
-
-def test_rows_append_never_rewrite() -> None:
-    make()
-    fp.record_evidence("iv", "mock", "score", 71, now=at(24, 9))
-    fp.record_evidence("iv", "mock", "score", 80.5, now=at(24, 10))
-    assert [r.value for r in evidence.read("iv")] == [71, 80.5]
-    assert not hasattr(fp, "delete_evidence")
-    assert not hasattr(evidence, "delete")
-
-
-def test_same_name_files_do_not_overwrite(tmp_path: Path) -> None:
-    make()
-    one = fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(10, 9))
-    two = fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(10, 9))
-    assert one.value != two.value
-
-
-@pytest.mark.parametrize(
-    ("milestone", "kind", "value"),
-    [
-        ("rules", "tap", True),  # wrong kind for the milestone
-        ("rules", "check", True),  # internal kind
-        ("trip_ask:2026-11-14", "answer", True),  # use answer_weekend
-        ("nope", "tap", True),  # not a milestone of the project
-        ("mock", "score", "high"),
-        ("mock", "score", True),
-        ("mock", "score", float("nan")),
-        ("mock", "score", -1),
-        ("slot", "slot", "tomorrow"),
-        ("slot", "slot", "2026-11-26T10:00:00+02:00"),  # aware
-        ("slot", "slot_confirm", False),
-        ("phone", "phone_check", {"plan": True}),
-        ("phone", "phone_check", {"plan": "yes", "due": 1, "seen_ms": 1}),
-        ("phone", "phone_check", "ok"),
-    ],
-)
-def test_bad_evidence_refused(milestone: str, kind: str, value: object) -> None:
-    make()
-    with pytest.raises(ValueError):  # noqa: PT011
-        fp.record_evidence("iv", milestone, kind, value, now=at(12, 9))
-
-
-def test_file_kind_needs_a_real_nonempty_file(tmp_path: Path) -> None:
-    make()
-    empty = tmp_path / "e.png"
-    empty.write_bytes(b"")
-    for bad in (None, tmp_path / "missing.png", empty, tmp_path):
-        with pytest.raises(ValueError, match="file"):
-            fp.record_evidence("iv", "rules", "file", file=bad, now=at(10, 9))
-
-
-def test_unknown_project_is_keyerror() -> None:
-    with pytest.raises(KeyError):
-        fp.record_evidence("ghost", "mock", "score", 1, now=at(24, 9))
-
-
-def test_focus_project_has_no_milestones() -> None:
-    fp.create_project(
-        "f",
-        name="F",
-        start=at(10).date(),
-        end=at(12).date(),
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    with pytest.raises(ValueError, match="milestone"):
-        fp.record_evidence("f", "mock", "score", 1, now=at(10, 9))
-
-
-def test_torn_and_foreign_lines_are_skipped(caplog: pytest.LogCaptureFixture) -> None:
-    make()
-    fp.record_evidence("iv", "mock", "score", 70, now=at(24, 9))
-    path = _paths.evidence_path("iv")
-    path.write_text(path.read_text() + '{"ms": 1, "milest\n[1]\n{"ms":"x"}\n')
-    with caplog.at_level("WARNING"):
-        rows = evidence.read("iv")
-    assert [r.value for r in rows] == [70]
-    assert "skipping" in caplog.text
-
-
-def test_missing_file_reads_empty() -> None:
-    assert evidence.read("nobody") == []
-
-
-def test_garbage_bytes_are_skipped_not_fatal() -> None:
-    make()
-    fp.record_evidence("iv", "mock", "score", 70, now=at(24, 9))
-    path = _paths.evidence_path("iv")
-    path.write_bytes(path.read_bytes() + b"\xff\xfe\x80 not utf-8\n")
-    assert [r.value for r in evidence.read("iv")] == [70]

focus_project/focus_project/tests/test_interview_model.py

@@ -1,175 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Project.kind, interview fields, validation, budget."""
-
-from __future__ import annotations
-
-from datetime import date
-import json
-from typing import TYPE_CHECKING
-
-import pytest
-
-import focus_project as fp
-from focus_project import _interview, _milestones
-from focus_project._errors import RegistryError
-from focus_project.tests.helpers import SAT1, START, TEST_AT, TEST_DAY, at, make
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-LIVE_ROW = {
-    "id": "hackerrank-prep",
-    "name": "HackerRank prep",
-    "start": "2026-11-02",
-    "end": "2026-11-05",
-    "replaces": ["leetcode", "reading", "tutor"],
-    "daily_minutes": 120,
-    "unlock_minutes": 30,
-    "subjects": ["a"],
-    "site_patterns": ["leetcode.com"],
-    "ended_on": None,
-    "created_ms": 1793603609345,
-    "work_url": "https://www.hackerrank.com/dashboard",
-    "history": [],
-}
-
-
-def test_live_projects_json_shape_loads(data: Path) -> None:
-    data.mkdir()
-    (data / "projects.json").write_text(json.dumps({"projects": [LIVE_ROW]}))
-    [p] = fp.list_projects()
-    assert p.kind == "focus"
-    assert p.test_at is None
-    assert p.trips == ()
-    assert p.weekend_answers == ()
-    assert p.daily_on(date(2026, 11, 3)) == 120
-    snap = fp.status(date(2026, 11, 3))
-    assert snap["milestones"] == []
-    assert snap["missed"] is False
-
-
-def test_focus_row_stays_free_of_interview_keys(data: Path) -> None:
-    fp.create_project(
-        "f",
-        name="F",
-        start=START,
-        end=TEST_DAY,
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    row = json.loads((data / "projects.json").read_text())["projects"][0]
-    assert row["kind"] == "focus"
-    assert not {"test_at", "invite_url", "trips", "weekend_answers"} & set(row)
-
-
-def test_interview_roundtrip_and_file_shape(data: Path) -> None:
-    p = make(
-        trips=[fp.Trip(date(2026, 11, 20), date(2026, 11, 22))],
-        weekend_answers={SAT1: False},
-    )
-    [back] = fp.list_projects()
-    assert back == p
-    assert back.kind == "interview"
-    assert back.test_at == TEST_AT
-    assert back.trips[0].entered_ms > 0
-    row = json.loads((data / "projects.json").read_text())["projects"][0]
-    assert row["test_at"] == "2026-11-27T10:00:00"
-    assert row["trips"][0]["start"] == "2026-11-20"
-    assert row["weekend_answers"] == {"2026-11-14": False}
-    assert row["invite_url"].startswith("https://")
-
-
-@pytest.mark.parametrize(
-    "bad",
-    [
-        {"test_at": None},
-        {"test_at": at(26, 10)},  # end != test day
-        {"test_at": at(9, 10), "end": date(2026, 11, 9)},
-        {"invite_url": "ftp://x.example/t"},
-        {"invite_url": "javascript:alert(1)"},
-        {"invite_url": "https://a b.example"},
-        {"kind": "bogus"},
-        {"trips": [fp.Trip(date(2026, 11, 28), date(2026, 11, 29))]},
-        {"trips": [fp.Trip(date(2026, 11, 20), date(2026, 11, 19))]},
-        {"weekend_answers": {date(2026, 11, 15): True}},  # a Sunday
-        {"weekend_answers": {date(2026, 11, 28): True}},  # after T
-        {"daily_minutes": 29},
-    ],
-)
-def test_interview_validation(bad: dict[str, object]) -> None:
-    with pytest.raises(ValueError):  # noqa: PT011
-        make(**bad)
-
-
-def test_focus_kind_rejects_interview_fields() -> None:
-    with pytest.raises(ValueError, match="interview"):
-        make(kind="focus")
-
-
-def test_empty_invite_url_is_allowed() -> None:
-    assert make(invite_url="").invite_url == ""
-
-
-def test_budget_30_weekdays_120_weekend() -> None:
-    p = make(unlock_minutes=60)
-    assert p.daily_on(date(2026, 11, 10)) == 30  # Tue
-    assert p.daily_on(date(2026, 11, 13)) == 30  # Fri
-    assert p.daily_on(SAT1) == 120
-    assert p.daily_on(date(2026, 11, 15)) == 120  # Sun
-    assert p.unlock_on(date(2026, 11, 10)) == 30  # clamped to the day's target
-    assert p.unlock_on(SAT1) == 60
-
-
-def test_budget_flows_through_status() -> None:
-    make()
-    assert fp.status(date(2026, 11, 10))["daily_minutes"] == 30
-    assert fp.status(SAT1)["daily_minutes"] == 120
-
-
-def test_interview_exempt_from_floor_but_focus_is_not() -> None:
-    make(daily_minutes=30, unlock_minutes=30)  # 3 gates, floor 90
-    with pytest.raises(ValueError, match="at least"):
-        fp.create_project(
-            "f",
-            name="F",
-            start=date(2026, 12, 1),
-            end=date(2026, 12, 2),
-            replaces=["leetcode", "reading"],
-            daily_minutes=30,
-            unlock_minutes=10,
-        )
-
-
-def test_non_interview_has_no_weekends() -> None:
-    focus = fp.create_project(
-        "f",
-        name="F",
-        start=START,
-        end=TEST_DAY,
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    assert _interview.saturdays(focus) == []
-
-
-def test_trip_without_entry_time_uses_the_nominal_due() -> None:
-    assert _milestones._trip_due(date(2026, 11, 13), 0) == at(12, 12)
-
-
-def test_non_boolean_weekend_answer_is_a_bad_row(data: Path) -> None:
-    make()
-    path = data / "projects.json"
-    rows = json.loads(path.read_text())
-    rows["projects"][0]["weekend_answers"] = {"2026-11-14": "false"}
-    path.write_text(json.dumps(rows))
-    with pytest.raises(RegistryError):
-        fp.list_projects()
-
-
-def test_interview_unlock_edit_keeps_the_ratchet() -> None:
-    make(unlock_minutes=60)
-    fp.update_project("iv", today=SAT1, unlock_minutes=10)
-    assert fp.list_projects()[0].unlock_on(SAT1) == 60  # lower holds from tomorrow
-    assert fp.list_projects()[0].unlock_on(date(2026, 11, 15)) == 10

focus_project/focus_project/tests/test_leetcode_source.py

@@ -14,8 +14,6 @@ import pytest
 import focus_project as fp
 from focus_project import minutes, sources
 from focus_project.sources import leetcode
-from focus_project.tests.test_backing import DAY, proj, rows
-from focus_project.tests.test_backing import at as day_at
 from focus_project.tests.test_minutes import D1, at, project
 
 KEY = b"k" * 32
@@ -129,29 +127,3 @@ def test_read_only_replay_of_real_ledger(monkeypatch: pytest.MonkeyPatch) -> Non
     assert got
     assert all(b - a == sources.HEARTBEAT_WINDOW_S for a, b in got)
     assert REAL.stat().st_mtime_ns == before
-
-
-def test_accepted_submission_backs_a_timer_span(tmp_path: Path) -> None:
-    """0.10.0: an accepted submission backs timer time on a backed day."""
-    proj()
-    rows(("timer", "start", day_at(9)), ("timer", "stop", day_at(9, 10)))
-    assert minutes.unbacked_minutes_for(DAY) == 10
-    write(tmp_path, ac_row(day_at(9, 10)))
-    assert minutes.unbacked_minutes_for(DAY) == 0
-    assert minutes.minutes_for(DAY) == 10
-
-
-def test_unreadable_ledger_leaves_heartbeat_backing(
-    monkeypatch: pytest.MonkeyPatch,
-) -> None:
-    def boom(_project: object) -> list[float]:
-        raise OSError
-
-    monkeypatch.setattr(leetcode, "submission_times", boom)
-    proj()
-    rows(
-        ("timer", "start", day_at(9)),
-        ("timer", "stop", day_at(9, 10)),
-        ("tutor", "heartbeat", day_at(9, 10)),
-    )
-    assert minutes.unbacked_minutes_for(DAY) == 0

focus_project/focus_project/tests/test_milestones.py

@@ -1,228 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""Milestone generation, states and the lock that falls out of them."""
-
-from __future__ import annotations
-
-from datetime import date
-from typing import TYPE_CHECKING
-
-import focus_project as fp
-from focus_project import evidence
-from focus_project.tests.helpers import SAT1, SAT2, at, make, session, shot, states
-
-if TYPE_CHECKING:
-    from pathlib import Path
-
-    import pytest
-
-
-def _due(p: fp.Project, now_day: int = 10) -> dict[str, str]:
-    return {m.id: m.due.isoformat() for m in fp.milestones(p, at(now_day, 8))}
-
-
-def test_table_of_dues() -> None:
-    assert _due(make()) == {
-        "rules": "2026-11-10T23:59:00",
-        "trip_ask:2026-11-14": "2026-11-12T23:59:00",
-        "trip_ask:2026-11-21": "2026-11-19T23:59:00",
-        "mock": "2026-11-25T23:59:00",
-        "slot": "2026-11-26T12:00:00",
-    }
-
-
-def test_sorted_by_due_and_all_pending_at_start() -> None:
-    project = make()
-    items = fp.milestones(project, at(10, 8))
-    assert [m.due for m in items] == sorted(m.due for m in items)
-    assert {m.state for m in items} == {"pending"}
-    assert items[0].to_dict() == {
-        "id": "rules",
-        "due": "2026-11-10T23:59:00",
-        "state": "pending",
-    }
-
-
-def test_focus_project_has_none() -> None:
-    p = fp.create_project(
-        "f",
-        name="F",
-        start=date(2026, 11, 10),
-        end=date(2026, 11, 12),
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    assert fp.milestones(p, at(11, 8)) == []
-
-
-def test_unanswered_weekend_is_a_trip_once_ask_is_due() -> None:
-    p = make()
-    assert "phone:2026-11-14" not in states(p, at(12, 23, 58))
-    got = states(p, at(13, 8))
-    assert got["trip_ask:2026-11-14"] == "missed"
-    assert got["phone:2026-11-14"] == "pending"  # due Fri 12:00
-    assert states(p, at(13, 12))["phone:2026-11-14"] == "missed"
-    assert _due(p, 13)["phone:2026-11-14"] == "2026-11-13T12:00:00"
-
-
-def test_answers_decide_the_phone_milestone() -> None:
-    make()
-    no = fp.answer_weekend("iv", SAT1, trip=False, now=at(11, 9))
-    assert "phone:2026-11-14" not in states(no, at(13, 13))
-    assert states(no, at(13, 13))["trip_ask:2026-11-14"] == "done"
-    yes = fp.answer_weekend("iv", SAT1, trip=True, now=at(11, 10))
-    assert states(yes, at(11, 11))["phone:2026-11-14"] == "pending"
-
-
-def test_explicit_trip_covers_its_saturday() -> None:
-    make()
-    p = fp.add_trip("iv", date(2026, 11, 13), date(2026, 11, 15), now=at(10, 9))
-    got = states(p, at(13, 8))
-    assert "phone:2026-11-13" in got
-    assert "phone:2026-11-14" not in got
-    assert _due(p)["phone:2026-11-13"] == "2026-11-12T12:00:00"
-
-
-def test_late_trip_is_due_the_moment_it_is_entered() -> None:
-    make()
-    p = fp.add_trip("iv", date(2026, 11, 13), date(2026, 11, 14), now=at(12, 8))
-    assert _due(p, 12)["phone:2026-11-13"] == "2026-11-12T08:00:00"
-    assert states(p, at(12, 8))["phone:2026-11-13"] == "missed"
-    far = fp.add_trip("iv", date(2026, 11, 20), date(2026, 11, 22), now=at(12, 9))
-    assert _due(far, 12)["phone:2026-11-20"] == "2026-11-19T12:00:00"
-
-
-def _clean(tmp_path: Path) -> None:
-    """Everything due before the slot is done on time; weekend 2 is no trip."""
-    fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(10, 9))
-    fp.answer_weekend("iv", SAT1, trip=False, now=at(11, 9))
-    fp.answer_weekend("iv", SAT2, trip=False, now=at(11, 9))
-    session(at(24, 7), 90)
-    fp.record_evidence("iv", "mock", "score", 70, now=at(24, 9))
-
-
-def test_slot_is_due_t_minus_1_at_noon_and_locks_that_evening(tmp_path: Path) -> None:
-    make()
-    _clean(tmp_path)
-    assert fp.owed_lock(at(26, 11, 59)) is None
-    owed = fp.owed_lock(at(26, 12))
-    assert owed is not None
-    assert owed.id == "slot"
-    assert owed.state == "missed"
-    assert fp.owed_lock(at(26, 20)) is not None
-
-
-def test_test_day_never_locks_but_shows_red(tmp_path: Path) -> None:
-    make()
-    assert fp.owed_lock(at(27, 9)) is None
-    assert fp.status(date(2026, 11, 27))["milestones"] is not None
-    assert {m.state for m in fp.milestones(fp.list_projects()[0], at(27, 9))} >= {
-        "missed"
-    }
-
-
-def test_owed_lock_is_the_earliest_missed() -> None:
-    make()
-    owed = fp.owed_lock(at(13, 8))
-    assert owed is not None
-    assert owed.id == "rules"
-
-
-def test_late_completion_unlocks(tmp_path: Path) -> None:
-    make()
-    fp.answer_weekend("iv", SAT1, trip=False, now=at(11, 9))
-    owed = fp.owed_lock(at(11, 10))
-    assert owed is not None
-    assert owed.id == "rules"
-    fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(11, 10))
-    p = fp.list_projects()[0]
-    assert states(p, at(11, 10))["rules"] == "late_done"
-    assert fp.owed_lock(at(11, 10, 1)) is None
-
-
-def test_on_time_is_done_at_the_boundary(tmp_path: Path) -> None:
-    make()
-    fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(10, 23, 58))
-    p = fp.list_projects()[0]
-    assert states(p, at(11, 8))["rules"] == "done"
-    assert states(p, at(10, 23, 59))["rules"] == "done"
-
-
-def test_slot_needs_entry_and_confirm_tap() -> None:
-    p = make()
-    fp.record_evidence("iv", "slot", "slot", "2026-11-26T20:00:00", now=at(25, 9))
-    assert states(p, at(25, 10))["slot"] == "pending"
-    assert states(p, at(26, 12))["slot"] == "missed"  # entered, never confirmed
-    fp.record_evidence("iv", "slot", "slot_confirm", value=True, now=at(25, 11))
-    assert states(p, at(25, 12))["slot"] == "done"
-    fp.record_evidence("iv", "slot", "slot", "2026-11-26T21:00:00", now=at(25, 13))
-    assert states(p, at(25, 14))["slot"] == "pending"  # moved: confirm again
-    fp.record_evidence("iv", "slot", "slot_confirm", value=True, now=at(26, 13))
-    assert states(p, at(26, 14))["slot"] == "late_done"
-    assert not [s for s in states(p, at(26, 14)).values() if s == "unknown"]
-
-
-def _phone_setup() -> None:
-    make()
-    fp.answer_weekend("iv", SAT1, trip=True, now=at(11, 9))
-
-
-def test_phone_milestone_states() -> None:
-    _phone_setup()
-    p = fp.list_projects()[0]
-    mid = "phone:2026-11-14"
-    assert states(p, at(13, 13))[mid] == "missed"  # no tap, no check
-    fp.record_evidence("iv", mid, "tap", value=True, now=at(12, 9))
-    assert states(p, at(13, 13))[mid] == "unknown"  # tap but nothing checked
-    evidence.append("iv", mid, "check", value=False, now=at(13, 11))
-    assert states(p, at(13, 13))[mid] == "missed"  # checked and failing
-    evidence.append("iv", mid, "check", value=None, now=at(13, 11, 30))
-    assert states(p, at(13, 13))[mid] == "unknown"  # source went dark
-    evidence.append("iv", mid, "check", value=True, now=at(13, 11, 40))
-    assert states(p, at(13, 13))[mid] == "done"
-
-
-def test_phone_passed_after_due_is_late_done() -> None:
-    _phone_setup()
-    mid = "phone:2026-11-14"
-    fp.record_evidence("iv", mid, "tap", value=True, now=at(12, 9))
-    evidence.append("iv", mid, "check", value=True, now=at(13, 14))
-    assert states(fp.list_projects()[0], at(13, 15))[mid] == "late_done"
-
-
-def test_unknown_locks_like_missed(tmp_path: Path) -> None:
-    _phone_setup()
-    mid = "phone:2026-11-14"
-    fp.record_evidence("iv", "rules", "file", file=shot(tmp_path), now=at(10, 9))
-    fp.answer_weekend("iv", SAT2, trip=False, now=at(11, 9))
-    fp.record_evidence("iv", mid, "tap", value=True, now=at(12, 9))
-    owed = fp.owed_lock(at(13, 13))
-    assert owed is not None
-    assert (owed.id, owed.state) == (mid, "unknown")
-
-
-def test_unreadable_registry_fails_open_and_loud(
-    data: Path, caplog: pytest.LogCaptureFixture
-) -> None:
-    data.mkdir()
-    (data / "projects.json").write_text("{nope")
-    with caplog.at_level("WARNING"):
-        assert fp.owed_lock(at(13, 8)) is None
-    assert "unreadable" in caplog.text
-
-
-def test_no_project_no_lock() -> None:
-    assert fp.owed_lock(at(13, 8)) is None
-
-
-def test_focus_project_never_locks_on_milestones() -> None:
-    fp.create_project(
-        "f",
-        name="F",
-        start=date(2026, 11, 10),
-        end=date(2026, 11, 20),
-        replaces=["leetcode"],
-        daily_minutes=30,
-        unlock_minutes=10,
-    )
-    assert fp.owed_lock(at(13, 8)) is None

focus_project/focus_project/tests/test_milestones_mock.py

@@ -1,61 +0,0 @@
-# Copyright (c) 2026 Krzysztof Rudnicki
-"""The mock milestone: a score counts only after a >= 90 min focus session."""
-
-from __future__ import annotations
-
-import focus_project as fp
-from focus_project.tests.helpers import at, make, session, states
-
-
-def test_mock_needs_a_score_after_a_long_session() -> None:
-    p = make()
-    assert states(p, at(26, 8))["mock"] == "missed"
-    session(at(25, 20), 90)
-    fp.record_evidence("iv", "mock", "score", 64, now=at(26, 8))
-    assert states(fp.list_projects()[0], at(26, 9))["mock"] == "late_done"
-
-
-def test_score_without_a_session_does_not_complete_mock() -> None:
-    p = make()
-    fp.record_evidence("iv", "mock", "score", 64, now=at(24, 9))
-    assert states(p, at(24, 10))["mock"] == "pending"
-    assert states(p, at(26, 8))["mock"] == "missed"
-    assert fp.owed_lock(at(26, 8)) is not None
-
-
-def test_short_session_does_not_complete_mock() -> None:
-    p = make()
-    session(at(24, 7), 89)
-    fp.record_evidence("iv", "mock", "score", 64, now=at(24, 9))
-    assert states(p, at(24, 10))["mock"] == "pending"
-
-
-def test_session_broken_by_a_gap_does_not_complete_mock() -> None:
-    p = make()
-    session(at(24, 6), 60)  # 06:00-07:00, then an 11 min hole (> 600 s)
-    session(at(24, 7, 11), 60)
-    fp.record_evidence("iv", "mock", "score", 64, now=at(24, 9))
-    assert states(p, at(24, 10))["mock"] == "pending"
-
-
-def test_session_must_end_before_the_score() -> None:
-    p = make()
-    session(at(24, 8), 90)  # ends 09:30
-    fp.record_evidence("iv", "mock", "score", 64, now=at(24, 9))
-    assert states(p, at(24, 10))["mock"] == "pending"
-    fp.record_evidence("iv", "mock", "score", 66, now=at(24, 9, 30))
-    assert states(p, at(24, 10))["mock"] == "done"  # ends exactly at the score
-
-
-def test_other_projects_sessions_do_not_count() -> None:
-    p = make()
-    session(at(24, 7), 90, pid="other")
-    fp.record_evidence("iv", "mock", "score", 64, now=at(24, 9))
-    assert states(p, at(24, 10))["mock"] == "pending"
-
-
-def test_session_on_time_but_score_late_is_late_done() -> None:
-    p = make()
-    session(at(25, 20), 100)
-    fp.record_evidence("iv", "mock", "score", 64, now=at(26, 0, 5))
-    assert states(p, at(26, 1))["mock"] == "late_done"