Cache read is charged per API round-trip on the whole conversation so far, so a long session's cost is driven by TURN COUNT, not message length. One measured session spent 267M cache-read against 856k output — a ratio of 290:1. The report priced tokens but never counted the turns spending them. Across the 263 stored transcripts this now shows 33,508 tool turns at an 8% batching rate, of which 4,437 (13%) decided nothing: polling a counter, running a linter, checking whether a push landed. At 606k average context re-read per turn that is ~2.7B cache-read tokens on turns a script could have taken. - turns.py classifies a command as poll / lint_test / vcs_check / substantive, deliberately conservative: state-changing commands (git commit, push) count as substantive even though they are scriptable, so the reported waste is a floor. A turn is only mechanical when EVERY command in it is. - Batching is grouped by requestId, not by transcript line: one API response can span several lines, and counting per line reports every turn as single-tool (0% batching where the real rate is 12%). - records.py gains TurnEfficiency (schema v3, additive — old records read as zeros) with derived rates written into to_dict so readers need not recompute. - report.py gains a "Turn efficiency" section pricing each mechanical kind. - _TurnTracker owns both the totals and the open turn, after pylint flagged _Accumulator over its attribute budget; _Accumulator.turns is now a property, so every existing caller is unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .github | ||
| .hippo | ||
| .vscode | ||
| bucket_catch | ||
| docs | ||
| kcd2_dice_solver | ||
| linux_configuration | ||
| meta | ||
| phone_focus_mode | ||
| python_pkg | ||
| third_party/agent-skills | ||
| .copilotignore | ||
| .fvmrc | ||
| .gitignore | ||
| .pre-commit-config.yaml | ||
| .prettierignore | ||
| CLAUDE.md | ||
| DESIGN_AUDIT_TODO.md | ||
| lint_python.sh | ||
| pyproject.toml | ||
| README.md | ||
| requirements.txt | ||
| run.sh | ||
testsAndMisc
A collection of personal projects, scripts, and experiments — from a GPS-based phone focus tool to Linux/Arch automation, with CI, linting, and pre-commit hooks across the board.
Highlights
Phone Focus Mode
Location-based app restriction for rooted Android. Automatically disables non-whitelisted apps within 500 m of home using ADB + Magisk. Features Haversine distance calculation, hysteresis to prevent toggling, fail-safe unlock, and state persistence. Bash, Android ADB.
Linux Configuration
Automated Arch Linux setup: fresh-install scripts, i3 window manager config, LaTeX environment, and system tests. Includes documentation and test result logging.
Scripts
Utility scripts for development workflows — build file validation, secret detection, and custom makepkg helpers.
Repository Layout
| Path | Description |
|---|---|
python_pkg/ |
Python packages (each maintained subpackage lives here) |
linux_configuration/ |
Arch Linux setup, i3 config, system maintenance scripts |
phone_focus_mode/ |
GPS-based Android focus enforcer |
scripts/ |
Workspace-level helper scripts and pre-commit hooks |
docs/ |
Reference docs and historical reports |
third_party/ |
Vendored upstream skills/agents |
meta/ |
Repo-wide tooling: pyproject.toml, requirements.txt, .pre-commit-config.yaml, run.sh, lint_python.sh, .fvmrc. Symlinked into the repo root so tools that auto-discover from root keep working. |
Archived / unmaintained projects live in the sibling repository
testsAndMisc-archive.
Tooling
- Python linting: Ruff with all rules enabled (see
meta/pyproject.toml) - Dependencies:
pip install -r meta/requirements.txt(combined runtime + dev) - CI: GitHub Actions — lint, build, and test on push
- Testing: pytest (Python), custom shell-based test harness for scripts