[ajianaz]

What

Part 3 of 3 of tole doctor: the persisted coverage ledger and the READY n-of-m summary. Builds on #388 (offline/catalog) and #390 (full).

Why

#378. Release validation needs the result of a run to outlive the terminal, and a one-line answer to "how many providers are proven, and what blocks the first one that is not?".

Changes

• doctor.rs: ledger at $CODECORA_HOME/tole/doctor-ledger.json — per provider + model: level reached, first blocker, fix command, timestamp. The provider identity is the redacted URL (no userinfo/query/fragment); the API key is never stored. Latest run wins per provider + model.
• Atomic write (temp file + rename); the temp file is created exclusively with mode 0600 after removing any stale file, so a pre-planted file or symlink is never written through. A missing ledger is empty; a corrupt one is an error and is never silently replaced.
• main.rs: tole doctor --summary prints READY n-of-m + first blocker + fix without running anything; --no-ledger skips recording.
• README, CONTRIBUTING (release validation result lands in the ledger), CHANGELOG.

Testing

• 4 new unit tests: record/summary/roundtrip and latest-wins, missing vs corrupt ledger, stale-tmp symlink not written through with 0600 mode, owner-only file mode.
• fmt, clippy -D warnings, cargo test --workspace pass; cora review --staged clean.
• Debug binary end to end against a local mock: full green run records READY 1-of-1, a no-tool-calling provider exits 1, the ledger file is 0600 and contains neither the API key nor URL credentials.
• Not done: a live run against a real provider (none configured here); feat(cli): provider doctor with tiered diagnostics + persisted coverage ledger #378 stays open for that.

Refs #378

[github-actions]

🔍 Cora AI Code Review

✅ No issues found. Code looks good!

---

Review powered by cora-code · BYOK · MIT