M1: Implement OpenAI Codex capacity collector #13

Merged

creatidy merged 15 commits from m1-openai-codex-collector into main

2026-09-04 20:56:32 +00:00

creatidy commented

2026-09-03 20:59:49 +00:00

(Migrated from github.com)

Closes #12 (left open for human review tracking).

Scope

Read-only OpenAI/Codex app-server capacity collector only. No routing, status/CLI, Ollama, model execution, prompts, retries, credential inspection, or live provider/account calls.

Implementation

Validated rateLimitResetCredits summaries, including {availableCount: 0, credits: []}, are supplemental telemetry and do not represent current quota exhaustion or withhold valid quota percentages.
Unsupported platforms are classified before root enumeration: discovery is limited to evidenced Linux x86-64 layouts; Linux ARM64 and Darwin are unsupported.
Additional bucket identities, exact tagged mirrors, strict schema/protocol/JSON validation, descriptor-bound no-follow discovery/execution, bounded output/timeouts, no prompts, and secret safety remain intact.

Verification

Focused Codex parser/acquisition tests: 181 passed.
make check: 400 tests passed. uv run basedpyright: 0 errors, 0 warnings, 0 notes.
git diff --check: clean. Linear history, no merge commits, normal non-force push.

Bounded limitations

No live-account test or live success recapture; transport tests use synthetic process fakes.
Discovery is limited to evidenced VS Code ChatGPT roots and Linux x86-64 descriptor execution; Darwin, ARM64, PATH/npm/standalone installs remain unsupported.
Protocol error classification, binary/version reporting, freshness policy, and CLI/status integration remain outside this issue.

Closes #12 (left open for human review tracking). ## Scope Read-only OpenAI/Codex app-server capacity collector only. No routing, status/CLI, Ollama, model execution, prompts, retries, credential inspection, or live provider/account calls. ## Implementation - Validated `rateLimitResetCredits` summaries, including `{availableCount: 0, credits: []}`, are supplemental telemetry and do not represent current quota exhaustion or withhold valid quota percentages. - Unsupported platforms are classified before root enumeration: discovery is limited to evidenced Linux x86-64 layouts; Linux ARM64 and Darwin are unsupported. - Additional bucket identities, exact tagged mirrors, strict schema/protocol/JSON validation, descriptor-bound no-follow discovery/execution, bounded output/timeouts, no prompts, and secret safety remain intact. ## Verification - Focused Codex parser/acquisition tests: 181 passed. - `make check`: 400 tests passed. `uv run basedpyright`: 0 errors, 0 warnings, 0 notes. - `git diff --check`: clean. Linear history, no merge commits, normal non-force push. ## Bounded limitations - No live-account test or live success recapture; transport tests use synthetic process fakes. - Discovery is limited to evidenced VS Code ChatGPT roots and Linux x86-64 descriptor execution; Darwin, ARM64, PATH/npm/standalone installs remain unsupported. - Protocol error classification, binary/version reporting, freshness policy, and CLI/status integration remain outside this issue.

Rows
Columns