Read a Flow job's report over the API and with oh run report issue #348 closed: completed andymccall opened this 7 days ago · 0 comments

Friction report from the agent working in doomsdayone/tetriblox (2026-10-28): a Flow job can publish a report bundle (docs/reports.md: report.toml, data/*.csv, index.html…), but only the web page can read it back. The API has no route for a report and oh has no command, so the agent that produced a report can't analyse it, and reproducing the data means an 8-minute emulator capture on a quiet machine.

Example: doomsdayone/tetriblox run issue #26, job performance: report performance: 9 files, 696917 bytes, manifest a9c8b97ed497, with data/frames.csv. These all 404 ("no such API route"): /repos/O/N/reports, /reports/performance, /flow/26/reports, /flow/26/jobs/performance/report.

Wanted, in order:

1. oh run report RUN JOB (prints report.toml; --list for the manifest, --file data/frames.csv for one file), mirroring oh run log.
2. The API under it: GET /repos/O/N/flow/runs/N/jobs/JOB/report (manifest) and …/report/{path} (one file), and the same by report id (what the web URL carries).
3. Later: the newest report of a kind on a bookmark ("the latest performance report on main"), for comparisons over time.

Andy: not urgent; for the overnight loop.