Skip to content

Export benchmark patch artifacts - #225

Closed
sethkarten wants to merge 1 commit into
mainfrom
feature/benchmark-patch-artifacts
Closed

Export benchmark patch artifacts#225
sethkarten wants to merge 1 commit into
mainfrom
feature/benchmark-patch-artifacts

Conversation

@sethkarten

@sethkarten sethkarten commented Jun 23, 2026

Copy link
Copy Markdown
Contributor

Summary

  • add a benchmark patch artifact exporter that writes patch.diff, metadata.json, summary.md, and optional trajectory.jsonl
  • capture tracked and untracked git diffs plus commit, branch, file lists, session stats, tests, failures, and final status metadata
  • mark dependency/setup churn as rejected_setup_churn unless explicitly allowed
  • expose the exporter through AgentSession, agent connections, /export patch [dir], and prime-agent daemon patch-artifact <session> [dir]

Testing

  • npm --prefix packages/coding-agent test -- patch-artifact.test.ts
  • ./node_modules/.bin/tsgo --noEmit
  • npm run check
  • isolated daemon smoke with temp socket/session dirs: create git repo, create session, export daemon patch-artifact, validate patch and metadata

Note

Add patch-artifact export command for benchmark patch artifacts

  • Adds patch-artifact.ts, a new module that exports createPatchArtifact, which captures git diffs (tracked and untracked), writes patch.diff, metadata.json, summary.md, and optionally trajectory.jsonl to an output directory.
  • Adds a patch-artifact <session> [dir] CLI subcommand to the daemon client and an /export patch [dir] interactive slash command variant.
  • Extends the AgentConnection interface and all implementations (daemon, in-process, deferred) with exportPatchArtifact(outputDir?), and adds a corresponding export_patch_artifact daemon protocol message.
  • Filters setup/dependency churn files (e.g. lockfiles, node_modules) by default, reporting them as rejectedFiles; set allowSetupChurn to include them.

Macroscope summarized f8c719f.

@sethkarten

Copy link
Copy Markdown
Contributor Author

Closing as out of scope for the autonomous-mode work. Patch extraction is not needed for this path right now; #222 covers autonomous continuation and gate-driven terminal evidence.

@sethkarten sethkarten closed this Jun 23, 2026
@kevinjosethomas
kevinjosethomas deleted the feature/benchmark-patch-artifacts branch June 26, 2026 00:37
@kevinjosethomas
kevinjosethomas restored the feature/benchmark-patch-artifacts branch June 26, 2026 00:37
@kevinjosethomas
kevinjosethomas deleted the feature/benchmark-patch-artifacts branch July 8, 2026 00:30
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant