feat(workspace): runInWorkspace seam — compounding worker clones over gitWorkspace#305
Merged
Merged
Conversation
…shared gitWorkspace
- runInWorkspace(ws, body, {commitOnInvalid}): materialize a clone -> run -> commit-back,
so a worker builds on the prior worker's committed artifact (test: a second worker builds
on the first, real git + python). Exported from the runtime barrel.
- atom-mcp-e2e: rewired onto a shared bare-repo gitWorkspace; workers compound (live run
delivered with a real commit rev) instead of isolated cwds.
- fix(bench): atom-humaneval tree element type (NonNullable) — clears the bench typecheck.
- docs(bench/HARNESS): supervisor-vs-blind is runBenchmark([sample,refine]) + the multistep
Environment via runStrategyEvolution; no bespoke gate/fitness/hillclimb (those were a
reinvention of the published suite).
- fix(skill): build-with-agent-runtime — createTopologyView/renderTopologyTree are /topology.
tangletools
approved these changes
Jun 16, 2026
tangletools
left a comment
tangletools
left a comment
Contributor
There was a problem hiding this comment.
✅ Auto-approved PR — 0ccfa2ba
Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.
tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-16T09:53:33Z
This was referenced Jun 16, 2026
drewstone
added a commit
that referenced
this pull request
Jun 16, 2026
…#307) * chore(release): 0.54.0 — expose the recursive agent atom + workspace seam Publishes coordinationDriverAgent, serveCoordinationMcp, driverChild (the recursive driver atom, #304) and runInWorkspace (the shared-workspace seam, #305), which merged into main without a version bump and so were absent from the published 0.53.0. Additive (new exports on the /runtime barrel) — no breaking changes, no fleet bump required. Verified: dist/runtime.js exposes all four; tests 945 pass. * fix(lint): clear 2 biome errors blocking the release (assignment-in-expr, unused import, import order) --------- Co-authored-by: Drew Stone <hello@webb.tools>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
A small seam so the supervisor's workers compound through one shared artifact instead of running in isolated cwds, plus the bench fixes/docs from the same session.
runInWorkspace(ws, body, {commitOnInvalid})(src/runtime/workspace.ts, exported from the runtime barrel) — materialize a clone of a sharedgitWorkspace→ runbody(cwd)→ commit the delivery back. A second worker builds on the first's committed output. ~25 lines over the existinggitWorkspace; no new workspace machinery.tests/loops/workspace.test.ts): "a second worker builds on the first, gated by a real test" — real git + realpython3, green.atom-mcp-e2erewired onto a shared bare-repogitWorkspace; workers compound (live run delivered with a real commitrev) instead of isolated cwds.atom-humanevaltree-element type viaNonNullable— clears the bench typecheck (pre-existing error).runBenchmark([sample, refine])+ the multistepEnvironmentviarunStrategyEvolution— the published suite, not a bespoke harness.build-with-agent-runtime—createTopologyView/renderTopologyTreeimport from/topology, not root.Verification
tscclean; benchtsc0 errors; biome clean on changed files.vitest run tests/loops/workspace.test.ts→ 3 passed, 1 skipped (jj absent).origin/main.Notes
The supervisor experiments (
supervised-commit0,observe-driver) intentionally do not land here — they move to the newsupervisor-labrepo (the product/experiment layer). This PR keepsagent-runtimeto the substrate seam + fixes.