Repository navigation
Record a model for background work a remote Planner ran - #731
Merged
Merged
Conversation
Under HARNESS=remote an operator may leave MODEL unset, and the attached client chooses its own model. Every background worker recorded config.model as its artifact's model provenance, and the artifact validators refuse an empty model, so research, document descriptions, research briefs and History summaries all failed after their model work had finished, with "model is invalid" or "artifact is invalid". Artifacts now record "attached-client" when no model is configured. What a worker asks for is unchanged: an empty model still reaches the attached client as no request. Assistant-model: Claude Opus 5.5 (fast) Assistant-workflow: inline Assistant-duration: 1h32m converged, estimated unmeasured Assistant-verification: agent-browser manual scenario passed: on a local HARNESS=remote server with MODEL unset and a real attached Planner, /research completed and published a 7-source report; before the fix the same run failed with "model is invalid" Assistant-verification: red-before-fix passed: the new research-workspace test fails with "model is invalid" without the change Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test (6737 pass, 0 fail) Assistant-verification: typecheck and lint passed: bun run types and bun run ci User-preference: Do not reveal internal details of downstream deployments in public Chopin artifacts Co-authored-by: Alex Lavaee <lavaman131@github.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Follow-up to #730. Once #730 let background workers run on a remote document's attached Planner, they still failed after doing their model work. Under
HARNESS=remotean operator may leaveMODELunset, and the attached client then chooses its own model. Every worker recordedconfig.modelas its artifact'smodelprovenance, and the artifact validators refuse an empty one. So:model is invalidafter the evidence was gathered;Background job … artifact is invalid.Fix.
workerModel()injobs/worker-session.tsrecordsattached-clientwhen no model is configured. The research evidence and answer artifacts, document description, research brief and History summary all use it. What a worker asks for is unchanged: an empty model still reaches the attached client as no request.Verified end to end. I ran a local server with
HARNESS=remote,BACKGROUND_JOBS=onandMODELunset, with a real Planner client attached over the agent connection./researchgathered evidence with twoweb_searchcalls on the client, published a report with 7 sources as a child document, and no job failed. Before this change the same run failed withmodel is invalid.Tests. A new
research-workspace.test.tscase runs the evidence definition withmodel: ""and parses its artifact. It fails withmodel is invalidwithout the change.Assistant-model: Claude Opus 5.5 (fast)
Assistant-workflow: inline
Assistant-duration: 1h32m converged, estimated unmeasured
Assistant-verification: agent-browser manual scenario passed: local HARNESS=remote server with MODEL unset and a real attached Planner; /research published a 7-source report, and failed with "model is invalid" before the fix
Assistant-verification: red-before-fix passed: the new research-workspace test fails without the change
Assistant-verification: bun test unit passed: GIT_CONFIG_GLOBAL=/dev/null bun test (6737 pass, 0 fail)
Assistant-verification: typecheck and lint passed: bun run types and bun run ci
User-preference: Do not reveal internal details of downstream deployments in public Chopin artifacts
Co-authored-by: Alex Lavaee lavaman131@github.com