Skip to content

fix(issues): expose actions, proposed_examples, and traces in issues list --format json#144

Merged
Palash Shah (Palashio) merged 3 commits into
mainfrom
fix/issues-list-expose-actions
May 29, 2026
Merged

fix(issues): expose actions, proposed_examples, and traces in issues list --format json#144
Palash Shah (Palashio) merged 3 commits into
mainfrom
fix/issues-list-expose-actions

Conversation

@Palashio
Copy link
Copy Markdown
Contributor

@Palashio Palash Shah (Palashio) commented May 28, 2026

Summary

issueToMap was stripping three fields agents need to read and update issues effectively:

  • actions — the evaluator spec; needed to test the existing evaluator against new traces and update it if the pattern has changed
  • proposed_examples — per-trace regression example assertions; needed to see what assertions already exist and re-propose with better ground truth when available
  • traces — linked evidence runs; needed to sample existing traces for evaluator regression checks

All three are now included when non-empty, decoded from raw JSON so they appear as structured objects.

Why this matters: The companion langchainplus PRs teach the issues agent a read→test→update loop for both evaluators and regression examples, symmetric with how it already reads and updates issue descriptions. Without these fields in the CLI output, agents can't see what already exists before deciding whether to update.

Related langchainplus PRs:

Test Plan

  • langsmith project issues list --project <name> --format json includes actions, proposed_examples, and traces on issues that have them
  • Issues with no actions / proposed_examples / traces omit those keys (no empty arrays or nulls)

Release Note

none

`issueToMap` was omitting the `actions` field (evaluator spec) and
`traces` field (linked evidence runs) from the JSON output of
`project issues list` and `project issues update`. Agents reading
the issues list couldn't see the existing evaluator, so they had no
way to run new evidence traces against it or update the evaluator
when it no longer covered a newly-linked trace.

Both fields are now included when non-empty, decoded from raw JSON
so they appear as structured objects rather than escaped strings.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add proposed_examples (per-trace regression example assertions) to
issueToMap so agents can read existing assertions alongside the
evaluator spec (actions) and linked traces.

Symmetric with the evaluator read→test→update loop: agents can now
read what assertions already exist for a trace, re-derive from the
trace evidence, and re-propose if they have concretely better ground
truth. The backend upserts on (issue_id, run_id) conflict, so
re-running propose overwrites rather than ignoring.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@Palashio Palash Shah (Palashio) changed the title fix(issues): expose actions and traces in issues list --format json fix(issues): expose actions, proposed_examples, and traces in issues list --format json May 28, 2026
@Palashio Palash Shah (Palashio) merged commit 4b04949 into main May 29, 2026
10 checks passed
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.

2 participants