* docs(upstream-sync): add triage ledger + cherry-pick hook plan
Seeds the upstream triage ledger (52 diverged commits from langchain-ai/open-swe
categorized: landed/won't-merge/deferred/untriaged) and the design plan for a
git-hook mechanism to keep it in sync during cherry-picks.
* feat(upstream-sync): jsonl-backed triage ledger + generator CLI
triage.jsonl is now the source of truth (52 rows migrated from triage.md);
triage.md is generated with a do-not-edit banner. scripts/triage.py provides
migrate/generate/reconcile/lookup/check-reject/set; make triage-render/check/reconcile
added (triage-check is CI-safe staleness gate). Stdlib-only so git hooks can call it.
* feat(upstream-sync): cherry-pick triage git hooks + git cp wrapper
post-commit journals each -x pick to an untracked .git-local journal; prepare-commit-msg
hard-blocks known-reject picks (commit-msg is a secondary backstop — clean picks skip it on
git 2.50.1), overridable via git cp --force / SH_CHERRYPICK_ALLOW_REJECT=1 /
sh.cherrypick.blockRejects=false. git-cp is the pre-apply guard + auto-reconcile. pre-push is
a SHIM that re-execs the global Sea Haven security pre-push so core.hooksPath=.githooks does
not shadow it; install-hooks.sh verifies that shim FIRST and refuses if it is missing.
* docs(upstream-sync): correct hook plan + git cp runbook
Record the verified git 2.50.1 finding that clean cherry-picks skip commit-msg, so the block
lives in prepare-commit-msg; note the locked HARD-BLOCK-by-default reject policy. CHERRYPICK.md
now leads with make install-hooks + git cp and explains the generated-md ledger.
* chore(upstream-sync): mark #1651 landed (Bedrock family fix on gateway-routing)
* docs(upstream-sync): rewrite CHERRYPICK.md as a repo-specific runbook
* feat(upstream-sync): add triage.py sync + make triage-sync
Discovers commits on dev..upstream/main not yet in the ledger and appends them
as untriaged (PR # and subject parsed from each commit), then bumps _meta
'last synced' to the upstream tip and regenerates triage.md. Closes the
discovery side of the workflow: triage-sync to pull in new work, git cp to land it.
* docs(upstream-sync): move cherry-pick runbook to PR1 branch as cherry-pick-runbook.md
* feat: add reviewer graph + eval target wiring
- New `reviewer` graph (`agent/reviewer.py`) registered in langgraph.json
alongside the main `agent` graph. Reuses the same sandbox lifecycle,
GH proxy auth, and middleware primitives from `agent.server`, but with
a narrower tool set, a reviewer-specific system prompt, no
commit/push, and the `task` (subagent) tool stripped via
`_ToolExclusionMiddleware` so review stays in one context.
- New `github_comment` tool: agents call it once per issue with
`(file, line, body, severity)` and the eval scores those calls
against golden comments.
- `ensure_no_empty_msg` middleware (the no_op nudge) is intentionally
*not* on the reviewer's stack — that middleware exists to enforce the
main agent's "always finalize via Slack/Linear/PR" contract, which
the reviewer doesn't have. The main agent's behavior is unchanged.
- `evals/reviewer/target.py`: send PR info as a user message, extract
every `github_comment` tool call (multiple expected per review) into
the run output.
- `evals/reviewer/judge.py`: per-example evaluator now returns a list
of metrics under `{"results": [...]}` so LangSmith averages each
numeric key (f1/precision/recall/tp/fp/fn) across the experiment in
the UI. Dropped the broken `aggregate_pr` summary evaluator that
reached for an attribute that doesn't exist on `RunTree`.
- `evals/reviewer/run_eval.py`: `--limit` now slices the dataset via
`client.list_examples(limit=N)` since `aevaluate` doesn't accept
`max_examples`.
- Makefile: `dev` and `run` targets now use `uv run` so they work
without an activated venv.
* resolve comments
---------
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
