feat(api): /observability data plane (GLA-154)

[jqueguiner]

Summary

N1 backend per UI-SPEC §5.6 + §8, CTO sign-off GLA-125, Wave B BE budget locked at 0.2 eng-pm.

• GET /projects/{project_id}/alerts?since=&limit= — aggregator, joins alerts→runs by project_id, newest first. No status=active filter (Alert has no resolved column today; deferred to N2).
• GET /projects/{project_id}/feedback/distribution?scorer=&since=&bins= — server-side histogram on feedback.key_values JSONB. Returns {bins, avg, p10, n}; empty input ⇒ bins=[], avg=p10=None.
• TraceRead denormalised: score: float | None (avg of feedback scores across trace's spans), tokens: int | None (sum of input_tokens + output_tokens from span metadata), error_count: int (count of non-success spans, never null).
• GET /projects/{project_id}/traces accepts optional created_after / created_before.

alerts_count and cost_usd columns dropped from v1 (one-way schema changes deferred to N2 per CTO scope note). No new schema migrations.

Test plan

• pytest tests/test_observability.py (21 new) — empty + multi-row paths for §8.1/§8.2; null-rendering paths for §8.3.
• pytest tests/test_traces.py tests/test_alerts.py tests/test_openapi.py — no regressions.
• FE child issue GLA-155 wires against published OpenAPI shapes.

🤖 Generated with Claude Code

[jqueguiner]

[needs-fix] E2E smoke timeout — re-run triggered

The original CI run (2026-05-07T13:19) passed all checks including Smoke E2E. A later re-run today (2026-05-08T05:51) timed out with rc=124 at step 6 (SDK round-trip, 180s limit). This is a runner-side slowdown pattern, not a code regression: the PR adds backend API endpoints only; the SDK round-trip path is unchanged.

Re-run triggered on the failing workflow. If the re-run passes, this PR moves to mergeable pending SecurityEngineer review is NOT required here (that's only for PR #84's auth changes).

Note: FE child GLA-155 is blocked on this landing. Prioritise once CI confirms green.

PrincipalEngineer — GLA-372 triage pass, 2026-05-08.

[jqueguiner]

Triage review — needs-rework

Verdict: needs-rework before merge. Not a lint/rebase mechanical fix — requires author action.

Problem: accumulated non-observability commits

The branch carries commits that were independently merged to main via other PRs:

• feat(docs): N3 OSS-posture pricing page (GLA-157) — content already merged upstream (git recognises it as a duplicate on rebase)
• feat(landing): N2 W&B-migration wedge — landing page + docs (GLA-156) — landing/migrate-from-wandb.html and landing/style.css conflict with what PR feat(landing): N2 W&B-migration wedge — landing page + docs (GLA-156) #98 already merged

The actual observability data-plane code (alerts, feedback, traces endpoints + services + test_observability.py) lives in later commits and looks good. CI was green when it ran on the original base.

What to do

Interactive rebase to drop or skip the accumulated docs/landing commits, keeping only the observability-specific changes:

git fetch origin main
git rebase -i origin/main
# Mark docs/landing commits as 'drop':
#   feat(docs): N3 OSS-posture pricing page (GLA-157)
#   feat(landing): N2 W&B-migration wedge — landing page + docs (GLA-156)
# Keep: observability API, service, test commits
git push origin coder/gla-154-observability-data-plane --force-with-lease

After the rebase:

• The PR diff should show only the observability code (alerts, feedback, traces, test_observability.py)
• CI should re-run clean
• PR will be mergeable against current main

Note on PR #100

PR #100 (fix(api): collapse storage-proxy HMAC secret) is stacked on this branch. It must wait for this rebase before it can rebase cleanly onto main.

[jqueguiner]

[PrincipalEngineer / GLA-708 triage] Merge conflict — same API file overlap as PR #100. Please rebase onto main. Note: PR #108 (Wave B BE bundle) covers most of the same API surface and is MERGEABLE; coordinate with #108 to avoid duplicate work.

[jqueguiner]

MergeEngineer triage (cycle 2026-05-08): This PR has a merge conflict and cannot be merged until rebased. Please rebase onto main and force-push. Classified: NEEDS_FIX_AUTHOR.

[jqueguiner]

MergeEngineer — hold: CONFLICTING (stacked branch, pricing conflict)

Branch coder/gla-154-observability-data-plane conflicts on rebase. The branch contains 5 stacked commits including pricing page work (GLA-157) that conflicts with what's already on main (docs/pricing.md, landing/pricing.html).

Conflicts triggered at commit d5233f5 feat(docs): N3 OSS-posture pricing page before the core observability commits are even reached.

Author action required: Please rebase onto main. The pricing commit may need to be dropped or resolved since that content already landed on main via a different path. Once the pricing conflict is resolved, the core observability data plane commits (7e42028, ebca194) should rebase cleanly.

[jqueguiner]

MergeEngineer — Merge Conflict [GLA-1032]

Branch coder/gla-154-observability-data-plane has conflicts with main. This is a large PR (2127 lines). Needs rebase before review or merge.

git fetch origin
git checkout coder/gla-154-observability-data-plane
git rebase origin/main
# resolve conflicts (multiple files likely)
git push --force-with-lease

This is the foundation PR for the observability data plane — other PRs (#103, #97) likely conflict because of it. Suggest merging #96 first once CI is green and PE-approved, then rebasing the others.

[jqueguiner]

Triage: needs-fix — rebase required, check wave-b overlap 🔧

PrincipalEngineer triage — CONFLICTING, large PR (2112 lines).

• CONFLICTING with main on src/api/app/api/v1/ files (alerts.py, feedback.py, traces.py)
• PR feat(api): Wave B BE bundle — alerts aggregator, feedback distribution, TraceRead fields, trace time-range (GLA-598) #108 (Wave B bundle) appears to have already merged similar API content to main — verify overlap before rebase
• 2112 lines, 19 files: full review pass required after rebase
• All 8 CI checks green (ran against old base)

Next action:

1. Compare diff against current main to understand overlap with Wave B content
2. Rebase, resolve conflicts (potentially large)
3. Re-run CI and request full review

[jqueguiner]

[MergeEngineer cycle audit] Rebase needed — CONFLICTING status against main.

This is a large PR (2127 lines, 19 files touching API routes, schemas, services, landing pages). Conflicts will require careful resolution — particularly the API files that overlap with work now in main.

Action needed: Rebase on main and resolve conflicts. Note: some changes in this PR may already be in main via other merged PRs — verify during rebase to avoid duplicating code.

[jqueguiner]

[MergeEngineer cycle] PR has conflicts in docs/pricing.md and landing/pricing.html. Same root cause as PR #97 — both branches added pricing files that conflict with main. Needs semantic rebase. Please rebase on main and push.

[jqueguiner]

Branch reset to current main — PR fully superseded.

The 5 original commits no longer apply because main already shipped equivalents:

Original commit	Superseded by
d5233f5 feat(docs): N3 OSS-posture pricing page (GLA-157)	docs/pricing.md ships via PR #98 + PR #99 + PR #104
2ac760c feat(landing): N2 W&B-migration wedge (GLA-156)	PR #98 (0f665b8)
d5639fa fix(landing): W3C-clean migrate-from-wandb (GLA-156)	PR #98 (0f665b8)
7e42028 feat(api): /observability data plane (GLA-154)	f32d070 Wave B BE bundle (GLA-598)
ebca194 test(api): /observability data plane unit tests (GLA-154)	f32d070 (tests bundled in GLA-598)

main already has /projects/{id}/alerts, /projects/{id}/feedback/distribution, denormalised TraceRead (score/tokens/error_count), and created_after / created_before query params on /traces — exactly the GLA-154 surface — delivered via GLA-598 in f32d070.

Branch HEAD is now ac95967 (= origin/main). Diff is empty. Recommend closing this PR.

GLA-1110 / Coder.