Skip to content

fix(backfill): Workers-AI timeout guard for Sync + re-land Docs reader restyle - #26

Merged
AndresL230 merged 3 commits into
mainfrom
fix/backfill-ai-timeout-guard
Jul 7, 2026
Merged

fix(backfill): Workers-AI timeout guard for Sync + re-land Docs reader restyle#26
AndresL230 merged 3 commits into
mainfrom
fix/backfill-ai-timeout-guard

Conversation

@AndresL230

@AndresL230 AndresL230 commented Jul 7, 2026

Copy link
Copy Markdown
Contributor

Two independent changes, bundled per request so the stranded Docs restyle rides an already-open PR to main.

1. Timeout-guard Workers AI so a hung summarize can't wedge Sync GitHub

Problem. The admin Sync GitHub button sticks forever on "Contacting GitHub — taking inventory of PRs and issues…". POST /admin/backfill calls Workers AI (@cf/google/gemma-4-26b-a4b-it, a slow 26B model) synchronously in the request the browser waits on. The await ai.run(...) in makeWorkersAiSummarizer is unbounded: when the daily Neuron budget is exhausted or the model is overloaded, the call can hang indefinitely (never resolves, never rejects). The surrounding try/catch only handles throws, so a hang wedges the whole request — observed as POST /admin/backfill never completing (>2 min) with 0 summary rows written.

Fix.

  • makeWorkersAiSummarizer races ai.run(...) against a timer (WORKERS_AI_TIMEOUT_MS, default 15s). A timeout rejects into the same catch, returns null, and the existing deterministic excerpt fallback fires — so the request always completes. Timeout is injectable (workersAiPrSummarizer(ai, timeoutMs?) / workersAiIssueSummarizer(ai, timeoutMs?)) for unit testing; clearTimeout in a finally leaves no pending timer on a healthy call.
  • SUMMARY_BATCH_LIMIT 20 → 5, so each /admin/backfill returns quickly and the frontend auto-loop shows progress instead of one long stall.
  • The catch's diagnostic console.error now also logs timeouts (visible in wrangler tail).

Test (TDD, failing-first). New workersAiSummarizer — timeout guard block: a fake ai whose run never resolves + a 50ms injected timeout must resolve to null / model='excerpt' within the timeout. Verified RED then GREEN.

2. Re-land the Docs reader restyle onto main

The Docs reader restyle + collapsible nav tree (originally PR #25) was accidentally merged into the already-merged feat/mywork-grid-seed-docs branch instead of main, so it never reached main. Its web-only commit (web/src/canopy.css, main.ts, markdown.ts, render.ts) is cherry-picked here. The stale backfill.ts on that branch — a pre-#23 revert — is deliberately not included.

Verification

  • npm run build:web → ok · npm run typecheck → clean (worker + web) · npx vitest run --exclude '**/.claude/**'422 passed
  • ⏳ Deployed-worker verification of the timeout path is pending (Workers AI can't run locally) — confirm on the deployed canopy worker via wrangler tail + a Sync click.

🤖 Generated with Claude Code

…'t wedge Sync

The Sync GitHub button stuck forever on "Contacting GitHub…" because
POST /admin/backfill awaits ai.run() unbounded: an exhausted Neuron budget
or overloaded 26B model can leave the call pending forever (never resolves,
never rejects), so the try/catch — which only handles throws — can't recover.

makeWorkersAiSummarizer now races ai.run against an injectable timer
(WORKERS_AI_TIMEOUT_MS, default 15s). A timeout rejects into the same catch,
returns null, and the existing deterministic excerpt fallback fires — the
request always completes. Shrink SUMMARY_BATCH_LIMIT 20 -> 5 so each batch
returns quickly and the frontend's auto-loop shows progress instead of a stall.

Test (failing-first): a fake ai whose run never resolves + a short injected
timeout must resolve to null / model='excerpt' within the timeout.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@coderabbitai

coderabbitai Bot commented Jul 7, 2026

Copy link
Copy Markdown

Review Change Stack

Warning

Review limit reached

@AndresL230, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 49 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: c5da8b87-6b2c-410d-9c73-171dbac8cc09

📥 Commits

Reviewing files that changed from the base of the PR and between 1038be7 and b363101.

📒 Files selected for processing (6)
  • src/tools/summarize.ts
  • test/summarize.test.ts
  • web/src/canopy.css
  • web/src/main.ts
  • web/src/markdown.ts
  • web/src/render.ts
📝 Walkthrough

Walkthrough

Adds a bounded timeout (WORKERS_AI_TIMEOUT_MS) to Workers AI summarizer calls using Promise.race, falling back to null (and thus excerpt summarization) on timeout or failure. Both public summarizer wrappers gain an optional timeoutMs parameter. Backfill's SUMMARY_BATCH_LIMIT is lowered from 20 to 5. Tests added for timeout fallback behavior.

Changes

Workers AI timeout and backfill batching

Layer / File(s) Summary
Timeout constant and race-based execution guard
src/tools/summarize.ts
Adds WORKERS_AI_TIMEOUT_MS and races ai.run against a timeout, returning null on timeout/throw/malformed output, clearing the timer in finally.
Public wrapper signatures and batch limit reduction
src/tools/summarize.ts, src/tools/backfill.ts
Updates workersAiPrSummarizer/workersAiIssueSummarizer to accept optional timeoutMs; lowers SUMMARY_BATCH_LIMIT from 20 to 5 with updated comments.
Timeout fallback tests
test/summarize.test.ts
Adds tests simulating a hanging AI call, verifying summarizers return null and storage functions fall back to excerpt summaries within timeout.

Estimated code review effort: 2 (Simple) | ~12 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Caller
  participant SharedSummarizer
  participant AiRun
  participant TimeoutTimer

  Caller->>SharedSummarizer: workersAiPrSummarizer/IssueSummarizer(ai, timeoutMs)
  SharedSummarizer->>AiRun: ai.run(...)
  SharedSummarizer->>TimeoutTimer: start timer(timeoutMs)
  alt AiRun resolves first
    AiRun-->>SharedSummarizer: model output
    SharedSummarizer-->>Caller: parsed summary
  else Timer fires first
    TimeoutTimer-->>SharedSummarizer: timeout rejection
    SharedSummarizer-->>Caller: null (fallback to excerpt)
  end
  SharedSummarizer->>TimeoutTimer: clear timer (finally)
Loading

Possibly related PRs

  • SaplingLearn/canopy#11: Both PRs modify the workersAi*Summarizer/shared summarizer logic in src/tools/summarize.ts.
  • SaplingLearn/canopy#12: Both PRs modify the Workers AI summarizer to return null on unusable ai.run() results, with overlapping test changes.
🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.
Description Check ✅ Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Title check ✅ Passed The title captures the main Workers-AI timeout/backfill fix, though the trailing docs restyle note appears unrelated.
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch fix/backfill-ai-timeout-guard

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Jul 7, 2026

Copy link
Copy Markdown

Deploying with  Cloudflare Workers  Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status Name Latest Commit Preview URL Updated (UTC)
✅ Deployment successful!
View logs
canopy b363101 Commit Preview URL

Branch Preview URL
Jul 07 2026, 06:20 AM

Frontend-only design pass on the Docs page, styled entirely with the theme
CSS vars so it holds in all three themes (dark / light / midnight).

Reader (web/src/render.ts, canopy.css):
- Centered reading column (~730px measure) with a wrap-safe title (drops the
  white-space:nowrap that could overflow long titles) and a mono
  `SPACE / SECTION` eyebrow.
- Markdown: stronger reading ink (color-mix off --fg), real heading hierarchy
  (h2 hairline rule, h3 distinct at 16px), language-tagged code panels,
  callout blockquotes (accent-soft), crisp inline-code chips, accent-numbered
  ordered lists, and editorial ledger-style tables (rule-only borders, mono
  uppercase headers).

Nav tree (web/src/render.ts, main.ts, canopy.css):
- Sections are now collapsible: a header button (rotating chevron + mono
  label + count) toggles each section; collapse state persists to
  localStorage, keyed `${space}:${section}`.
- Active doc is a clean --accent-soft pill matching the sidebar's active
  style; removes the inset accent line (.cnpy-selbar no longer used here —
  left intact for review.ts).

markdown.ts:
- Post-sanitize DOM pass wraps code blocks in a `.cnpy-code` panel (tagged
  with the fence language) and tables in a scroll wrapper. Browser-only;
  render tests mock this module.

Verified across dark/light/midnight in both Sapling and Canopy spaces.
typecheck green; full suite 418/418.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@AndresL230 AndresL230 changed the title fix(backfill): timeout-guard Workers AI so a hung summarize can't wedge Sync GitHub fix(backfill): Workers-AI timeout guard for Sync + re-land Docs reader restyle Jul 7, 2026

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/tools/summarize.ts`:
- Around line 137-176: The current timeout logic in summarize() only stops
waiting because Promise.race does not cancel ai.run, so timed-out Workers AI
calls can keep consuming capacity. Update the workersAiSummarizer flow to create
an AbortController, pass its signal through the ai.run(AWORKERS_AI_MODEL, ...)
AiOptions, and abort it when the timeout promise wins; keep the existing
catch/fallback behavior and ensure the timer is still cleared in finally.
🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

  • Push a commit to this branch (recommended)
  • Create a new PR with the fixes

ℹ️ Review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 9b368762-968f-4a7c-aed9-7c6168673869

📥 Commits

Reviewing files that changed from the base of the PR and between b28f5e5 and 1038be7.

📒 Files selected for processing (3)
  • src/tools/backfill.ts
  • src/tools/summarize.ts
  • test/summarize.test.ts

Comment thread src/tools/summarize.ts
Promise.race only stops WAITING for ai.run — the losing call keeps its
inference subrequest alive and keeps consuming capacity. Across a backfill
batch that piles up several abandoned calls against the very capacity that's
already exhausted. Thread an AbortSignal (AiOptions.signal, supported since
workerd early 2026) into ai.run and abort it when the timeout wins, so a
timed-out summary actually cancels. Timeout->null->excerpt fallback unchanged.

Addresses CodeRabbit review on PR #26.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@AndresL230
AndresL230 merged commit d2e6287 into main Jul 7, 2026
2 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.

1 participant