Skip to content

🚀 Prepare Production Release (Initiated: 2026-07-04) - #511

Merged
annachayn merged 12 commits into
mainfrom
develop
Jul 6, 2026
Merged

annachayn merged 12 commits into
mainfrom
develop

Conversation

@github-actions

@github-actions github-actions Bot commented Jul 4, 2026 •

Copy link
Copy Markdown
Contributor

This Pull Request includes changes from develop to be released into main.
Important: Merging this PR will trigger an automatic deployment to production.


📝 Summary of Changes (from develop to main)

(Last updated: Sat, 04 Jul 2026 14:42:57 GMT)


This PR is automatically managed by the manage-release-pr.yml workflow.

…pletes (#510)

The streaming fix was validated on a test branch but did not make it into the
merge to develop (#508), so develop still has the non-streaming .create() with a
2-minute timeout — which hangs facebook/instagram/tiktok. Step 1
(validateDocuments) calls Anthropic with the web_search server tool
(max_uses: 10), a long tool-heavy request that can take ~13 min on large
platforms. Non-streaming, it holds one HTTP connection open for the whole
server-side search loop and trips the request timeout; the SDK retries timeouts
and retryWithDelay wraps that again, so it thrashes to the job cap instead of
completing. Switch callAnthropic to .stream().finalMessage() and drop maxRetries
to 1. Verified end-to-end on facebook (run completed, quality check valid,
PR opened). Leaves the JSON-unwrap logic from #508 intact.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@vercel

vercel Bot commented Jul 4, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated (UTC)
tools Ready Ready Preview, Comment Jul 4, 2026 2:43pm

Request Review

…513)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…515)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…509)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…un real comparison (#516)

Two root causes behind runs that hung on 'Step 1: validating current documents'
and then reported 'no changes needed':

1. Step 1 used the AGENTIC web_search_20260209 (max_uses: 10), which runs a
   server-side code-execution/dynamic-filtering loop per use — highly variable,
   ~13 min on large platforms. Switch to the classic web_search_20250305
   (~2-3s per search, no code-exec step); keep max_uses: 10 for thorough
   discovery. The latency came from the agentic version, not the search count.

2. The orchestrator short-circuited to completed_no_changes based on Step 1's
   LLM+web_search verdict. When the search underperformed, the model
   rubber-stamped every document 'valid', so the pipeline skipped the real work
   and reported 'no changes' even when policies had changed. Remove the
   short-circuit: always scrape + abstract + compare. Change detection is done
   authoritatively in Step 5 (comparePlatformPolicies) on real document content;
   Step 1 is now only for discovering new/removed documents.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…520)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…517)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…#518)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…#523)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
…521)

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
@annachayn
annachayn merged commit a787359 into main Jul 6, 2026
14 of 15 checks passed

This branch was successfully deployed

1 active deployment
Preview — 694912fa Deployed Jul 4, 2026 by vercel[bot]
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