Implement real AI summarize callbacks with smart tone mode

[alex-fedotyev]

Summary

Replace the April Fools local summary generator with real AI summarize callbacks for event, pattern, and trace contexts, while enforcing token-aware context compaction and robust provider error handling.

• Added backend summarize pipeline at POST /ai/summarize (reusing existing AI provider/model setup from chart assistant):
  • strict typed validation per summary kind (event | pattern | trace)
  • bounded payload compaction before prompting
  • concise, action-oriented prompt templates with trace-specific prioritization (errors + critical path)
  • provider/API error handling consistent with existing AI routes
• Migrated summary UI from local phrase templates to backend callbacks:
  • event side panel summary
  • pattern side panel summary
  • trace tab summary (new button)
• Implemented smart-mode style behavior exactly as requested:
  • default prompt/tone always when ?smart=true is absent
  • style selector only appears when ?smart=true
  • style choice persists in localStorage for subsequent summaries in smart mode
• Removed obsolete April Fools generator files and wiring.
• Addressed CI Knip findings by removing unused exports.

Screenshots or video

Before	After

How to test locally or on Vercel

1. Configure AI env vars (same as chart assistant), start app/api.
2. Open an event side panel and click Summarize; confirm AI summary returns.
3. Open a pattern side panel and click Summarize; confirm summary uses pattern sample/stat context.
4. Open trace tab and click Summarize; confirm summary includes error and critical-path context.
5. Without ?smart=true, verify no style selector and outgoing tone is default.
6. With ?smart=true, select Detective Noir, summarize, refresh, and verify style persistence.

References

• Linear Issue: HDX-3992
• Related PRs: Incremental UI updates #2023

Linear Issue: HDX-3992

  

[changeset-bot]

⚠️ No Changeset found

Latest commit: a8f30c2

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
hyperdx-oss	Ready	Preview, Comment	Apr 11, 2026 2:15am

[github-actions]

E2E Test Results

✅ All tests passed • 129 passed • 3 skipped • 1071s

Status	Count
✅ Passed	129
❌ Failed	0
⚠️ Flaky	4
⏭️ Skipped	3

Tests ran across 4 shards in parallel.

View full report →

[github-actions]

🔴 Tier 4 — Critical

Touches auth, data models, config, tasks, OTel pipeline, ClickHouse, or CI/CD.

Why this tier:

• Large diff: 2699 lines changed (threshold: 1000)

Additional context: touches API routes or data models

Review process: Deep review from a domain expert. Synchronous walkthrough may be required.
SLA: Schedule synchronous review within 2 business days.

Stats

• Files changed: 21
• Lines changed: 2699 (+ 349 in test files, excluded from tier calculation)
• Branch: cursor/ai-true-summarize-callbacks-e018
• Author: alex-fedotyev

To override this classification, remove the review/tier-4 label and apply a different review/tier-* label. Manual overrides are preserved on subsequent pushes.

[github-actions]

PR Review

• ⚠️ computeCriticalPath DFS has no cycle detection (requestBuilders.ts) → The memoization only prevents re-entering a key after it's been set, but memo.set happens after all children are visited. Cycles like A→B→A will cause infinite recursion and a browser tab crash. Add a visiting set and skip nodes already in the active call stack.

• ⚠️ No rate limiting on /ai/summarize → A rate limiter utility already exists (packages/api/src/utils/rateLimiter.ts) but isn't applied here (or on /assistant). Authenticated users can spam this endpoint to amplify AI provider costs. Consider applying the existing rateLimiter to both AI routes.

• ⚠️ AISummarizeTraceButton always fetches trace rows (AISummarizeTraceButton.tsx:62-80) → enabled: true causes useEventsAroundFocus to fire on every trace panel render, not just when the user clicks summarize. This duplicates queries already made by DBTraceWaterfallChartContainer. Change to enabled: isOpen || isGenerating or derive rows lazily.

• ℹ️ Redundant alias (ai.ts:26) → const summaryBodySchema = AISummaryRequestSchema; is unused indirection — use AISummaryRequestSchema directly in validateRequest.