perf: SQL-based conversation pagination + missing DB indexes

[BillChirico]

Summary

Performance audit of recently merged features — specifically the AI conversations viewer (PR #121) and AI feedback (PR #190). Found three bottlenecks, all fixed here.

---

Issues Found & Fixed

🔴 Critical: 10,000-row in-memory pagination (conversations list)

File: src/api/routes/conversations.js — GET /guilds/:id/conversations

Every page request fetched up to 10,000 message rows into Node heap, grouped them in JavaScript using groupMessagesIntoConversations(), then sliced the result for the requested page. For active servers this means:

• 10k rows × ~200 bytes/row ≈ 2MB per request allocated and GC'd
• O(n) grouping pass over the full dataset for every page 1, 2, 3...
• Response time scales with conversation volume, not page size

Fix: Replaced with a single SQL CTE that uses window functions (LAG + SUM OVER) to identify conversation boundaries and aggregate summaries directly in Postgres. COUNT(*) OVER() provides the total count without a second query. Pagination via LIMIT/OFFSET on summary rows — Node only processes 25 rows (default page size).

🟡 Missing indexes (4 new indexes, migration 004)

Index	Table	Columns	Fixes
idx_ai_feedback_guild_created	ai_feedback	(guild_id, created_at DESC)	getFeedbackTrend() / getRecentFeedback() full guild scan
idx_conversations_content_trgm	conversations	content gin_trgm_ops	ILIKE '%...%' search sequential scan
idx_conversations_guild_created	conversations	(guild_id, created_at DESC)	Default 30-day listing with no channel filter
idx_flagged_messages_guild_message	flagged_messages	(guild_id, message_id)	Conversation detail + flag endpoints

The pg_trgm extension is enabled idempotently (CREATE EXTENSION IF NOT EXISTS).

🟢 Minor: Sequential verification queries in flag POST

File: src/api/routes/conversations.js — POST /guilds/:id/conversations/:conversationId/flag

msgCheck and anchorCheck were independent lookups running sequentially (2× DB round-trips). Changed to Promise.all — 1× round-trip.

---

Test Results

• ✅ 41/41 conversation tests pass (updated 1 test to match new SQL response shape)
• ✅ Pre-existing failures in redis.test.js, triage.coverage.test.js, and ai-feedback.test.js are unrelated to this PR (confirmed present on main before this branch)

---

Notes

• buildConversationSummary() helper removed — its logic is now handled SQL-side
• groupMessagesIntoConversations() is still used by the conversation detail endpoint and its export is tested; no changes there
• The pg_trgm ILIKE index requires the pg_trgm extension to be available (standard on Postgres 12+)

[coderabbitai]

Warning

Rate limit exceeded

@BillChirico has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 0 minutes and 40 seconds before requesting another review.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

📥 Commits

Reviewing files that changed from the base of the PR and between bcf04e2 and 4bffbfb.

📒 Files selected for processing (3)

• migrations/004_performance_indexes.cjs
• src/api/routes/conversations.js
• tests/api/routes/conversations.test.js

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch perf/conversation-query-optimizations

---

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

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[Copilot]

Pull request overview

Improves performance of the conversations viewer and related endpoints by moving conversation pagination/grouping into SQL and adding missing database indexes to support common query patterns.

Changes:

• Replaced Node in-memory grouping/pagination with a Postgres CTE + window-function based conversation summarization and SQL-level pagination.
• Added migration 004 with composite + trigram indexes (and pg_trgm enablement) to eliminate sequential scans and expensive sorts.
• Parallelized independent verification queries in the flagging endpoint.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
src/api/routes/conversations.js	Moves conversation grouping/pagination into SQL and refactors response shaping; parallelizes verification queries for flagging.
tests/api/routes/conversations.test.js	Updates conversation pagination test to match the new SQL summary row shape (total + preview expectations).
migrations/004_performance_indexes.cjs	Adds pg_trgm extension enablement and several targeted indexes for conversations/feedback/flags queries.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[greptile-apps]

Greptile Summary

Excellent performance optimization that addresses three identified bottlenecks from the AI conversations feature. The critical fix replaces 10k-row in-memory pagination with a sophisticated SQL CTE using window functions (LAG, SUM OVER) to group and paginate conversations directly in Postgres, reducing heap allocation from ~2MB to ~5KB per request and changing complexity from O(n) to O(1) for pagination.

Major Changes:

• SQL-based conversation grouping using 3-stage CTE (lag_step → numbered → summaries) with proper 15-minute gap detection
• 4 new composite indexes in migration 004: ai_feedback(guild_id, created_at), conversations GIN/pg_trgm for ILIKE search, conversations(guild_id, created_at), flagged_messages(guild_id, message_id)
• Parallel query optimization in flag endpoint (sequential → Promise.all)
• Removed buildConversationSummary() helper (logic moved to SQL)

Technical Highlights:

• Window function logic is correct: LAG detects gaps, SUM OVER assigns conversation numbers, COUNT(*) OVER() provides total without second query
• Participant parsing uses lastIndexOf(':::') to correctly handle usernames containing the separator
• All user input remains properly parameterized (constant CONVERSATION_GAP_MINUTES * 60 is safe)
• Maintains identical behavior to previous implementation while eliminating memory bottleneck
• Test coverage updated and all 41 tests passing

Confidence Score: 5/5

• This PR is safe to merge with high confidence - well-tested performance optimization with no breaking changes
• Score of 5 reflects: (1) sophisticated but correct SQL window function logic that properly handles conversation grouping, (2) comprehensive test coverage with all 41 tests passing, (3) proper database migration with idempotent operations and down migration, (4) all user input properly parameterized following security best practices, (5) significant performance improvement without changing API behavior, (6) clear documentation and thorough PR description demonstrating deep understanding of the problem
• No files require special attention - all changes are well-implemented and thoroughly tested

Important Files Changed

Filename	Overview
migrations/004_performance_indexes.cjs	Adds 4 well-justified indexes: ai_feedback(guild_id,created_at), conversations GIN/trgm for ILIKE, conversations(guild_id,created_at), flagged_messages(guild_id,message_id). Includes proper up/down migrations and idempotent extension setup.
src/api/routes/conversations.js	Replaces 10k-row in-memory pagination with SQL-based CTE using window functions (LAG/SUM OVER) for conversation grouping. Removes buildConversationSummary helper. Optimizes flag endpoint with Promise.all for parallel verification queries.
tests/api/routes/conversations.test.js	Updates one test to match new SQL response shape (pre-aggregated conversation summary rows instead of individual message rows). All other tests remain unchanged and passing.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    Start[GET /conversations request] --> OldOrNew{Implementation}
    
    OldOrNew -->|Old: PR #121| OldFetch[Fetch up to 10k rows<br/>ORDER BY created_at DESC]
    OldFetch --> OldReverse[Reverse rows to ASC]
    OldReverse --> OldGroup[groupMessagesIntoConversations<br/>in Node.js memory]
    OldGroup --> OldPaginate[Slice array for page]
    OldPaginate --> OldBuild[buildConversationSummary<br/>for each conversation]
    OldBuild --> OldReturn[Return JSON]
    
    OldOrNew -->|New: This PR| NewCTE[Execute SQL CTE with<br/>window functions]
    NewCTE --> NewLag[lag_step: LAG to detect<br/>15-min gaps]
    NewLag --> NewNum[numbered: SUM OVER<br/>to assign conv_num]
    NewNum --> NewSum[summaries: GROUP BY<br/>channel + conv_num]
    NewSum --> NewPage[LIMIT/OFFSET on<br/>summary rows]
    NewPage --> NewMap[Map results in Node<br/>parse participant_pairs]
    NewMap --> NewReturn[Return JSON]
    
    OldReturn --> Impact1[2MB heap per request<br/>O n grouping]
    NewReturn --> Impact2[~5KB heap per request<br/>O 1 pagination]
    
    style OldFetch fill:#ffcccc
    style OldGroup fill:#ffcccc
    style OldPaginate fill:#ffcccc
    style NewCTE fill:#ccffcc
    style NewPage fill:#ccffcc
    style Impact1 fill:#ff9999
    style Impact2 fill:#99ff99

Loading

_{Last reviewed commit: 4bffbfb}