feat(conversation-memory): reduce token usage in summarization

[baleen37]

Summary

• Increase chunk size from 8 to 32 exchanges (~75% fewer API calls)
• Skip agent/subagent conversations (files starting with agent-)
• Add characterization tests for the new behavior

Test plan

• Build passes: npm run build
• Tests pass: npm test (9 tests)
• Code changes verified in dist output

Expected impact

• 1765 exchanges: 222 API calls → 56 API calls (~75% reduction)
• 400 exchanges: 51 API calls → 13 API calls (~75% reduction)
• Agent/subagent conversations are now skipped during indexing

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Agent-generated files are now excluded from conversation indexing and sync operations.

• Performance Improvements

  • Increased conversation chunk size from 8 to 32 exchanges, reducing API calls for large conversations.

• Tests

  • Added test coverage for file filtering and conversation chunking logic.

[coderabbitai]

Caution

Review failed

The pull request is closed.

📝 Walkthrough

Walkthrough

This PR introduces filtering to exclude agent-generated files (prefixed with "agent-") from indexing and sync operations, increases the hierarchical summarization chunk size from 8 to 32 exchanges to optimize API calls, adds comprehensive test suites for filtering and chunking logic, and sets up a test script in package.json.

Changes

Cohort / File(s)	Summary
Test Infrastructure plugins/conversation-memory/package.json	Added "test" script using bun test for running test suites.
Indexer Logic plugins/conversation-memory/src/core/indexer.ts, plugins/conversation-memory/src/core/indexer.test.ts	Implemented filtering to exclude files starting with "agent-" in indexConversations, indexSession, and indexUnprocessed. Added test suite validating filter behavior across multiple scenarios.
Summarizer Logic plugins/conversation-memory/src/core/summarizer.ts, plugins/conversation-memory/src/core/summarizer.test.ts	Increased chunk size from 8 to 32 exchanges in hierarchical summarization to reduce API calls. Added test suite validating chunking behavior with various input sizes.
Sync Logic plugins/conversation-memory/src/core/sync.ts	Updated file collection logic to exclude files starting with "agent-" when listing JSONL files.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~15 minutes

Suggested labels

enhancement

Poem

🐰 A bunny hops through code so neat,
Filters out those agent treats,
Chunks grow large, from eight to thirty-two,
Tests ensure it all rings true! 🎉

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/reduce-converstation

---

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.}