feat: Add compaction detection and context efficiency analysis (closes #69)

[evansenter]

Summary

• Add result_size_bytes column to track message sizes for context burn rate analysis
• Detect compaction events during ingestion (when summaries contain "continued from a previous conversation")
• Add 4 new MCP tools and CLI commands for context efficiency analysis:
  • get_compaction_events - List context reset events
  • get_pre_compaction_events - Show events before a compaction
  • get_large_tool_results - Find bloated tool results consuming context
  • get_session_efficiency - Analyze burn rate and efficiency metrics

Test plan

• All 352 tests pass
• New tests for get_pre_compaction_events (3 tests)
• New tests for calculate_result_size and detect_compaction helpers (14 tests)
• Migration v9 properly adds column with idempotency check

🤖 Generated with Claude Code

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR adds compaction detection and context efficiency analysis capabilities: a new result_size_bytes column for tracking message sizes, detection of compaction events during ingestion, and 4 new MCP tools/CLI commands for analyzing context efficiency patterns.

Issues Found

Critical

None

Important

None

Suggestions

• cli.py - Missing CLI formatter for get_pre_compaction_events. The command exists (line 1109) but unlike _format_compactions, _format_large_results, and _format_efficiency, there's no @_register_formatter for pre-compaction events. It falls back to JSON which is inconsistent with the other new commands that have human-readable formatters.

• queries.py:2241 - Issue feat: Compaction detection and context efficiency analysis #69 specified read_to_edit_ratio and files_read_multiple_times metrics for get_session_efficiency(), but these are not implemented. The current implementation has useful signals but doesn't match the full issue spec. Consider whether these should be deferred to a follow-up or added here.

Verdict

REQUEST_CHANGES - Minor suggestions for consistency and issue spec alignment.

---

Automated review by Claude Code

[evansenter]

Feedback Addressed

Implemented

• [Suggestion] Missing CLI formatter for get_pre_compaction_events - Added _format_pre_compaction formatter for human-readable output
• [Suggestion] Missing metrics from Issue feat: Compaction detection and context efficiency analysis #69 spec - Added read_to_edit_ratio and files_read_multiple_times to get_session_efficiency() output

[claude]

Prompt: evansenter/dotfiles/.../claude-review.md

Code Review

Summary

This PR adds compaction detection and context efficiency analysis capabilities: a new result_size_bytes column for tracking message sizes, detection of compaction events during ingestion (via entry_type='compaction'), and 4 new MCP tools/CLI commands for analyzing context efficiency patterns. The implementation addresses all requirements from Issue #69.

Issues Found

Critical

None

Important

None

Suggestions

None

Previous Feedback Status

The previous review identified two suggestions:

• Missing CLI formatter for get_pre_compaction_events - ✅ Implemented (_format_pre_compaction at cli.py:638-660)
• Missing read_to_edit_ratio and files_read_multiple_times metrics - ✅ Implemented in get_session_efficiency() at queries.py:2339-2340

Notes

• Test coverage is comprehensive: 352 tests pass, with specific tests for the new query functions, helper functions (calculate_result_size, detect_compaction), and storage operations
• get_pre_compaction_events is intentionally excluded from benchmark (cli.py:1348-1350) since it requires specific session_id and timestamp parameters - this is consistent with the existing pattern for parameter-dependent tools
• Migration v9 properly adds the column with idempotency check

Verdict

APPROVE - Code looks good, no issues found. All previous feedback has been addressed.

---

Automated review by Claude Code