v0.2.0

What's New in v0.2.0

Added

• Agent-friendly CLI harness: formats list, formats describe, schema, validate, normalize, diagnose
• Cursor and Claude Code transcript JSONL ingestion
• agent-trace format support for academic trajectory datasets
• Agent-readable diagnosis reports with issue codes, confidence, and recommendations
• Context diff evidence: turn-to-turn token summary, large addition hints, artifact churn hints
• Tool context hotspot diagnosis: TOOL_USE_DOMINATES_CONTEXT, TOP_TOOL_CONTEXT_HOTSPOT
• Agent Skill distribution (.plugin/plugin.json, .claude-plugin/marketplace.json)
• Persistence View: heatmap of content blocks across requests
• Multi-turn dataset examples (SWE-agent, lmcache, OpenHands)
• HTML timeline markers for large additions and high tool-use turns

Changed

• Default analysis now includes content repeat and field repeat analyzers
• Repository URLs updated to yanpgwang/context-profiler

Full Changelog: v0.1.0...v0.2.0

v0.1.0

Initial release of context-profiler.

• Token distribution analysis by role, content type, and tool name
• Interactive HTML report with icicle visualization, timeline, and diff view
• CLI report with Rich formatting
• Support for OpenAI, Anthropic, Langfuse, JSONL, and directory inputs
• Snapshot mode (single request) and session mode (multi-request)