Initial commit: Prompt Performance Analytics Dashboard
- Agent 1: Prompt Analyzer (FastAPI + Anthropic Claude)
- Scores prompts on 5 dimensions (clarity, token efficiency,
goal alignment, structure, vagueness)
- Generates optimized rewrites with token savings
- Per-project Context Store for pattern learning
- Agent 2: Analytics Reporter
- SQLite storage with async writes
- Human vs Agent mode tracking
- Rewrite adoption metrics
- MCP Server
- analyze_prompt and get_analysis_history tools
- Claude Desktop integration via claude_desktop_config.json
- Dashboard UI
- Live KPIs, quality trend chart, common mistakes donut chart
- Interaction feed with per-project filtering
- CLI test script (test_agent.py) for direct API testing
704cccf729