[ROADMAP] Phase 1 — Solidify (Pure Python, no Rust yet)

[Wolfvin]

Goal

Make CodeLens the best pure-Python code intelligence tool for AI agents before adding Rust complexity. These changes are high-impact, low-risk, and can start now.

Checklist

• [ARCH] Replace flat registry with true graph data model (nodes + edges) #8 — True graph data model (nodes + edges) — foundation for everything
• [FEATURE] get_architecture — single-call codebase overview for agents #19 — get_architecture one-call codebase overview
• [PERF] Token-efficient output: structured compact format for all MCP tools #17 — Token-efficient compact output + pagination
• [FEATURE] Hybrid type resolution: import-aware call edge refinement #13 — Hybrid type resolution (import-aware call edges)
• [FEATURE] Git-diff aware incremental re-index (replace file-watcher polling) #14 — Git-diff aware incremental re-index
• Fix vuln DB staleness (OSV.dev API, update scheduler)

Definition of Done

• Agent can orient in a new codebase in 1 tool call (get_architecture)
• 5 structural queries cost <5k tokens total
• Full scan on 500-file repo under 1 second
• Call graph is import-aware (not just same-file)

Why this before Rust

Architecture fixes (#8, #13) will themselves improve speed significantly — RAM-first pipeline + bulk SQLite inserts can recover 5-10x even in pure Python. Profile after Phase 1 before deciding what to port to Rust.