Skip to content

ROADMAP.md

Latest commit

 

History

History
48 lines (33 loc) · 2.58 KB

ROADMAP.md

File metadata and controls

48 lines (33 loc) · 2.58 KB
Raw

codei — Strategic Roadmap 🚀

This roadmap outlines the long-term vision to scale codei into a professional AI context retrieval engine.

🟢 Phase 1: Foundation & Language Expansion (Current focus)

  • Multi-language Support: Integrate tree-sitter to support Python, Go, Rust, and Java (currently limited to TypeScript).
  • Incremental Build Optimization: Improve the hashing mechanism to detect file changes even faster and only rebuild affected sub-trees.
  • Documentation Overhaul: Complete multilingual READMEs and add detailed API usage examples.
  • Telemetry (Local only): Add stats for token reduction and cost-saving metrics per query.

🟡 Phase 2: Ecosystem & Better Integration (The Connectivity Phase)

  • Native MCP (Model Context Protocol) Support: Build an official MCP server for codei to enable "plug-and-play" integration with Claude Desktop, Cursor, and Windsurf without needing custom scripts.
  • Official VS Code Extension: A sidebar to visualize the index tree and a status bar to monitor the local server.
  • CI/CD Integration: Official GitHub Action and GitLab Runner scripts for auto-indexing on merge.

🟠 Phase 3: Hybrid Search & Local Intelligence (The Reasoning Phase)

  • Hybrid Retrieval (RAG + Reasoning):
    • Implement a small, local vector store (e.g., SQLite with sqlite-vec or FAISS) for keyword-based jumping.
    • Combine vector search with our existing Hierarchical Reasoning Tree for 100% accuracy.
  • Local LLM First: Optimization for Small Language Models (SLMs) like phi-4 or deepseek-v3 running locally via Ollama to reduce costs to zero.
  • Context Window Packing: Intelligently pack multiple related files into a single context snippet to avoid multiple queries.

🔴 Phase 4: Enterprise & Scale (The Maturity Phase)

  • Monorepo Scalability: Optimize memory usage for projects with 10k+ files (using disk-based tree nodes).
  • Web Dashboard: A local web UI to:
    • Explore the generated tree visually.
    • Manage global settings and API keys.
    • Review query history and token costs.
  • Code Graph Analysis: Go beyond imports — detect cross-file patterns (Shared state, Event emitters) that text-based search misses.

🛠 Active Technical Challenges to Solve:

  1. Circular Dependencies: Handling infinite loops in tree traversal.
  2. Context Fragmentation: Ensuring that splitting code into symbols doesn't lose the "bigger picture" of the logic.
  3. Private LLMs: Ensuring sensitive code remains on-premise for enterprise clients.