CodexA

Appearance

Roadmap ​

CodexA development roadmap. All phases through v0.5.0 are complete.

v0.5.0 — All Phases Complete ​

CodexA v0.5.0 ships with all planned features implemented across 42 development phases.

Phases 33–37 — Search Dominance ​

  • Phase 33: JSONL streaming, scored output, snippet control, exclude/no-ignore flags
  • Phase 34: Incremental indexing (--add, --inspect), model-consistency guard, Ctrl+C safety
  • Phase 35: Tantivy full-text engine (Rust-native, cfg-gated)
  • Phase 36: MCP Server v2 with cursor-based pagination, --serve shorthand, Claude Desktop auto-config
  • Phase 37: --hybrid/--sem shorthands, .codexaignore auto-creation, PyInstaller single-binary

Phase 32 — Rust Search Engine Core ​

  • Native codexa-core Rust crate via PyO3 — HNSW, BM25, tree-sitter chunker (10 languages), parallel scanner, optional ONNX and Tantivy

Phase 31 — RAG Pipeline ​

  • 4-stage Retrieve → Deduplicate → Re-rank → Assemble pipeline
  • Configurable strategies, cross-encoder re-ranking, token-aware budgets, source citations

Phases 38–42 — Platform Expansion ​

  • Phase 38: Incremental Embedding Models & Model Hub — --switch-model, model verification, multi-model index, benchmark memory metrics
  • Phase 39: Pre-built Wheels & Platform Distribution — manylinux/macOS/Windows wheels, Scoop, Chocolatey, PyInstaller binaries
  • Phase 40: Code Editor Compatibility — Zed, JetBrains, Neovim, Vim, Sublime Text, Emacs, Helix, Eclipse, Cursor/Windsurf
  • Phase 41: Multi-Agent Orchestration — SessionManager, shared discovery, semantic diff, RAG code generation, bridge session endpoints
  • Phase 42: Cross-Language Intelligence — FFI pattern detection, polyglot dependency graphs, language-aware search boosting, universal call graph

Earlier Phases (1–30) ​

  • CLI framework, repository indexing, semantic search, code parsing, context engine, AI features
  • Platform evolution, AI coding assistant platform, external AI cooperation
  • Multi-repo workspaces, multi-language parsing, platform enhancements
  • Open source readiness, web interface, CI/CD pipeline, advanced AI workflows
  • VS Code extension, priority features, power features, UI/UX polish
  • Model profiles, persistent index, LLM caching, mypy strict, deep coverage
  • AI agent tooling protocol, developer workflow intelligence, code quality metrics
  • Self-improving development loop, competitive feature parity & distribution

What's Next? ​

All planned phases are complete. CodexA is now in community-driven development mode.

If you have ideas for improvements, new features, or integrations, please open an issue on GitHub and tell us what you'd like to see.