codeatlas

Roadmap

CodeAtlas is being developed in five phases. Each phase builds on the previous one.

Phase 1: Core Ingestion & Basic Graph :white_check_mark:

Status: Complete

• GitHub repository cloning (shallow, public repos)
• ZIP file upload and extraction
• Tree-sitter parsing for Python, JavaScript, TypeScript
• Extraction of functions, classes, imports, inheritance
• NetworkX graph construction
• Cytoscape.js interactive visualization
• Landing page with URL input and drag-and-drop upload
• Node detail panel (file path, line number, type)
• Color-coded nodes by type (module, class, function, external)
• Backend test suite (9 tests passing)

Phase 2: Advanced Visuals :white_check_mark:

Status: Complete

• Java language parser (classes, interfaces, methods, constructors, imports)
• Function call extraction across all parsers (Python, JS/TS, Java)
• Call chain analysis — trace function invocations across files (BFS traversal)
• Search by name, type, or file path (debounced, with highlighting)
• Dependency scope filtering (all / internal / external)
• Graph export to PNG (client-side) and JSON (server-side download)
• Edge weight visualization
• Background task processing with progress polling
• Demo mode with bundled sample project
• Dashboard split-view layout (graph canvas + collapsible detail panel)
• Node detail modal with focus management
• Backend test suite expanded to 21 tests

Phase 3: History Integration :white_check_mark:

Status: Complete

• Full Git history analysis (opt-in full clone with sampled snapshots)
• Time slider to view graph at any commit (with play/pause animation)
• Churn metrics — lines changed per module over time
• Change highlighting — added/removed/modified nodes between commits
• Contributor visualization — who works on what (panel + graph highlighting)
• Graph diff between any two commits
• Animated graph morphing for history transitions
• JSON import — re-upload exported graphs to rebuild

Phase 4: RAG Agent :white_check_mark:

Status: Complete

• Vector embeddings of graph data and commit messages (FAISS + sentence-transformers)
• Tool-augmented LLM agent with agentic loop (search, inspect, trace calls, query history)
• Multi-provider LLM support (Anthropic, OpenAI, Google Gemini, xAI Grok, Ollama)
• Chat-like query interface in the dashboard (tabbed panel with Details/Ask AI)
• Response types: text explanations, highlighted subgraphs, code snippets
• Confidence scoring, follow-up suggestions, and local-only fallback mode
• Conversation memory with bounded history
• Provider-agnostic tool format conversion (Anthropic-style canonical → OpenAI/Gemini native)

Phase 5: Polish & Extensibility :white_check_mark:

Status: Complete

• OAuth authentication (GitHub + Google) with JWT tokens
• User API key management — encrypted storage per provider (Fernet)
• User model preferences — choose preferred LLM provider and model
• Atlas history tracking — previously generated graphs per user
• Plugin system for custom language parsers
• Real-time collaboration via WebSockets (presence indicators)
• Shareable graph links
• Docker containerization (multi-stage single-container build)
• Cloud deployment support (Railway, Fly.io)
• API rate limiting (SlowAPI middleware)
• 5 color themes (Dark, Light, Neon, Sunset, High Contrast)
• Resizable sidebar panel with drag handle
• Professional graph styling — connection-based node sizing, directory clustering, refined edge rendering
• Adaptive fCoSE layout tuned to graph size (small/medium/large)
• Graph legend with node types and edge types
• Neo4j integration for persistent graph storage (optional)

Remaining

• Comprehensive frontend testing (Jest + React Testing Library)
• End-to-end tests (Cypress)
• Full accessibility audit (keyboard navigation, screen reader support)
• Performance profiling for graphs with 10,000+ nodes