Turn an unfamiliar GitHub repository into an interactive engineering map.
- Live Demo: DevFlow App
- GitHub Repository: GitHub Repository
- Hackathon: WeMakeDevs × Zerops
Understanding an unfamiliar repository requires manually exploring and cross-referencing:
- README files and fragmented documentation
- Package manifests across monorepo packages and subdirectories
- Directory structures, entry points, and routing modules
- Core services, API endpoints, and middleware layers
- External dependencies and architecture boundaries
This manual overhead becomes overwhelming when onboarding to large codebases, evaluating open-source libraries, or exploring unfamiliar tech stacks with sparse documentation.
DevFlow automates repository comprehension by turning raw source code into structured engineering intelligence and interactive architectural maps.
GitHub Repository
↓
Analysis Job
↓
Repository Inspection
↓
Repository Intelligence
↓
Architecture + API Analysis
↓
FalkorDB Knowledge Graph
↓
Interactive Repository Report
↓
Ask DevFlow
- Deterministic Extraction: DevFlow clones and statically inspects the codebase to extract exact facts about languages, frameworks, dependencies, package managers, and API routes without relying on ungrounded AI guesses.
- Knowledge Graph Projection: Extracted facts and structural relationships are projected into a FalkorDB graph database.
- Graph-Grounded Q&A ("Ask DevFlow"): When users ask questions about the repository, DevFlow queries the FalkorDB knowledge graph for exact factual context before passing it to Gemini, ensuring precise, grounded answers rather than generic hallucinations.
- Repository Intelligence: Comprehensive breakdown of repository purpose, structure, and classification.
- Language & Framework Detection: Extension mapping and framework identification with confidence scoring.
- Dependency Intelligence: Multi-manifest manifest parsing for Node.js, Python, Rust, Go, Java, and more.
- Architecture Intelligence: Workspace layout analysis, decoupled client/server inspection, and containerization detection.
- API Surface Intelligence: Automated route detection across controllers, handlers, and API endpoints.
- Repository Health Signals: Evaluation of CI workflows, Docker setups, and standard project configuration files.
- Interactive Architecture Visualization: Dynamic visual maps powered by D3.
- FalkorDB Knowledge Graph: Robust graph-based entity and relationship storage.
- Graph-Grounded AI Questions: Context-aware querying via Gemini backed by real graph facts.
- Asynchronous Repository Analysis: Robust background worker architecture managing job state transitions.
- Analysis Progress Tracking: Real-time state updates across queued, running, completed, and failed stages.
flowchart TD
A[GitHub Repository] --> B[Analysis API]
B --> C[Supabase Job]
C --> D[Background Worker]
D --> E[Repository Inspection]
E --> F[Repository Intelligence]
F --> G[Architecture Intelligence]
F --> H[API Surface Intelligence]
F --> I[Dependency Intelligence]
F --> J[Supabase]
F --> K[FalkorDB]
K --> L[Knowledge Graph]
L --> M[Repository Report]
L --> N[Ask DevFlow]
N --> O[Gemini]
# Install workspace dependencies with pnpm
pnpm install --frozen-lockfile
# Start full-stack development server (API + Web on Port 3000)
pnpm run dev
# Start background analysis worker
pnpm run worker
# Run typecheck across all workspace packages
pnpm run lint
# Run unit tests
pnpm test
# Build production bundle
pnpm run buildDevFlow is fully configured for zero-downtime, multi-service deployment on Zerops using zerops.yaml:
frontend: Static SPA deployment powered by Vite.api: Node.js 22 Express API server handling analysis endpoints and health checks (GET /health).worker: Background Node.js 22 job processing daemon.