Workflow Orchestrator is a real-time collaborative workflow canvas built for teams that plan and execute projects visually. The canvas is powered by React Flow and Socket.io, so multiple people can map tasks, draw dependencies, and watch changes land live — secured end to end with JWT-based RBAC and email OTP authentication.
The most interesting part is the AI layer. I designed a Python RAG pipeline using SentenceTransformers, ChromaDB, and Llama 3.3 (via the Groq API) that ingests complex requirements PDFs, vectorizes them, and auto-parses them into structured execution plans — turning a 40-page spec into an actionable task graph in seconds.
To close the loop with developer workflows, I established bidirectional GitHub Issue sync via Octokit and Webhooks, with custom MongoDB triggers that dynamically manage developer availability and project statuses. Planning on the canvas and execution on GitHub stay perfectly in sync, in both directions.
- Real-time multiplayer canvas with React Flow + Socket.io
- RAG pipeline: PDF → vectors → structured execution plans
- Bidirectional GitHub Issue sync via Octokit/Webhooks
- JWT-based RBAC with email OTP authentication