Fluxy AI is an advanced AI Research Copilot built using LangGraph, LangChain, RAG pipelines, and hybrid retrieval systems. It combines real-time web search with document-based contextual retrieval to deliver grounded, reliable, and production-ready AI responses.
The system intelligently routes user queries between vector search and live internet search, merges contextual information, generates answers, critiques its own responses, and retries low-quality outputs automatically.
- 🔐 Authentication & session management
- 💬 Real-time AI chat with streaming responses
- 📄 PDF upload and document understanding
- 🧠 RAG (Retrieval-Augmented Generation)
- 🌐 Tavily-powered live web search
- 🔀 Hybrid retrieval (Web + Vector DB)
- 🤖 LangGraph-based agent workflows
- 🧾 Source citations & grounded answers
- ♻️ Self-critique and retry loop for hallucination reduction
- 🧠 Conversational memory support
- 📊 Workflow observability using LangSmith
- 🐳 Docker-ready scalable architecture
- Node.js
- Express.js
- LangChain
- LangGraph
- OpenAI / Groq
- Pinecone / ChromaDB
- Tavily API
- React
- Tailwind CSS
- TanStack Query
- Markdown Rendering
- Streaming UI
User Query ↓ LangGraph Workflow ↓ Router Node → decides between:
- Web Search
- Vector Search
- Hybrid Retrieval
↓ Context Merge ↓ LLM Generation ↓ Critic Node ↓ Retry Loop (if needed) ↓ Final Grounded Response
Fluxy AI is designed to demonstrate:
- AI system design
- Agent orchestration
- Stateful workflows
- Retrieval quality optimization
- Hallucination reduction
- Production-level backend architecture
- Scalable AI application development
- Multi-agent collaboration
- Voice interaction
- Advanced memory systems
- Workspace/project-based knowledge bases
- Team collaboration
- Code execution sandbox
- Fine-grained observability dashboard
Most AI projects stop at “chat with PDF.” Fluxy AI focuses on building a complete AI engineering system with:
- Hybrid retrieval
- Conditional workflows
- Self-evaluation
- Reliability pipelines
- Scalable architecture
This project reflects real-world AI infrastructure and production thinking.