hubENTERPRISE
RAG Playground
Interactive RAG PipelineInteractive web application demonstrating the complete Retrieval-Augmented Generation pipeline with visual stage-by-stage exploration. Experiment with document processing, embeddings, vector search, and AI generation.
Architecture Snapshot
- 9-stage RAG pipeline: Ingestion → Preprocessing → Chunking → Embedding → Indexing → Retrieval → Reranking → Generation → Evaluation.
- Multi-provider support: Google Gemini, OpenAI embeddings (768/1536/3072-dim) and GPT/Gemini/Groq LLMs.
- Dual storage: Supabase S3 cloud with local fallback, Qdrant vector DB with HNSW indexing.
- Real-time visual pipeline with React Flow, comprehensive logging, and evaluation metrics.
Next.js 16FastAPIQdrantOpenAIGeminiReact FlowZustandPython 3.11