Repository: github.com/funkybooboo/insighthub

A flexible platform for building, configuring, and comparing Retrieval-Augmented Generation (RAG) systems across any dataset. Enables rapid experimentation with different RAG architectures and configurations.

Key Features

  • Modular RAG Pipeline: Configurable components for embedding, retrieval, and generation
  • Multi-Model Support: Integration with multiple embedding models and LLMs
  • Vector Database Flexibility: Support for Pinecone, Qdrant, and other vector stores
  • Comparative Analysis: Side-by-side comparison of different RAG configurations
  • Dataset-Agnostic: Works with any dataset format for maximum flexibility

Technology Stack

  • Language: Python
  • Vector Databases: Pinecone, Qdrant
  • LLM Integration: OpenAI API, Hugging Face models, LangChain
  • Embeddings: Sentence transformers, OpenAI embeddings
  • Framework: Custom RAG pipeline architecture

Use Cases

  • Research and experimentation with RAG architectures
  • Benchmarking different retrieval and generation strategies
  • Building domain-specific question-answering systems
  • Prototyping AI-powered knowledge bases