bragai/bRAG-langchain
BRAG Langchain
bRAG-langchain is a framework for building retrieval augmented generation pipelines using LangChain to connect documents with language models. It functions as a vector store orchestrator that manages document indexing and retrieval strategies to improve context accuracy.
The system implements an advanced retrieval pipeline featuring a semantic query router that directs natural language inputs to specific data sources or prompts. It includes a metadata filtering engine that translates natural language queries into structured schemas to narrow search results.
The project covers hybrid search optimization through query expansion and reciprocal rank fusion. It supports multi-vector indexing and the storage of multiple document representations to increase retrieval precision.
Features
- RAG Pipelines - Provides a framework for building retrieval augmented generation pipelines that connect document stores and language models.
- Hybrid Search Systems - Combines multiple retrieval methods and rank fusion to optimize the context provided to language models.
- RAG Frameworks - Provides a specialized framework using LangChain to build and orchestrate retrieval augmented generation pipelines.
- Semantic Routers - Routes natural language inputs to specific data sources or prompts based on semantic similarity.
- Vector Store Orchestrators - Indexes documents and manages retrieval strategies across vector databases to improve context accuracy.
- Metadata Filtering - Implements metadata-filtered retrieval to narrow search results using structured document attributes.
- Natural Language to Metadata Translators - Translates natural language inputs into structured metadata schemas to narrow search results by specific attributes.
- Semantic Query Routing - Directs natural language inputs to specific data sources or prompts based on semantic similarity.
- Advanced Retrieval Techniques - Employs query expansion and diversification techniques to capture a wider set of relevant documents.
- AI Query Generators - Implements the generation of multiple query variations to increase the recall of relevant documents for AI responses.
- Retrieval Re-ranking - Provides reciprocal rank fusion to re-order retrieved context for improved language model generation quality.
- Multi-Stage Retrieval Pipelines - Implements a multi-stage retrieval workflow using query diversification and reciprocal rank fusion.
- Multi-Representation Indexing - Supports storing different document representations, such as summaries or tokens, to improve search precision.
- Query Expansion - Generates multiple prompt variations to retrieve a broader set of context documents from vector databases.
- Search Result Fusion Algorithms - Merges multiple search result lists using reciprocal rank fusion to surface the most relevant context.
- Multi-Vector Indexing - Stores multiple representations of a single document to improve retrieval precision across different query types.
- RAG Component Modularity - Connects document loaders, vector stores, and language models into a modular, linear sequence for context-aware generation.
- RAG Frameworks and Toolkits - Comprehensive tutorial repository for learning RAG implementation.
- Retrieval Augmented Generation - Resources for building custom retrieval-augmented applications.
- RAG and Document Processing - Tutorial series for advanced RAG development.
