Flexible GraphRAG v0.8.0 adds Rust-based CocoIndex pipeline and Visual Langflow flows

Products·launched

Custom RAG pipeline from scratch without LangChain/LlamaIndex

A production-style Retrieval-Augmented Generation system built from scratch, implementing hybrid retrieval (FAISS + BM25), reciprocal rank fusion, cross-encoder reranking, intelligent PDF ingestion with semantic chunking, query analysis, Mermaid diagram retrieval, and modular context/prompt builders. Uses Groq for response generation. Designed for developers who want to understand and control the internals of RAG.

A production-style Retrieval-Augmented Generation system built from scratch, implementing hybrid retrieval (FAISS + BM25), reciprocal rank fusion, cross-encoder reranking, intelligent PDF ingestion with semantic chunking, query analysis, Mermaid diagram retrieval, and modular context/prompt builders. Uses Groq for response generation. Designed for developers who want to understand and control the internals of RAG.

Releases·Tool Release

Developer releases Agentic GraphRAG Blueprint for knowledge-graph QA over documents

A developer shared a first-version reference architecture for question answering over large document collections, combining knowledge graphs with vector search instead of plain chunk retrieval. The blueprint targets cross-document synthesis and cost-efficient incremental ingestion, making it relevant to teams building RAG pipelines.

A developer shared a first-version reference architecture for question answering over large document collections, combining knowledge graphs with vector search instead of plain chunk retrieval. The blueprint targets cross-document synthesis and cost-efficient incremental ingestion, making it relevant to teams building RAG pipelines.

A developer on Reddit asks for a GraphRAG framework that is production-ready, customizable, and includes batteries, noting that many libraries come and go or carry too much infrastructure. The request emphasizes support for indexing, storing, retrieving, and API integration for agents, with evals and scalability to thousands of documents and hundreds of thousands of relationships.

Cairn is an open-source GraphRAG domain knowledge layer that continuously distills scattered skill documents into a queryable, evolvable, human-editable two-layer knowledge graph. It versions, hot-swaps, and rolls back the graph via a GitOps pipeline, and serves it to AI agents over MCP. The project ships as a three-part suite: DK Build, MCP Service, and Viewer.

Products·launched

PipesHub - open-source enterprise RAG infrastructure with connectors and permissions

PipesHub is an open-source context layer for building AI applications on top of company data. It handles the unglamorous infrastructure of enterprise RAG: connectors to sources like Google Drive, Slack, Jira, Confluence, SharePoint, S3, email and databases, permission preservation, incremental syncing of changing documents, and citations that point back to source documents. It targets developers building internal AI apps who would otherwise hand-roll this pipeline.

PipesHub is an open-source context layer for building AI applications on top of company data. It handles the unglamorous infrastructure of enterprise RAG: connectors to sources like Google Drive, Slack, Jira, Confluence, SharePoint, S3, email and databases, permission preservation, incremental syncing of changing documents, and citations that point back to source documents. It targets developers building internal AI apps who would otherwise hand-roll this pipeline.