Airweave open‑source retrieval layer
Airweave, an open‑source context retrieval layer, launched to power self‑hosted assistants that search Notion/GitHub/Jira with semantic, keyword and agentic search modes — positioning itself as a free rival to proprietary context layers. The project underlines how startups and teams are re‑architecting RAG with modular retrieval middle layers. (x.com)
Airweave’s public repository shows the stack uses FastAPI for the backend, PostgreSQL for metadata, Vespa for vector storage, Temporal for worker orchestration, Redis for pub/sub, and supports Docker Compose for dev and Kubernetes for production. (github.com) The retrieval pipeline runs semantic and BM25 keyword search in parallel, fuses results with reciprocal rank fusion (RRF), applies recency weighting, and exposes AI reranking to produce a final ranked set for agents. (bestofshowhn.com) Built-in connectors and the docs list first‑class integrations for Notion, GitHub, Jira, Slack, Google Drive and other productivity and developer tools, with the platform describing “40+ connectors” available through the unified connector framework. (docs.airweave.ai) Airweave exposes REST APIs plus Python and TypeScript SDKs, an MCP server for agent tool discovery, and a CLI and quickstart examples in the repo for local self‑hosting and programmatic workflows. (airweave.ai) The project is published under an MIT license and the main GitHub repo shows roughly 6.1k stars, ~733 forks and over 4,600 commits as of the current snapshot. (github.com) Airweave lists itself as a YC X25 company founded in Spring 2025 by Rauf Akdemir and Lennert Jansen and the YC profile indicates a small early team and active hiring. (ycombinator.com) A current limitation flagged in the repo is that the Jira connector primarily indexes core fields (summary/description) and lacks some issue‑level fields used in production workflows, while the docs provide a guide for adding custom connectors if teams need deeper field coverage. (github.com)
