#3636th of 67 in RAG, knowledge and searchFor data and documents, documents and ocr

knowhere

Document parsing and retrieval backend that builds navigable memory for agents

Stars
4.1k
License
Apache-2.0
Last commit
Nov 2026
Last release
Nov 2026
Language
Python
Stars, last 28 weeks+4,023
2 May 2026: 38 stars9 May 2026: 39 stars16 May 2026: 244 stars23 May 2026: 458 stars30 May 2026: 761 stars6 Jun 2026: 1,031 stars13 Jun 2026: 1,345 stars20 Jun 2026: 1,669 stars27 Jun 2026: 1,785 stars4 Jul 2026: 1,836 stars11 Jul 2026: 1,889 stars18 Jul 2026: 1,909 stars25 Jul 2026: 2,001 stars1 Aug 2026: 2,136 stars8 Aug 2026: 2,287 stars15 Aug 2026: 2,430 stars22 Aug 2026: 2,591 stars29 Aug 2026: 2,736 stars5 Sept 2026: 2,891 stars12 Sept 2026: 3,157 stars19 Sept 2026: 3,337 stars26 Sept 2026: 3,458 stars3 Oct 2026: 3,573 stars10 Oct 2026: 3,626 stars17 Oct 2026: 3,709 stars24 Oct 2026: 3,798 stars31 Oct 2026: 3,888 stars7 Nov 2026: 3,966 stars14 Nov 2026: 4,061 stars

38 on 2 May 20264,061 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Knowhere ingests PDFs, slides, Office files, images and audio, then builds hierarchy-aware chunks, a cross-document entity graph and page-grounded citations. PDF and PowerPoint go through vision-model parsing, other formats are parsed natively, and agents query the result through built-in retrieval, MCP or a REST tools endpoint. It is a Python API plus worker that needs Postgres, Redis and S3-compatible storage.

Who it is for: Engineers building agentic RAG over messy PDFs, slides and office documents

Strengths

  • Progressive ingestion makes documents searchable before background enrichment finishes
  • Citations keep source document, section path, page number and rendered page image
  • External agents can call the same corpus tools via MCP or POST /v2/retrieval/tools/invoke
  • Model-agnostic: DeepSeek, OpenAI, Qwen, GLM or any compatible provider

Weaknesses

  • Setup needs Postgres, Redis, S3-compatible storage and several external API keys
  • PDF and PowerPoint parsing depends on a vision-capable model provider
  • Anonymous PostHog telemetry is on by default in self-hosted installs
  • Benchmark figures come from an internal evaluation, not independent runs

What it needs

  • Docker + Compose
  • Needs PostgreSQL, Redis, S3-compatible storage, LLM provider API key, Vision-capable model provider, VERCEL_KEY for embeddings, uv
  • Models: DeepSeek, OpenAI, Qwen, GLM, Volcengine
  • port 5005

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in RAG, knowledge and search

See all 67
Also in RAG, knowledge and search
RankProjectScore
1Open NotebookSelf-hosted NotebookLM alternative for research notebooks, chat, and podcasts42.1k starsMIT92 out of 100
2LightRAGGraph-based RAG framework with a REST server and WebUI40.8k starsMIT92 out of 100
3RAGFlowRAG engine with document parsing, agentic retrieval and citations Live demo ↗ (opens in a new tab)94.4k starsApache-2.090 out of 100
#4deeptutorTutoring workspace with knowledge bases, guided learning, memory and quizzes60.2k starsApache-2.090 out of 100
#5MaxKBEnterprise knowledge-base agent platform with RAG, workflows and MCP tools23.2k starsGPL-3.089 out of 100