Repolode
Submit
netease-youdao

netease-youdao/QAnything

A local knowledge base Q&A system that lets you ask questions over your own files, including PDF, Word, Excel, Markdown and images, with offline deployment.

Open-source alternative to:ChatPDFNotion AI

View on GitHubWebsite
Stars
14,205
Forks
1,351
Open issues
407
Last commit
2 months ago

What is it?

QAnything is a question-answering system for local knowledge bases. You upload files in formats such as PDF, DOCX, PPTX, XLSX, Markdown, email, TXT, images, CSV and web pages, and it answers questions based on their contents. It parses documents, splits them into chunks, and retrieves relevant passages before passing them to a language model.

Retrieval uses a two-stage pipeline: embedding-based search followed by reranking, which the project says keeps accuracy stable as the amount of data grows. Hybrid search combining BM25 and embeddings is also supported. The default setup runs on CPU and is started with Docker Compose, and the components for PDF parsing, OCR, embedding and reranking can be swapped or called independently over HTTP. The LLM is accessed through an OpenAI-compatible interface, so tools such as Ollama can supply local models.

Beyond standard Q&A, it offers a quick-start mode for chatting with a file without creating a knowledge base, a retrieval-only mode that returns search results without calling a model, and custom bots. Version 2.0 also adds an API and online editing of parsed chunks.

Key features

  • 01Parses PDF, Word, PowerPoint, Excel, Markdown, email, images, CSV and web pages
  • 02Two-stage retrieval with embedding search and reranking, plus hybrid BM25 search
  • 03Runs on CPU by default and starts with a single Docker Compose command
  • 04Works with any OpenAI-compatible LLM endpoint, including local models via Ollama
  • 05Supports offline deployment without network access
  • 06Offers quick-start, retrieval-only and custom bot modes, with an HTTP API

Use cases

  • Asking questions over a folder of internal manuals, reports or contracts without uploading them to a cloud service
  • Building an offline question-answering tool for an air-gapped office or lab
  • Searching a mixed set of PDFs, spreadsheets and emails with a single query
  • Prototyping a RAG pipeline with swappable embedding, reranking and OCR components

Quick start

  1. 1

    Pull the qanything repository

  2. 2

    Enter the project root directory and run the startup command from the README

  3. 3

    Open the service and start using it

Things to know

  • Licensed under AGPL-3.0, a copyleft license that requires sharing modified source code when offered as a network service.
  • Commercial use of the bundled LLM is subject to the QwenLM license, as the README notes.
  • Running it requires Docker and a local deployment, so self-hosting takes setup effort.

In collections

Private AI knowledge bases and RAGPrivacy & local-first tools

Topics

#rag#knowledge-base#document-qa#llm#embedding#reranking#offline#docker

FAQ

What is QAnything?
A local knowledge base Q&A system that lets you ask questions over your own files, including PDF, Word, Excel, Markdown and images, with offline deployment. QAnything is a question-answering system for local knowledge bases. You upload files in formats such as PDF, DOCX, PPTX, XLSX, Markdown, email, TXT, images, CSV and web pages, and it answers questions based on their contents. It parses documents, splits them into chunks, and retrieves relevant passages before passing them to a language model.
Is QAnything free to use? What license does it have?
QAnything is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
Is QAnything still maintained?
Updated within 6 months. The last commit to QAnything was on Sep 23, 2026.
What can I use QAnything for?
Asking questions over a folder of internal manuals, reports or contracts without uploading them to a cloud service Building an offline question-answering tool for an air-gapped office or lab Searching a mixed set of PDFs, spreadsheets and emails with a single query Prototyping a RAG pipeline with swappable embedding, reranking and OCR components
How do I install QAnything?
Pull the qanything repository Enter the project root directory and run the startup command from the README Open the service and start using it
Where can I get QAnything?
The source code and installation instructions are on GitHub: https://github.com/netease-youdao/QAnything

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects