QAnything General Introduction
QAnything (Question and Answer based on Anything) is a local knowledge base Q&A system launched by NetEase, which supports all kinds of file formats and databases and can be installed and used offline. It can handle PDF, Word, PPT, XLS and other formats of documents, support for cross-language Q&A, and provide large data volume Q&A support, with high performance, user-friendly, multi-knowledge base Q&A capabilities and data security features.
The system is based on the self-developed RAG (Retrieval Augmented Generation) engine, providing efficient and accurate Q&A services.QAnything is suitable for a variety of scenarios, such as internal document management, legal counseling, and governmental services, to help enterprises improve the efficiency of information acquisition and decision-making.
Function List
- Support a variety of file formats: PDF, Word, PPT, Excel, Markdown, TXT, pictures, etc.
- Local deployment: no need for an Internet connection to use, ensuring data security
- Efficient retrieval: based on RAG engine, providing high accuracy semantic retrieval
- Flexible Workflow: Automating Tasks with Agents
- Content generation: Generate complete outlines and article content based on references
- Scenario customization: model and search optimization based on enterprise needs
QAnything Help
System Requirements: Linux with at least 4GB of GPU memory, Windows systems require WSL subsystems
How to install: clone via git and run the startup scripts
How to use: Q&A can be operated via web front-end or API interface
FAQ: Provides answers to frequently asked questions
Technical Support: Provides community support and developer email consulting services
Installation process
- Download QAnything: AccessGitHubpage to download the latest version of QAnything.
- environmental preparation: Ensure that Docker and Docker Compose are installed on your system.
- Pulling the code base: Execute in the terminal
git clone https://github.com/netease-youdao/QAnything.git
Command. - Go to the project directory: Implementation
cd QAnything
Go to the project root directory. - Starting services: Implementation
docker-compose up -d
command to start the QAnything service.
Usage Process
- Uploading files: Upload files to be parsed through QAnything's interface, which supports PDF, Word, PPT, Excel and many other formats.
- Search Q&A: Enter a question in the search box and QAnything will retrieve and generate an answer based on the content of the uploaded file.
- View Results: The system displays relevant answers and references that users can click on to view details.
- Content generation: Users can choose to generate outlines or articles, and the system will automatically generate content based on references.
Functional operation details
- File UploadClick the "Upload Files" button to select the files to be parsed, support batch upload.
- Issue retrieval: Enter a question in the search box, click the "Search" button, and the system will display the relevant answer.
- Answer View: Click on the answer cards for detailed answers and references.
- Content generation: In the content generation module, enter keywords or references, click the "Generate" button, the system will automatically generate an outline or article.
QAnything Core Skills
One-stage retrieval (embedding)
Model name | Retrieval | STS | PairClassification | Classification | Reranking | Clustering | on average |
---|---|---|---|---|---|---|---|
bge-base-en-v1.5 | 37.14 | 55.06 | 75.45 | 59.73 | 43.05 | 37.74 | 47.20 |
bge-base-zh-v1.5 | 47.60 | 63.72 | 77.40 | 63.38 | 54.85 | 32.56 | 53.60 |
bge-large-en-v1.5 | 37.15 | 54.09 | 75.00 | 59.24 | 42.68 | 37.32 | 46.82 |
bge-large-zh-v1.5 | 47.54 | 64.73 | 79.14 | 64.19 | 55.88 | 33.26 | 54.21 |
jina-embeddings-v2-base-en | 31.58 | 54.28 | 74.84 | 58.42 | 41.16 | 34.67 | 44.29 |
m3e-base | 46.29 | 63.93 | 71.84 | 64.08 | 52.38 | 37.84 | 53.54 |
m3e-large | 34.85 | 59.74 | 67.69 | 60.07 | 48.99 | 31.62 | 46.78 |
bce-embedding-base_v1 | 57.60 | 65.73 | 74.96 | 69.00 | 57.29 | 38.95 | 59.43 |
- A more detailed review of the results is detailed inEmbedding model metrics summaryThe
Second-stage search (rerank)
Model name | Reranking | on average |
---|---|---|
bge-reranker-base | 57.78 | 57.78 |
bge-reranker-large | 59.69 | 59.69 |
bce-reranker-base_v1 | 60.06 | 60.06 |
- A more detailed review of the results is detailed inSummary of Reranker model indicators
QAnything Application Scenarios
- Cross-lingual: multiple English-language essay questions and answers
- Information extraction
- A hodgepodge of documents
- Web Q&A