Document Language: English | 简体中文
This project is a high-performance knowledge base system based on the latest technology stack, designed for large-scale information retrieval and intelligent search. Leveraging Microsoft's Semantic Kernel for deep learning and natural language processing, combined with .NET 8 and React framework, with the backend using MasaFramework, it has implemented an efficient, user-friendly, and scalable intelligent vector search platform. Our goal is to provide an intelligent search solution that can understand and process complex queries, helping users quickly and accurately obtain the information they need.
- Frontend Framework: React + LobeChat + TypeScript
- Backend Framework: MasaFramework based on .NET 8
- Implemented dynamic functions based on the JS V8 engine
- Vector Search Engine: Utilizes PostgreSQL's vector plugin to optimize search performance and also provides DISK
- Deep Learning and NLP: Microsoft's Semantic Kernel to enhance the semantic understanding capability of search
- License: Apache-2.0, encouraging community contributions and usage
- Intelligent Search: Leveraging Semantic Kernel's deep learning and natural language processing technology to understand complex queries and provide precise search results.
- High Performance: Optimizes vector search performance through pgsql's vector plugin, ensuring quick responses even with large amounts of data.
- Modern Frontend: Utilizes the React + LobeUI frontend framework, offering responsive design and user-friendly interfaces.
- Powerful Backend: Based on the latest .NET 8 and MasaFramework, ensuring code efficiency and maintainability.
- Open Source and Community-Driven: Adopts the Apache-2.0 license, encouraging developers and enterprises to use and contribute.
- Powerful dynamic JS Function and provides Monaco for convenient intelligent code suggestions.
- Powerful QA question-answer split mode for more intelligent knowledge base responses.
- Fast integration with Feishu bots.
- Fast integration with WeChat Official Accounts.
- Fast integration with enterprise projects.
Ensure you have installed the .NET 8 SDK, PostgreSQL database, the PostgreSQL vector plugin, and have configured the corresponding environments.
- Clone the repository:
git clone --recursive https://github.com/AIDotNet/fast-wiki.git
-
Install node.js, the latest version (https://nodejs.p2hp.com/).
-
Delete the package-lock.json file and node_modules directory in the web directory.
-
Run the following commands in the web directory:
npm i
npm run build
- Copy the contents under the dist directory in the web directory to the "\fast-wiki\src\Service\FastWiki.Service\wwwroot" directory (create one if wwwroot doesn't exist).
- Install dependencies:
Execute the following in the project root directory:
cd src/Service/FastWiki.Service
dotnet restore
- Database Configuration:
Ensure your PostgreSQL database is running properly and has the necessary databases created. Modify the database connection string in appsettings.json according to your configuration.
Execute the following in the project root directory:
dotnet run
This will start the backend service. Visit http://localhost:5124/ to see the frontend page.
Default username and password: admin Aa123456
FastWikiService environment variables:
- QUANTIZE_MAX_TASK: Maximum concurrency of quantization tasks, default is 3
- OPENAI_CHAT_ENDPOINT: Address of the OpenAI API
- OPENAI_CHAT_EMBEDDING_ENDPOINT: Address of the Embedding API
- OPENAI_CHAT_TOKEN: Token for the OpenAI API
- OPENAI_EMBEDDING_TOKEN: Token for Embedding, default is empty, if empty, the conversation Token is used
- DEFAULT_TYPE: Business database type default sqlite|[pgsql|postgres]
- DEFAULT_CONNECTION: Business database connection string
- WIKI_TYPE: Wiki database type default disk|[pgsql|postgres]
- WIKI_CONNECTION: Wiki database connection string (if disk, it should be a directory)
We welcome all forms of contributions, whether it's feature requests, bug reports, code submissions, documentation, or any other type of support. Please refer to CONTRIBUTING.md to get started.
This project is licensed under Apache-2.0. See the LICENSE file for details.