AI LocalBase
AI LocalBase is a local-first AI knowledge base system used to access individual or team documents to vector retrieval, RAG dialogues, and MCP tool calls.
##Functional characteristics - Knowledge base management: Support the creation, selection, and deletion of knowledge bases, and use different knowledge bases to isolate personal notes, project materials, team documents or test data. - Document upload and index: Support TXT, Markdown, PDF, CSV, XLSX file upload, and also support catalog batch upload. The backend will extract text, segment content, and write it to the vector library. - Retrieve-enhanced Q & A: Generate query vectors for user questions, recall relevant fragments from Qdrant, and inject the content of the hit document into the chat context. - Chat model configuration: Provider, Base URL, Model, API Key, Temperature and number of context messages can be configured separately in the "Chat Model" area of "Settings". - Embedding model configuration: Embedding Provider, Base URL, Model and API Key can be configured independently to facilitate the use of Olama or OpenAI Compatible services. - Rapid mode and thinking mode: Rapid models can be used for daily question and answer, and thinking mode models can be configured separately for complex reasoning. - Chat history persistence: The conversation list, message content and update time are saved to SQLite, and historical conversations can still be viewed after the app is restarted. - Configuration persistence: Chat / Embedding / MCP configuration and knowledge base status are saved to the backend data directory; thinking pattern models are saved separately in localStorage of the current browser. - MCP Server: Built-in MCP entry in HTTP/JSON-RPC form, providing knowledge base list, document list, retrieval, upload, session reading and writing, and dangerous deletion tools. - MCP authority protection: The MCP interface uses a Bearer Token for authentication, and dangerous tools also require a second confirmation of the Token. ###First use process 1. Open application address 2. The current online interface will first display a default session "New Conversation", and automatically select a "Default Knowledge Base". 3. Click "Settings" at the bottom of the left to open the "AI Settings" pop-up window. In the pop-up window, you will see three sections: "Chat Model","Embedding Model" and "MCP Settings". 4. First configure the Provider, Base URL, Model, and API Key in the "Chat Model". This page is saved instantly and does not require an additional click of the Save button. 5. Then configure the Provider, Base URL, Model and API Key in the "Embedding Model". 6. If there is a computing power warehouse, please refer to the configuration below, or see the attached figure - Chat Provider：`OpenAI Compatible` - Chat Base URL: `https://ollama-ai. Your device name.heiyu.space/v1` - Chat Model：`qwen3:8b` - Chat API Key: `Your api key` - Thinking pattern model: `gpt-oss: 20b` - Embedding Provider：`OpenAI Compatible` - Embedding Base URL: `https://ollama-ai. Your device name.heiyu.space/v1` - Embedding Model：`nomic-embed-text:latest` - Embedding API Key: `Your api key` 7. Click "Knowledge Base" at the bottom on the left to enter "Knowledge Base Management". 8. You can directly use the "Default Knowledge Base" automatically created by the system, or you can click "+ New Knowledge Base" to create a dedicated knowledge base. 9. Click "Upload File" or "Upload Catalog" in the Knowledge Base card to upload a TXT, Markdown, PDF, CSV or XLSX document. 10. After waiting for the document status to change to "indexed", go back to the chat interface to ask questions; at the bottom, you can switch between "Quick Mode" and "ThinkingMode".
