AutoFlow Knowledge Base

AutoFlow Knowledge Base is aimed at individuals, teams and enterprise knowledge base scenarios. It supports uploading Markdown, text and other documents into the knowledge base, automatically cutting into document blocks and generating vector indexes. Users can directly ask questions to the knowledge base through chat. The system will combine vector retrieval, query rewriting, knowledge map retrieval and large language model to generate answers, and display relevant sources to facilitate users to trace the basis of answers. The application supports the configuration of different large language models and Embedding Models. It can be connected to cloud APIs or used with local Embedding/Reranker services. Administrators can manage the knowledge base, data sources, model configuration and system settings, and ordinary users can conduct public Q & A without login. Suitable for product document Q & A, technical manual retrieval, internal knowledge base, personal data collation, learning assistant and other scenarios. The local Embedding image is large and the first download takes time. Cloud APIs can be used. The database can be tidb cloud in the cloud or pingkai caloud cloud database. You can also use locally deployed tidb, or use the TiDB Vector Lab customized for lazy cats with its own SQL Edit function. Instructions for use: A series of settings will be made during installation. At least: SECRET_KEY: At least 32 bits Administrator password: at least 16 digits LLM API Key: Your DeepSeek Key Other local Embedding and other parameters can be left as default. After the installation is completed, you need to click login at the bottom left to automatically inject confidential information and automatically login, and then use Knowledge Bases to create and upload documents;Chat Engines makes settings and selects the Knowledge Bases you just created.