AutoFlow: GraphRAG-based conversational knowledge base/deep web search tool with dialogs that can be integrated into other websites

General Introduction

AutoFlow is an open source tool developed by PingCAP to build graph-based knowledge bases with TiDB serverless vector storage. It integrates LlamaIndex and DSPy framework to support complex dialog search and knowledge graph editing. Users can embed a dialog search window into their website with a simple JavaScript code snippet to enhance the user experience.

Online experience: https://tidb.ai/

Function List

• knowledge graph construction: Based on TiDB vector storage, it supports the creation and editing of knowledge graphs.
• Dialog Search: Integrates with the LlamaIndex and DSPy frameworks to provide similar Perplexity The dialog search function of the
• Serverless Storage: Supports large-scale data storage and management with TiDB serverless vector storage.
• Embedded Search Window: Embed a dialog search window into a website via a JavaScript code snippet.
• Docker Deployment: Supports local deployment using Docker and Docker Compose.

Using Help

Installation and deployment

1. local deployment::
   • Ensure that Docker and Docker Compose are installed.
   • Clone the AutoFlow repository:

    git clone https://github.com/pingcap/autoflow.git
   

   • Go to the project directory and start the service:

    cd autoflow
   docker-compose up -d
   

   • To access a locally deployed AutoFlow instance, the default address is http://localhost:3000The
2. Embedded Search Window::
   • Copy the following JavaScript code snippet and embed it in your website: html
¨K16K

   • The code snippet will display a dialog search window in the bottom right corner of the site where users can search and interact directly.

Function Operation Guide

1. knowledge graph construction::
   • Access the AutoFlow administration interface and select the Knowledge Graph module.
   • Use the built-in editing tools to add, modify or delete knowledge nodes and relationships.
   • After you save your changes, the Knowledge Graph will be automatically updated and available for dialog search.
2. Dialog Search::
   • Enter a query in the dialog search window and the system will provide relevant answers based on the knowledge graph and vector database.
   • Users can further refine their queries through dialogs to get more precise information.
3. Serverless Storage Management::
   • Access the TiDB administration interface to view and manage stored data.
   • Supports data import, export and backup to ensure data security and high availability.

Detailed Operation Procedure

• Adding Knowledge Nodes::
  1. Click the "Add Node" button.
  2. Enter the node name and description and select the node type.
  3. Save the node and the system will automatically update the knowledge graph.
• Editing Knowledge Relationships::
  1. Select two nodes and click the "Add Relationship" button.
  2. Enter a relationship description and select a relationship type.
  3. Save the relationship and the Knowledge Graph will be automatically updated.
• Dialog Search Usage::
  1. Enter a question in the search window, such as "What is AutoFlow?
  2. The system will provide answers based on the knowledge graph and display the relevant nodes and relationships.
  3. Users can click on the nodes to view details or continue to ask questions for more information.