This project provides a framework for building intelligent customer service chatbots with memory capabilities, leveraging LangGraph, DeepSeek-R1, FastAPI, and Gradio. It supports various large language models, including GPT, domestic models via OneApi, Ollama, and Alibaba's Tongyi Qianwen, offering a flexible and extensible solution for conversational AI applications.

How It Works

The core of the project utilizes LangGraph to define a stateful, graph-based workflow for chatbot interactions. This graph structure allows for complex conversational flows, including dynamic routing, tool integration, and memory management. Short-term memory is maintained within the graph's thread, while long-term memory is persisted using PostgreSQL, enabling continuous and context-aware conversations. The system supports tool calling, dynamic routing based on tool types (retrieval vs. non-retrieval), and includes mechanisms for document relevance scoring and query rewriting.

Quick Start & Requirements

• Installation: Clone the repository from GitHub or Gitee. Install dependencies using pip install langgraph==0.2.74 langchain-openai==0.3.6 fastapi==0.115.8 uvicorn==0.34.0 gradio==5.18.0. For PostgreSQL persistence, install langgraph-checkpoint-postgres and psycopg2.
• LLM Access: Requires access to LLM APIs (e.g., OpenAI, OneApi, Ollama, Tongyi Qianwen). Configuration is managed in llms.py.
• Database: PostgreSQL with pgvector extension is required for persistent memory. Docker can be used to set up the database.
• Dependencies: Python 3.x, FastAPI, Gradio, LangChain, LangGraph, PostgreSQL, NLTK, pdfminer.six.
• Documentation: Links to Bilibili and YouTube videos are provided for setup and demonstrations.

Highlighted Details

• Supports multiple LLM backends, including local Ollama models and various API-based services.
• Implements both short-term (in-thread) and long-term (PostgreSQL) memory management for conversational context.
• Features dynamic routing and tool integration, allowing the chatbot to call external tools and adapt its responses.
• Includes a FastAPI backend for API services and a Gradio frontend for a user-friendly web interface with session management.
• Offers visualization of the LangGraph state machine as a PNG file.

Maintenance & Community

• The project is actively developed by NanGePlus.
• Links to GitHub and Gitee repositories are provided.

Licensing & Compatibility

• The specific license is not explicitly stated in the provided text, but the project appears to be open-source. Compatibility for commercial use would require license verification.

Limitations & Caveats

• The README mentions a potential openai.BadRequestError when using certain models (e.g., OneApi, Qwen) with specific embeddings, requiring a modification in the langchain_openai/embeddings/base.py source code.
• Users need to configure API keys, model parameters, and service IPs/ports according to their environment.