This repository provides a web UI for local knowledge-grounded chat using LangChain and ChatGLM-6B/related models, with multiple embedding options and deployment guidance. It supports uploading text formats and exposes HuggingFace, ModelScope, and OpenXLab integrations.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
A Python-based web UI that uses LangChain and ChatGLM-6B series models to build local knowledge-powered chat applications. It supports uploading text formats (txt, docx, md, pdf, etc.) and includes embedding model options such as GanymedeNil/text2vec-large-chinese and nghuyong variants. It provides both HuggingFace and ModelScope deployment paths.
How it works
The project coordinates LLMs (e.g., ChatGLM-6B, Belle) with embedding models to enable retrieval-augmented question answering over local documents. It references dependencies and deployment docs for setup across ModelScope and HuggingFace environments.
Getting started
Requires Python >= 3.8.1. Deployment details are available in the docs:
- 部署文档
- 视频教程
Supported models include:
- large language model: ChatGLM-6B, ChatGLM-6B-int8, ChatGLM-6B-int4, ChatGLM-6B-int4-qe, Vicuna-7b-1.1, Vicuna-13b-1.1, BELLE-LLaMA-7B-2M, BELLE-LLaMA-13B-2M, internlm-chat-7b-8k, internlm-chat-7b-v1_1, internlm-chat-7b
- Embedding model: text2vec-large-chinese, ernie-3.0-base-zh, ernie-3.0-nano-zh, ernie-3.0-xbase-zh, simbert-base-chinese, paraphrase-multilingual-MiniLM-L12-v2
Commands
Detailed deployment instructions are available in the repository’s docs:
- 部署文档
- 视频教程
Notes
The README lists multiple external platforms for online体验, including HuggingFace Spaces and ModelScope, but does not provide a single install script in the truncated portion. The project emphasizes local knowledge QA with text support and various model embeddings.
Recent releases
The latest releases section indicates: - none
Traction
Stars: 3316 Forks: 494 Open issues: 46
Behind the repo
No startup/company link is provided in the truncated README.
Caveats
License: Apache-2.0. Created: 2023-04-10. Last push: 2024-04-15. The project notes initial-stage status and invites community contributions. No specific licensing caveats beyond the Apache-2.0 license are stated in the provided text.