LangChain LAB | 中文

This experiment will verify different use cases of LLM using LangChain, which include chat, role-playing, and document-based QA.

Prerequisites

• python 3.11+
• make
• poetry

Quick Start

Start the Docker container by binding langchain-lab to external port 8080.

docker run -d --rm -p 8080:8080 \
-e OPENAI_API_BASE=https://api.openai.com/v1 \
-e OPENAI_API_KEY=sk-xx \
coolbeevip/langchain-lab

Visit http://localhost:8080 to access the langchain-lab web page.

Development Guide

Create a .env file in the project's root directory and input OpenAI's API key and base url in the file.

OPENAI_API_BASE=https://api.openai.com/v1
OPENAI_API_KEY=sk-xx

Install the dependencies.

make install

Run the project.

make run

Screenshot

Chat Chat-Role RAG Agent

Environmental Variables

You can configure additional parameters either through environment variables or the .env file.

OPENAI_API_BASE / OPENAI_API_KEY

To set the API KEY, use the OPENAI_API_KEY variable. Specify the call address by setting the OPENAI_API_BASE variable. If desired. Alternatively, you can choose not to set these parameters initially and configure them on the page later.

HUGGINGFACE_CATCH_PATH

To set the hugging face cache directory, use HUGGINGFACE_CATCH_PATH. The default value is ./huggingface. When using a Compatible OpenAI API, you can utilize the EMBED huggingface models. After the initial selection, these models will be downloaded to the cache directory. The currently available models are as follows:

• moka-ai/m3e-base
• sentence-transformers/msmarco-distilbert-base-v4
• shibing624/text2vec-base-chinese