https://github.com/misbahsy/chat-your-data-self-hosted

This repo guides you through building a chatbot on your own data with self hosted LLM
https://github.com/misbahsy/chat-your-data-self-hosted

Last synced: 10 days ago
JSON representation

This repo guides you through building a chatbot on your own data with self hosted LLM

• Host: GitHub
• URL: https://github.com/misbahsy/chat-your-data-self-hosted
• Owner: misbahsy
• License: mit
• Created: 2023-02-07T22:16:11.000Z (almost 2 years ago)
• Default Branch: master
• Last Pushed: 2023-02-14T02:22:25.000Z (almost 2 years ago)
• Last Synced: 2024-12-25T08:20:17.638Z (17 days ago)
• Language: Python
• Size: 116 KB
• Stars: 90
• Watchers: 3
• Forks: 10
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• project-awesome - misbahsy/chat-your-data-self-hosted - This repo guides you through building a chatbot on your own data with self hosted LLM (Python)

README

        
# Chat-Your-Data

Create a ChatGPT like experience over your custom docs using [LangChain](https://github.com/hwchase17/langchain). This repo can help you use models hosted on HuggingFace for embedding and for text generation.

The explanation of [this blog post](https://blog.langchain.dev/tutorial-chatgpt-over-your-data/) can help you understand the reason for different files and the steps involved. We will primarily replace OpenAI API with huggingface based models.

## Environment Variable

Please set huggingface token as `huggingfacehub_api_token=[your-token]`. Token can be generated in the HuggingFace account settings.

## Ingest data

Ingestion of data is done over the `state_of_the_union.txt` file. 

Therefore, the only thing that is needed is to be done to ingest data is run `python ingest_data.py`

## Query data

Custom prompts are used to ground the answers in the state of the union text file.

## Running the Application

By running `python app.py` from the command line you can easily interact with your ChatGPT over your own data.