/chat_llm

A RAG pipeline using LangChain, OpenAI API, Gradio to chat with your PDFs!

Primary LanguagePythonApache License 2.0Apache-2.0

Chat with your own data!

An app that lets you chat with your documents (PDFs); the app makes use of the OpenAI LLM using LangChain as the framework.

.
├── README.md
├── chat.py                             --> Main app
├── config.conf                         --> Configuration file that allows one to change the models etc
├── input_docs                          --> Document directory to store your PDFs
│   └── Python-ML.pdf
├── logs                                --> Log directory; app logs as well as user logs
├── requirements.txt                    
├── restart_service_on_changes.py       --> Restarts the `chat.py` if there are changes in the `input_docs` folder
└── translator_langchain.py             --> Used for supporting the translation of the queries

How it works?

chat.py is the main app that connects to the LLM (OpenAI) based on the configuration that you give in the config.conf file. First, we split the document based on the configurations and then create a database index which will be used later to query. The database that we use is a vector database called, 'ChromaDB'. This gets created when the script is executed for the first time. Also, for 'ChromaDB' to work, one needs to have a Sqlite3 DB, so make sure to install one.

Create a .env file and add your OpenAI API key as:

OPENAI_API_KEY=sk-XYZ

The app makes use of 'Gradio' framework; so make sure to run the following command to install the required libraries:

pip install -r requirements.txt

Once the required libraries are installed, you can start the application by:

python3 chat.py, will output the link which you can paste into your browser to access the app.

The chat.py in turn makes use of translator_langchain.py to provide a translation feature.

In case, the user wants the indexing to be updated whenever there are changes to the input_docs like when a new PDF file is added/removed; the user can make use of restart_service_on_changes.py -- this will start the chat.py internally, so you need to make sure to close any previous running instance of chat.py.

Note: If you are using restart_service_on_changes.py --> then expect brief service disruption whenever there are changes to the input_docs folder as the script will stop the chat.py momentarily to index the new files.