dvijkalsi / docgpt-opensource

Here is a an open source version of documentGPT using which you can ingest your documents onto a vector database and use a gpt model to answer questions from it

DOCGPT-opensource

Ingest your documents using a vector database and hugging-face embeddings and use an open-source gpt model to ask questions related to the document using a Gradio frontend chat interface

step 1

Create a virtual environment and run pip install -r requirements.txt

step 2

Create a folder 'model' and add the quantized groovy model which you can get from https://gpt4all.io/ In conversation.py Edit MODEL_PATH = "" with the absolute location of your model

step 3

Create a folder 'source' and add all the documents that you want to ingest

step 4

Run index.py which should create a vector database in the folder 'db' ingesting the documents

step 5

Run bot.py to create a frontend using gradio and then you can ask your questions related to the document there

YAY, YOU'RE DONE!!