About

Vectorize a Google Drive folder content to a vector database to query and obtain answers using ChatGPT.

Technology Stack

• Google Drive
• LangChain
  • Chroma DB
• ChatGPT
• Flask

Project Structure

A very simple and straightforward personal chatbot API project written in Python, using ChatGPT, LangChain, and Flask, to answer questions about your files in Google Drive.

+-- db --+
|        |
|        +-- ChromaDB files
|
+-- flask --+
|           |
|           +-- app.py
|
+-- gdrive_copy --+
                  |
                  +-- copy_gdrive.py

Requirements

Google Drive API enabled

You need a Google Account to log into the Google Cloud Console and enable the Google Drive API. This is the most tedious part.

ChatGPT API

You need an OpenAI Account and enable the ChatGPT API.

Procedure

Considering that all the requirements are met, follow the steps below:

1. Clone the repository

   git clone https://github.com/robson-koji/ChatGPT-GDrive-Integration.git
   

2. Install Python Requirements You may need to install the necessary dependencies (Flask, OpenAI, etc.) using pip or any package manager you prefer before running the code.

3. Set Environment Variables Set the following environment variables to use the system:

   - OPENAI_API_KEY
   (e.g., export OPENAI_API_KEY=<YOUR OPENAI API KEY> in Unix-based systems)
   
   - GDRIVE_FOLDER_ID
   (e.g., export GDRIVE_FOLDER_ID=<YOUR GDRIVE_FOLDER_ID> in Unix-based systems)
   

4. Run the script below. It will download data from your Google Drive folder and load it into your local Chroma DB:

   cd <PROJECT PATH>/gdrive_copy/
   python copy_gdrive.py
   

5. Run the Flask application and turn on the endpoint to start querying your data.

   python app.py or flask run
   

Flask Application

I've created a Flask application and defined a GET endpoint /api/chat. The endpoint expects a query parameter called query to be passed with the GET request. It then runs the qa.run() function to obtain an answer based on the provided query and returns the answer as the API response.

The API will be accessible at http://localhost:5000/api/chat.

Using Examples

You can ask anything that is contained in the documents from your Google Drive folder. If everything is set correctly, ChatGPT will answer as expected.

Start making a GET requests to this endpoint by appending the query parameter like http://localhost:5000/api/chat?query="hi".