gmail-chat enables to index the emails in your Gmail account, create vector embeddings for each, and store them in Pinecone. Then, you can ask questions about your emails in natural language, using conversational models.

• Poetry package manager for Python. Can be installed by running

    $ brew install poetry
  

  on MacOS X.

• Python 3.11

• Required Python modules can be installed by running

    $ poetry update
  

• The following environment variables must be set:

  • OPENAI_API_KEY
  • PINECONE_ENVIRONMENT
  • PINECONE_API_KEY

• Optionally, you can download a credentials.json file from Google Cloud Console to provide the credentials for the Gmail API.

In order to enable gmail-chat to index your emails, you need to authorize API access by creating a OAuth authorization using the Google Cloud Console.

1. Go to https://console.cloud.google.com/apis.
2. Create a new project
3. Make sure Gmail API is enabled on https://console.cloud.google.com/apis/api/gmail.googleapis.com
4. Go to the Credentials screen (https://console.cloud.google.com/apis/credentials)
5. Click CREATE CREDENTIALS and choose OAuth Client ID
6. Add http://localhost:52102 to the Authorized Redirect URIs list
7. Download the client secret JSON to the directory from which you intend to run gmail-chat

You can run gmail-chat by executing

$ poetry run gmail-chat <command> [query]

where command is one of the following supported commands:

• create - Creates a Pinecone index for storing the email embeddings.
• delete - Deletes the Pinecone index.
• index - Indexes Gmail emails, and stores their vector embeddings in Pinecone.
• ask - Ask a question about the email content. Requires a query argument to be passed.
• chat - Run a conversation bot that can answer a series of questions about the indexed emails.