Document Chat + RAG

This is a simple email RAG system that uses the any custom API to generate contextual search results for emails in your Gmail inbox.

This repo extends Uppy React SDK to add a Gmail inbox picker and a button to upload emails to the S3 bucket.

At the core it uses the Uppy Companion to upload email markdowns to a S3 compatible bucket and then uses the any RAG API to generate contextual search results.

RAG API is currently set to Minerva (github) but can be easily swapped out for any other RAG API.

For a demo, see here.

1. Setup

1. Clone the repository
2. Copy .env.example into .env, change the environment variables
3. Install the dependencies
4. Run the server

2. Running the project

You'll need to spin up the companion app, I run it via docker (podman)

podman run \
  --name=companion \
  --rm \
  -p 0.0.0.0:3020:3020 \
  -v /Users/dev/drive-integration/tmp/companion:/data/companion:z \
  -e COMPANION_CLIENT_ORIGINS="true" \
  -e COMPANION_SECRET="" \
  -e COMPANION_DOMAIN="" \
  -e COMPANION_DATADIR="/data/companion" \
  -e COMPANION_ENABLE_GOOGLE_PICKER_ENDPOINT="true" \
  -e COMPANION_AWS_BUCKET="" \
  -e COMPANION_AWS_ENDPOINT="" \
  -e COMPANION_AWS_KEY="" \
  -e COMPANION_AWS_SECRET="" \
  -e COMPANION_GOOGLE_KEY="" \
  -e COMPANION_GOOGLE_SECRET="" \
  transloadit/companion

You might also have to set up cors for your S3 bucket

[
  {
    "origin": [
      "http://localhost:5173",
      "https://yourdomain.com",
    ],
    "method": ["GET", "PUT", "POST"],
    "maxAgeSeconds": 3000,
    "responseHeader": [
      "Authorization",
      "x-amz-date",
      "x-amz-content-sha256",
      "content-type",
      "location"
    ]
  },
  {
    "origin": ["*"],
    "method": ["GET"],
    "maxAgeSeconds": 3000,
    "responseHeader": ["Content-Type", "Location"]
  }
]

Lastly, just run yarn dev locally to start.

3. Future Work

• Add Notion integration
• Add LLM to generate conversation