Raggle By Context.ai

Raggle is an open source, self-hosted alternative to Glean and ChatGPT: an employee LLM assistant with enterprise RAG search that can be deployed in your private cloud, without sharing data.

• Perform RAG over all your team's data, today supporting Google Docs and Gmail, with more integrations coming soon
• Chat with many frontier LLMs, with all the features you’d expect - like conversation history and document upload
• Coming soon: support for self-hosted LLMs, and more search integrations

Get Started

Raggle is designed to be fully self-hosted and can be deployed on most cloud infrastructure that supports Docker images.

Configuration

1. Generate encryption credentials for your deployment.

   $ docker run ghcr.io/contextco/chat:latest bin/generate-keys
   
   ENCRYPTION_KEY=xxx
   ENCRYPTION_DETERMINISTIC_KEY=xxx
   ENCRYPTION_KEY_DERIVATION_SALT=xxx
   SECRET_KEY_BASE=xxx
   

   Copy the keys into your .env file (or otherwise into the environment where you will deploy this Docker image).

2. Create a new Google OAuth application that can be used to authenticate your team and allow their Google Workspace data to be indexed. Follow the instructions in the wiki to setup this integration.

3. Add your OpenAI and Anthropic (optional) API keys as environment variables:

   OPENAI_KEY=xxx
   ANTHROPIC_API_KEY=xxx
   

Deployment

A Raggle instance can be run in one of two ways:

• Monolithic This version runs all dependencies within a single Docker image and is the simplest to setup.

  Pros:

  • All dependencies are included within a single Docker image, so no additional external services are required.
  • Easiest to deploy and get started with.

  Cons:

  • Requires provisioning and mounting of a persistent disk to maintain database state across restarts.
  • Harder to manage backups and versioning of persisted data.

• Server Only Allows you to run the server and workers as independent services.

  Pros:

  • Can reuse existing database server resources.
  • Requires slightly less resources to deploy.

  Cons:

  • Requires maintaining additional external Redis and Postgres dependencies, which may be overhead if you don't already maintain these.
  • Some additional setup required to point the server instance at your external dependencies.

Monolithic Deployment

 # Assuming your environment variables are configured in .env

 docker run --env-file=.env \
   -p 3000:3000 \
   -v sidekick-data:/var/lib/postgresql/15/main \
   ghcr.io/contextco/chat:latest bin/monolith

Server-Only Deployment

For a server-only deployment you will need to configure two additional environment variables:

• DATABASE_URL pointing to an accessible Postgres server (eg: postgresql://user:password@host:5432/my_database)
• REDIS_URL pointing to an accessible Redis server (eg: redis://username:password@host:6379)

 # Assuming your environment variables are configured in .env

 docker run --env-file=.env ghcr.io/contextco/chat:latest -p 3000:3000 bin/server-only

Feedback?

We would love to hear it! Open an issue and let us know, or email us at henry@context.ai