/Open-Assistant

OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.

Primary LanguagePythonApache License 2.0Apache-2.0

Open-Assistant

Open Assistant is a project meant to give everyone access to a great chat based large language model.

We believe that by doing this we will create a revolution in innovation in language. In the same way that stable-diffusion helped the world make art and images in new ways we hope Open Assistant can help improve the world by improving language itself.

Do you want to try it out?

If you are interested in taking a look at the current state of the project, You can set up an entire stack needed to run Open-Assistant, including the website, backend, and associated dependent services.

To start the demo, Run this in the root directory of the repository:

docker compose up --build

Then, navigate to http://localhost:3000 (It may take some time to boot up) and interact with the website.

Note: When logging in via email, navigate to http://localhost:1080 to get the magic email login link.

The Plan

We want to get to an initial MVP as fast as possible, by following the 3-steps outlined in the InstructGPT paper.

  1. Collect high-quality human generated Instruction-Fulfillment samples (prompt + response), goal >50k. We design a crowdsourced process to collect and reviewed prompts. We do not want to train on flooding/toxic/spam/junk/personal information data. We will have a leaderboard to motivate the community that shows progress and the most active users. Swag will be given to the top-contributors.
  2. For each of the collected prompts we will sample multiple completions. Completions of one prompt will then be shown randomly to users to rank them from best to worst. Again this should happen crowd-sourced, e.g. we need to deal with unreliable potentially malicious users. At least multiple votes by independent users have to be collected to measure the overall agreement. The gathered ranking-data will be used to train a reward model.
  3. Now follows the RLHF training phase based on the prompts and the reward model.

We can then take the resulting model and continue with completion sampling step 2 for a next iteration.

The Vision

We are not going to stop at replicating ChatGPT. We want to build the assistant of the future, able to not only write email and cover letters, but do meaningful work, use APIs, dynamically research information, and much more, with the ability to be personalized and extended by anyone. And we want to do this in a way that is open and accessible, which means we must not only build a great assistant, but also make it small and efficient enough to run on consumer hardware.

Slide Decks

Vision & Roadmap

Important Data Structures

How can you help?

All open source projects begins with people like you. Open source is the belief that if we collaborate we can together gift our knowledge and technology to the world for the benefit of humanity.

I’m in! Now what?

Join the OpenAssistant Contributors Discord Server!, this is for work coordination.

Join the LAION Discord Server!, it has a dedicated channel and is more public.

and / or the YK Discord Server, also has a dedicated, but not as active, channel.

Visit the Notion

Taking on Tasks

We have a growing task list of issues. Find an issue that appeals to you and make a comment that you'd like to work on it. Include in your comment a brief description of how you'll solve the problem and if there are any open questions you want to discuss. Once a project coordinator has assigned the issue to you, start working on it.

If the issue is currently unclear but you are interested, please post in Discord and someone can help clarify the issue with more detail.

Always Welcome: Documentation markdowns in docs/, docstrings, diagrams of the system architecture, and other documentation.

Submitting Work

We're all working on different parts of Open Assistant together. To make contributions smoothly we recommend the following:

  1. Fork this project repository and clone it to your local machine. (Read more About Forks)
  2. Before working on any changes, try to sync the forked repository to keep it up-to-date with the upstream repository.
  3. Work on a small focused change that only touches on a few files.
  4. Run pre-commit and make sure all files have formatting fixed. This simplifies life for reviewers.
  5. Package up a small bit of work that solves part of the problem into a Pull Request and send it out for review.
  6. If you're lucky, we can merge your change into main without any problems. If there's changes to files you're working on, resolve them by:
  7. First try rebase as suggested in these instructions.
  8. If rebase feels too painful, merge as suggested in these instructions.
  9. Once you've resolved any conflicts, finish the review and merge into main.
  10. Merge in your change and move onto a new issue or the second step of your current issue.

Additionally, if someone is working on an issue that interests you, ask if they need help on it or would like suggestions on how to approach the issue. If so, share wildly. If they seem to have a good handle on it, let them work on their solution until a challenge comes up.

When does a review finish

A review finishes when all blocking comments are addressed and at least one owning reviewer has approved the PR. Be sure to acknowledge any non-blocking comments either by making the request change, explaining why it's not being addressed now, or filing an issue to handle it later.

Developer Setup

Work is organized in the project board.

Anything that is in the Todo column and not assigned, is up for grabs. Meaning we'd be happy for anyone to do these tasks.

If you want to work on something, assign yourself to it or write a comment that you want to work on it and what you plan to do.

  • To get started with development, if you want to work on the backend, have a look at scripts/backend-development/README.md.
  • If you want to work on any frontend, have a look at scripts/frontend-development/README.md to make a backend available.

There is also a minimal implementation of a frontend in the text-frontend folder.

We are using Python 3.10 for the backend.

Check out the High-Level Protocol Architecture

Website

The website is built using Next.js and is in the website folder.

Pre-commit

Install pre-commit and run pre-commit install to install the pre-commit hooks.

In case you haven't done this, have already committed, and CI is failing, you can run pre-commit run --all-files to run the pre-commit hooks on all files.

Deployment

Upon making a release on GitHub, all docker images are automatically built and pushed to ghcr.io. The docker images are tagged with the release version, and the latest tag. Further, the ansible playbook in ansible/dev.yaml is run to automatically deploy the built release to the dev machine.