LLM App Challenge with Helix 1.0

To celebrate the Helix 1.0 launch we're going to run a little competition :-)

Hey, I’m Luke, one of the co-organizers of the “Making Open Source and Local LLMs Work in Practice” event next week on September 12 in San Francisco. I’m looking forward to meeting many folks at the event!

I’m also a co-founder of Helix.ml and we just put out our 1.0 release yesterday. So I wanted to use the meetup as an opportunity to invite everyone to try out the release and give us feedback.

What does Helix do?

Helix lets you build our version of “GPTs” (that we call “Helix Apps”). These let you brand a chat interface, define system prompts, store knowledge, interact with APIs and run gptscripts. Check out the docs for what you can do with them: https://docs.helix.ml/helix/develop/apps/

Why should I use Helix Apps instead of OpenAI?

There are two key differences between Helix Apps and OpenAI GPTs:

• Helix works with open source models. You can install Helix locally alongside Ollama, on a server with a GPU, or connect it to an external LLM provider.

• Helix Apps are defined by a yaml manifest. We think of them like Docker Compose or Kubernetes manifests for LLM Apps. We believe this is important because it allows developers to version control these apps and treat them as deployable artifacts, rather than just clicking buttons in a web interface. At the same time, our simple UI means that anyone in the business can prototype an app. Making it accessible, while also version controlling things under the hood, we believe is a good mix because it means anyone can prototype an LLM app and then DevOps and quality teams can iterate on the models, apply evals and get the app into production.

• Helix Apps can be exposed via an OpenAI-compatible chat completions API. This means you can integrate them into other systems easily, rather than having to access them via the ChatGPT web interface.

• There is a CLI for deploying the helix app manifest: helix apply -f app.yaml (docs)

Why should I bother?

We hope that Helix will be a useful tool for you to build LLM applications that run locally and on your company’s private infrastructure to enable workflows like RAG and API integrations on sensitive data, and that should be motivation enough! But also, if you complete this LLM App Challenge, and if you are able to make it to our meetup on 9/12, you will be rewarded with one of these wonderful t-shirts that Tamao has created 😊 We will hand out the t-shirts to everyone (while supplies last!) who completed the challenge live after the lightning talks at the event!

What is the challenge?

1. Deploy Helix on your own machine or server and use it to develop a cool Helix app that uses system prompts, RAG/knowledge, API integrations, and/or gptscripts

2. Follow our new quickstart guide here: https://docs.helix.ml/helix/private-deployment/controlplane/

3. Then develop a helix app, along the lines of https://aispec.org/ - see the docs here https://docs.helix.ml/helix/develop/apps/

4. Use helix apply -f to deploy it to your private Helix deployment

5. (Optional) Embed the app on a website with the chat widget https://docs.helix.ml/helix/develop/helix-chat-widget/ or make it available in your own Helix app store

6. Make sure to record a video of your app (like a hackathon submission) and post it your on youtube/twitter/linkedin/blog etc. Then come and show it off in our Discord!

7. Bonus points for:

   1. Cool apps that do something innovative with API integrations, knowledge or gptscripts

   2. Deploying the helix controlplane on a public DNS name (docs)

   3. Integrating with business systems we haven’t integrated with before

If you have trouble setting up Helix on your own machine/infrastructure you can also build and deploy the app on our SaaS.

How to submit your app/video/blog

Join our Discord and post in the #general channel!

Need help?

Just ask on Discord - we’re a friendly bunch and will try to help as much as we can :-) Our docs are also a great resource.

We can’t wait to see what you build!

To get some inspiration, check this demo where we build a depressed android that knows about the news (RAG/knowledge), and a crypto bro who can check the current price of cryptos (API integration with coindesk API):

Can I participate if I’m not coming to the meetup?

Absolutely! Please join in remotely and show us what you can build wherever you are in the world. We can’t promise to mail shirts everywhere, but I’ll definitely bring some back to Bristol and London.

Cheers,
Luke