Replies: 2 comments 1 reply
-
|
Update: this has run pretty far ahead of what happy upstream has. But we should still be able to chop it up and merge back pieces if there is an appetite for it. This repo links together my 3 forks and adds some e2e testing, container management, etc: https://github.com/rrnewton/happy-devbox The goal is to make it super easy for anyone to get started with self-hosted Happy.
My primary use case is to combine running Claude in a loop, driven by a weighted random table of prompts, with the observability of being able to watch or intervene remotely via Happy. |
Beta Was this translation helpful? Give feedback.
-
|
I created something similar (except for the ElevenLabs parts, as I'm personally using Wispr Flow/Willow) in https://github.com/leeroybrun/happy-local. Wanted a very low footprint and minimal dependencies, so it uses https://github.com/leeroybrun/happy-server-light, which doesn't need Docker/Postgres/Redis/S3/Minio. Focused on providing an easy-to-setup and easy-to-use fully local Happy stack. You can either use my own forks or the original Happy ones (to which I already sent some PRs). It also builds the UI to serve it directly (no Expo dev server running) and facilitates setting up Tailscale Serve (HTTPS) to easily connect and use the UI remotely (from a mobile, for example). This allows you to easily start the stack on your computer and then access the UI using HTTPS using your Tailscale hostname from anywhere in your Tailscale network. Just dropping that here in case it's helpful for someone else! |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Happy is OSS, but getting set up to hack on it to test a new feature seems ... complicated. I want to test new buttons I'd like to add, or bugfixes I want to work on. But I need to be able to get all three pieces of it running locally, self-hosted and talking to eachother.
The below branch contains my attempt. It's one repo to clone, drop into a container, start happy server, cli, and web client and connect a web browser to the web client in the container: https://github.com/rrnewton/claude_template/tree/happy
The thing is, I was not able to do this without making SOME modifications to the happy repo. I wanted this to be a non-invasive dev harness to submodule all 3 repos without modification. Also, I'm still having a lot of trouble with authentication setups. It seems like there's some missing support for easily redirecting the CLI and clients to connect to a self-hosted setup to connect everything together easily.
If this is not the preferred development setup for the happy (cli + server + client) services then what is? How do the main developers do their dev setup?
Beta Was this translation helpful? Give feedback.
All reactions