Launching today
Assemblix
Chat agents as workflows - with voice & avatars
1 follower
Chat agents as workflows - with voice & avatars
1 follower
Assemblix is a visual engine for building conversational AI agents. Design the whole chat as a graph - branching logic, any LLM (OpenAI, Gemini, DeepSeek), knowledge bases, and typed state that remembers context across turns. When text isn't enough, flip a switch to give any agent real-time voice or a live, lip-synced AI avatar. Test instantly, replay every run, bring your own keys โ and self-host it all. Open and source-available.










Hey Product Hunt ๐
We built Assemblix to solve our own problem - and it turned out to be a much bigger one.
We run a school that teaches managers to communicate, and we needed to create a lot of realistic practice: interactive exercises where someone rehearses a tough 1:1 or a negotiation and gets feedback. No tool fit. These aren't one-shot prompts - they're long, branching conversations that have to remember what was said, and the people building them aren't developers.
So we ended up building what we actually needed: a visual engine for chat AI agents. You design the whole conversation as a graph โ drag nodes onto a canvas, branch the logic, plug in any LLM and your own knowledge, and it keeps typed state that carries context across the entire chat. When text isn't enough, one switch gives an agent real-time voice or a live AI avatar.
And it turns out an engine for chat workflows is exactly what you need far beyond us - support agents, tutors, interviewers, onboarding, role-play. Anything that's a real conversation instead of a single prompt.
What you get:
- Visual builder - chat agents as a node graph, no code
- Any LLM - OpenAI, Gemini, DeepSeek โ plus knowledge bases
- Typed state that persists across every turn
- Real-time voice or a live AI avatar
- Test & replay every run ยท bring your own keys ยท self-host it all (open & source-available)
We'd genuinely love your feedback โ and a โญ on GitHub means a lot to a small team ๐