Zeno

Inspiration

We wanted to build an AI assistant that goes beyond answering questions and can actually get things done. Most assistants aren't built to be easily extended with custom tools and background workers.

What it does

ZENO is a voice-first AI assistant that delegates complex tasks to AI workers called Bits. They can research, process files, run code, and handle long-running tasks while you keep talking to ZENO.

How we built it

We used FastAPI, Pipecat, React, and Kotlin, with support for multiple LLM providers. Each Bit runs in its own Docker sandbox with access to tools, files, and connected apps.

Challenges we ran into

Getting voice interaction, background agents, and sandboxed execution to work together was challenging. We also had to figure out how to let workers run independently while keeping their progress and results accessible.

Accomplishments that we're proud of

We built a system where multiple AI workers can run independently, share results, and pick up tasks again when needed. We're especially proud of making ZENO extensible instead of locking it into a single set of capabilities.

What we learned

We learned a lot about building agent loops, connecting different AI models and tools, and managing long-running tasks. We also learned how much architecture matters when building AI systems that need to do more than chat.

What's next for ZENO

We want to add more integrations, improve worker coordination, and make ZENO easier to customize. The long-term goal is a personal AI assistant that can handle more of your work with less supervision.

Built With

• elevenlabs
• fastapi
• react