Breaking

Developer builds voice assistant 'IRIS' over a weekend linking Gemini Live and Hermes Agent

June 29, 2026 at 14:16 EDT

AI Agents
Foundation Models

Independent developer Ashutosh Shrivastava built and demoed IRIS, a voice assistant that pairs Google's Gemini Live conversation feature with Nous Research's autonomous Hermes Agent, over a single weekend. When a user speaks naturally to Gemini Live, tasks requiring execution are handed off to Hermes Agent running in the background; once the work is done, Gemini Live reports the results and resumes the conversation.

Late June 2026 · Weekend Build

IRIS: A "Build-Your-Own-Jarvis" Voice AI, Coded in One Weekend

A developer fused Google's Gemini Live voice chat with Nous Research's open-source Hermes Agent — pairing a conversation layer that talks with an execution layer that works, all controllable hands-free by webcam gestures.

140k+

GitHub stars for Hermes Agent, the engine at IRIS's core

Weekend to prototype the full voice + gesture integration

Layers: a conversation manager and an execution worker

The Manager / Worker Flow

Conversation Layer

Gemini Live

Talks naturally with the user as the "manager"

→

Execution Layer

Hermes Agent

Runs the task in the background as the "worker"

→

Resume

Surface Results

Gemini proactively reports back, dialogue continues

Delegate · Continue · Review when ready — the conversation never stops while work runs asynchronously.

Built by AI, for the Builder

GPT-5.5 — backend & architecture
Claude Opus 4.8 — frontend & UI
MediaPipe — webcam gesture control
Runs on macOS, source public on GitHub

Developer Reception

"Everyone can build their own Jarvis now."

Praised as "futuristic" with a "manager/worker flow that feels good" — yet still a prototype, with calls for further refinement.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.