An Electron and Three.js companion robot prototype driven by structured JSON actions from a local Ollama model.
What it does
Turns a local Ollama model into an animated desktop robot that can respond through structured actions, speech, expressions, camera snapshots, and offline fallbacks without depending on cloud AI.
Brobot Companion Lab desktop app screenshot with a Three.js robot face and model controlsBrobot Companion Lab architecture diagram showing inputs, Electron, Ollama, JSON actions, and robot face output
Key Highlights
Built an Electron desktop wrapper with a secure preload bridge.
Rendered a Three.js robot face pod with animated eyes, mouth, neck motion, and ambient lighting.
Integrated Ollama chat through a local server at 127.0.0.1:11434.
Designed a brobot.action.v1 JSON schema so the LLM controls speech, eyes, mouth, neck, environment, voice, and behavior predictably.
Added text input, image upload, voice dictation, speech output, camera snapshots, and offline fallback behaviors.
Role and Contributions
Designed the structured LLM action contract for robot behavior.
Built the Electron shell and browser-side companion robot experience.
Implemented multimodal input paths and local fallback behavior.
Kept the prototype local-first by using Ollama instead of a remote model dependency.