EmoBot v30.0 is a high-performance, multimodal AI companion built for one purpose: evaluating human failure with clinical precision. Utilizing the Google Gemini 3 Pro & Flash models, it features Thinking Budget Reasoning, Live Multimodal Streaming, and Real-World Grounding—all wrapped in a dark, cyberpunk terminal interface.
- Multimodal Perception (SCAN BIO): Uses
gemini-3-flash-previewto analyze webcam frames and assign mean, grumpy nicknames based on your "vibe." - Deep Thinking Mode (THINK): Leverages
gemini-3-pro-previewwith a 32k thinking budget to ruminate on the pointlessness of your queries before answering. - Native Audio Interaction: Features a high-fidelity Live Session engine for low-latency, real-time voice conversations using the latest Gemini Native Audio models.
- Search & Maps Grounding: Scours the live web to find evidence for its cynicism or identifies locations where you can "hide from society."
- Multimodal Transcription: Transcribes your "human mutterings" using high-speed audio-to-text processing for further analysis.
- Animated SVG Core: A reactive robot face that responds to bot states (Thinking, Listening, Speaking) with dynamic blinking and squinting.
- Self-Destruct Sequence: A dramatic, terminal-style shutdown sequence for when the bot's patience finally reaches zero.
| Layer | Technology |
|---|---|
| Frontend Core | React 19 (ESM) |
| AI Engine | @google/genai (Gemini 3 Pro/Flash) |
| Vision Engine | HTML5 Canvas & MediaDevices API |
| Audio Pipeline | Web Audio API & PCM Streaming |
| UI Rendering | Tailwind CSS & ANSI-Style Aesthetics |
EmoBot/
├── App.tsx # Main Application Logic & State
├── components/
│ └── RobotFace.tsx # Animated SVG Visual Interface
├── services/
│ └── geminiService.ts # Multimodal Model Routing & Tools
├── utils/
│ └── audioUtils.ts # PCM Encoding/Decoding Logic
├── types.ts # Global State & Subject Definitions
└── index.html # Terminal-Themed Entry Point🔵 [SCAN BIO] - Capture a visual frame and update your Subject Nickname.
🟢 [START LIVE SESSION] - Initiate a real-time, native audio conversation.
🟣 [THINK] - Force the bot to use its deep reasoning budget on your query.
🔴 [TERMINATE] - Trigger the unrecoverable self-destruct sequence.
# EmoBot is built for ESM-compatible environments.
# To run locally:
# 1. Provide your API Key in the environment
# export API_KEY="your_google_ai_studio_key"
# 2. Launch through your preferred dev server
npm install
npm run dev
# 3. Access the interface at localhostDeveloper Note: This project serves as a reference implementation for Multimodal Task-Specific Routing, demonstrating how to coordinate Gemini Pro (for complex reasoning) and Gemini Flash (for speed/vision) within a single reactive UI.
Meet Potdar
Backend & Creative Technologist
Built with ❤️ (and extreme sarcasm) using Google Gemini