MyModels connects directly to local models via Ollama or executes neural models 100% inside your browser using WebGPU hardware acceleration, WebAssembly CPU execution, and Chrome's native Prompt API (Gemini Nano).
Key capabilities include multiline prompt composition with draft traversal history, streaming token rendering, dual-tab SVG graphics visualization, sequential multi-instruction batch queue, hardware diagnostics, and signed HMAC-SHA256 state backups.
Multiline prompt editor. Auto-expands as text is typed. Shift+Enter for newlines. Up/Down arrows to traverse historical prompt drafts.
[data-tool="prompt-input"]
Categorized selector supporting Ollama daemon models, Chrome Built-in Gemini Nano, and In-Browser WebGPU/WASM models.
[data-tool="model-selector"]
Inspects WebGPU adapter limits, FP16 shader capability, Chrome Prompt API status, and cached offline model weights.
[data-tool="hardware-diagnostics"]
Collapses or expands the left-hand navigation sidebar containing chronological chat history and session controls.
[data-tool="sidebar-toggle"]
Enables batch queuing of multiple prompt instructions and file payloads, executing sequentially with autonomous streaming completion.
[data-tool="command-queue"]
Dedicated rendering canvas for inline SVG vector art generated by LLMs, featuring zoom controls and raw XML source inspection.
[data-tool="svg-viewer"]
Switches between Forgeboro obsidian glassmorphism dark theme and clean high-contrast light mode.
[data-tool="theme-toggle"]
| Shortcut | Action Description |
|---|---|
| Enter | Submit current prompt to model |
| Shift + Enter | Insert newline without submitting |
| ↑ / ↓ | Traverse backward/forward in prompt draft history |
| ? | Open in-app interactive Help & Documentation modal |
| Escape | Close active modal dialog or abort streaming completion |
| Ctrl + Shift + O | Start new chat session |
MyModels supports zero-install local AI running directly in your browser tab. Model weights downloaded by Transformers.js or WebLLM are cached permanently in browser storage for full offline use.
In Chrome 127+, enable chrome://flags/#prompt-api-for-gemini-nano and chrome://flags/#optimization-guide-on-device-model (Enabled BypassPerfRequirement). In chrome://components, update "Optimization Guide On Device Model".
Models like SmolLM2 (135M/360M), Qwen 2.5 (0.5B/1.5B), and Llama 3.2 1B execute via WebGPU shaders with automatic CPU WebAssembly fallback.
Autonomous agents and testing harnesses can inspect and control MyModels programmatically via the global controller:
window.__MYMODEL_AGENT_SPEC__ = {
appName: "MyModels by Forgeboro",
version: "V1-[hash]",
getTool: (id) => document.querySelector(`[data-tool="${id}"]`),
getAction: (id) => document.querySelector(`[data-action="${id}"]`),
spec: { ... }
};
The raw machine-readable specification is accessible at /agent-spec.json.