App Settings

Configure app defaults, manage browser-cached models, and view storage metrics.

Defaults Configuration

This model will be selected by default when initiating a new chat conversation.

WebGPU executes inference on your GPU. WASM runs on the CPU and acts as a universal fallback.

Quantization reduces the precision of model weights to save RAM/VRAM and speed up generation.

0.7

Higher values increase randomness/creativity, lower values make responses more focused and deterministic.

The maximum length of generated response. Larger sizes require more memory and take longer to complete.

Turn off to use greedy decoding for maximum consistency (highly recommended for code generation).

Agent Features (Client-side)

In-Browser Agentic Tool CallingActive

Allow local LLMs to dynamically run client-side JavaScript calculations, fetch web page content, and generate files.


Sandboxed Code Execution (Code Interpreter)Active

Execute generated code safely inside a browser sandbox and let the model iteratively self-correct code execution failures.


Client-Side RAG (Local Search)Backlog

Use an embedding model to index uploaded files locally in your browser and inject relevant segments into the chat context.


Background Helper AgentsBacklog

Enable background routing to the optimal model size, auto-titling of chat sessions, and automated prompt extensions.

Storage & Cache Metrics

Total Conversations

0

Model Cache Storage

Calculating...

Hugging Face Cache Management

Models downloaded via Transformers.js are saved in the browser's Cache Storage. Cleared history only deletes text logs. To delete downloaded models, you can clear browser site data / cache.

Danger Zone

Clear Chat History

Permanently delete all chat sessions and messages stored locally in localStorage.