Firefox AI Runtime

Live views of the Remote Settings collections powering the Firefox AI runtime and Smart Window

Inference Options

Which ML model is deployed for each Firefox feature — task type, model ID, quantization, and revision.

ml-inference-options

ONNX Runtime Binaries

The WebAssembly runtime files (ONNX and wllama) hosted in Remote Settings for on-device inference.

ml-onnx-runtime

Model Allow/Deny List

URL prefixes that are allowed or denied when Firefox loads models from external hosts.

ml-model-allow-deny-list

Words Block List

Language-specific n-gram block lists used for content filtering during ML inference.

ml-inference-words-block-list

AI Window Prompts

System prompts and configuration for each Firefox Smart Window feature and AI model.

ai-window-prompts

AI Window Tools

Tool/function definitions the Smart Window assistant can call — tab management, memories, search, and more.

ai-window-tools

AI Window Telemetry Prompts

LLM prompts that classify conversations to produce telemetry, with their allowed output values.

ai-window-telemetry-prompts

AI Window Telemetry Triggers

Sampling rules that decide when telemetry prompts run.

ai-window-telemetry-triggers

Inference Request Options

Per-feature inference request options (system prompt, input size) stored as a JSON blob.

ml-inference-request-options