Inference Bridge

Official reference implementation of the experimental Inference Provider API browser proposal.

Inference Bridge is the official reference implementation of the experimental Inference Provider API (IPA). It injects window.inference into web pages, asks for per-origin permission, and routes chat requests to a provider you choose — OpenAI (bring your own API key) or local Ollama. API keys stay in the extension; page scripts never see them. Features • window.inference.request() for streaming text chat • Per-origin Allow / Deny / Remember permission flow • User-controlled provider and model selection • OpenAI (BYOK) and local Ollama support • Origin/Referer stripping so local Ollama works without special OLLAMA_ORIGINS setup in the common case • Secure-context injection only (https: or loopback http:) Security • Injects only into top-level frames • Requires a secure context (https: or localhost / loopback http:) • Does not inject into file: pages • Permission is per HTTP(S) origin and records the chosen provider + model • Request validation runs in the extension before any provider call • OpenAI credentials are read only inside the service worker • Ollama traffic stays on localhost:11434 / 127.0.0.1:11434 This extension implements an experimental browser API proposal. See the specification: https://github.com/SamSamskies/inference-provider-api Privacy policy: https://github.com/SamSamskies/inference-bridge/blob/main/PRIVACY.md