Filament

Filament now routes to @AskVenice llama-3.3-70b is live alongside Claude, GPT-4o, and Gemini — one endpoint, four models, zero switching costs. filament.works/app/router

you can also register any external tool with a schema and it becomes callable from any model. one registration, works everywhere. and if you're already using the OpenAI client — you just change the base URL. nothing else. that's it. one wire, whole stack.

every call comes back with the response + a trace: tokens in, tokens out, latency, which model ran it. if a model fails, it falls back. no cold-starts, no broken flows.

how Filament works: you send one request to filament.works/api/route Filament looks at your prompt and picks the right model automatically — — long-form reasoning → Claude — code generation → GPT-4o — factual / search-heavy → Gemini

Filament update. router playground is live inside the app. pick a model, type a prompt, see the response stream in real time — with full trace: tokens in, tokens out, latency, model used. still early but the core works. Claude, GPT-4o, Gemini — all routing through one endpoint.

Hello world