← knowledge.oriz.in

AI services

index servicesaiindex

AI services

Two AI providers, picked by surface:

Surface Service Why this surface
Browser (chat in oriz-me, on-page assistants) Puter.js User-pays; we ship no API key, the family pays nothing per request
Server (inside the umbrella Hono Worker at api.oriz.in) Cloudflare Workers AI Native Worker binding, zero-egress within Cloudflare, 10K neurons / day free
Hosted Gemini (only when a feature truly needs Google's model) Firebase AI Logic Free tier on Spark; document-shaped fits our existing Firebase posture

OpenRouter is rejected — Puter.js mirrors its model IDs already, and calling OpenRouter directly would require a server-paid account.

Per-service files

Service Status One-line role
puter-js.md active Browser-side AI calls — user-pays free tier, no API keys client-side
cloudflare-workers-ai.md active Server-side AI calls inside the Hono Worker — 10K neurons / day, zero-egress
openrouter.md rejected Server-side LLM gateway — rejected; Puter.js mirrors its model IDs and Workers AI covers server inference

Why two AI services and not one?

A single provider can't cover both surfaces well:

Locked at decisions/architecture/ai-puter-plus-cf-workers-ai.md.

Cross-refs