OmniRoute Integration

OmniRoute Integration

OmniGlyph ships two ways: as the standalone proxy described in Quick Start, and as a native compression engine inside OmniRoute — the free AI gateway.

Inside OmniRoute it runs as the omniglyph engine, either standalone (single mode) or stacked with the other engines in the compression pipeline. The same guarantees hold there:

• the exact per-provider billing math is the shared package export (src/core/index.ts), so the gateway and the proxy agree token-for-token;
• the fail-closed model gate still decides which models may receive images;
• image-aware token accounting keeps the gateway's usage numbers honest.

Which one should I use?

You want…	Use
Drop-in savings for one client (Claude Code, etc.)	the standalone proxy (omniglyph)
A gateway across many providers with free tiers, fallback, and stacked compression	OmniRoute with the omniglyph engine enabled

Library use

The billing math is importable without running the proxy:

import { anthropicImageTokens, resolveAnthropicVisionTier } from "omniglyph";

See the package exports map for the ./transform, ./measurement, ./applicability, ./proxy, ./node, and ./worker entry points.