Skip to content

Migration guidesFrom Unify

Off the Unify router

Unify offered one OpenAI-compatible API across providers, addressed as model@provider, with routing to the best model per prompt. Its site now describes a different product. We could not read its router docs today, so this page claims only what we could confirm, and shows the move.

Pivoted to continual learning; router docs no longer public; GitHub repo archived 2026-10-16

INFRO accounts are self-serve. The catalog, the docs, and the API contract are public, the gateway is deployed, and anything beyond v1 is marked Later. Plans and trial

Where things stand

What we could read

unify.ai presents continual learning for AI teams (Continual-ARC). The unifyai/unify repository was archived by its owner on 2026-10-16. docs.unify.ai and the API's endpoint listing redirect to a sign-in.

What we could not

An official notice that the router is shut down, and whether paid traffic is still served. If yours still works, there is time; plan the move anyway.

The current Python package

unifyai on PyPI is now a logging and evaluation SDK (unify.activate, unify.log), not a router client.

INFRO is one of several. It is the right one only if this matches.

  • INFRO runs on provider accounts you hold. If you used Unify to avoid opening accounts at each provider, an aggregator such as OpenRouter fits better, and can itself be one of your INFRO connections.
  • INFRO does not route to the best model per prompt by benchmark. A route's prompt classifier sends classes of prompts to targets you choose.
  • If you are using the new unifyai package for logging and evaluation, that is a different job, and INFRO does not do it.

The edits, one row each. Everything not listed here stays as it is.

 UnifyINFRO
Base URLhttps://api.unify.ai/v0https://api.infro.io/v1
KeyUNIFY_KEY, a Unify key.An INFRO gateway key; provider accounts as connections.
Model idgpt-5.6-terra@openai, llama-4-maverick@together-ai.openai/gpt-5.6-terra; the provider is whichever of your connections serves it, narrowed with routing.providers (together).
RoutingRouter strings for best-model routing.A route, optionally with a prompt classifier.

Before · Unify

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.unify.ai/v0",
    api_key=os.environ["UNIFY_KEY"],
)

client.chat.completions.create(
    model="gpt-5.6-terra@openai",
    messages=[{"role": "user", "content": "Hello"}],
)

After · INFRO

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.infro.io/v1",
    api_key=os.environ["INFRO_API_KEY"],
)

client.chat.completions.create(
    model="openai/gpt-5.6-terra",
    messages=[{"role": "user", "content": "Hello"}],
    extra_body={"routing": {"providers": {"allow": ["openai"]}}},
)

`model@provider` becomes two things

The model becomes INFRO's vendor/model id. The provider becomes a connection you hold, and routing.providers.allow when a request must use that one.

List your endpoints first

Write down every model@provider your code uses before the old API stops answering; each one needs a provider account on your side.

Failover

Two connections that serve the same model (OpenAI and Azure OpenAI, Anthropic and Vertex AI) give failover with no code change.

Unify's objects against INFRO's, from Unify's public documentation read on 19 October 2026 and checked against what the gateway implements. One row says INFRO has nothing.

UnifyWhat it does thereIn INFRO
Universal APIdifferentOne OpenAI-compatible endpoint across providers.One OpenAI-compatible endpoint across your own provider accounts.
`model@provider` endpointsdifferentModel and host in one id.vendor/model, plus routing.providers to choose the host.
Best-model routingdifferentRoutes each prompt to the best model.A route's prompt classifier sends classes of prompts to targets you choose. No benchmark-based choice.
Logging and evaluation SDKnothingThe current unifyai package.A request log and traces; no evaluation.

Our reading of Unify’s published documentation, not a trial or a benchmark. If we have read something wrong, tell us and we will fix the row.

What INFRO does not do

  • Provide model access without your own provider accounts.
  • Choose the best model per prompt from benchmarks.
  • Evaluate or train on your traffic.

What does not come across

  • Anything stored in your Unify console. Export it while you can sign in.

Questions

Is the Unify router shut down?
We found no official notice. Its site now presents a different product, its docs require sign-in and its repository was archived on 2026-10-16.
What replaces `model@provider`?
vendor/model as the model, and the provider as one of your connections, chosen with routing.providers when it matters.
Who pays for model usage on INFRO?
You, directly to your providers. INFRO is a subscription with 0% markup on model spend.

Move with us, or on your own

The migration program maps your Unify setup to INFRO connections, routes and budgets on a call and moves the first service with you. Your providers keep billing you directly; INFRO is a subscription and adds 0% to model spend.

Start 14-day trial Book a migration sessionRead the quickstart

Sources

Every statement on this page about Unify comes from one of these, read on 19 October 2026. Nothing is inferred from a trial, a benchmark, or a conversation.