Smart routing

Smart routing lets you call a capability instead of a specific model. Routic automatically picks the best available model for that capability.

Why smart routing

Scenario	Canonical model name	Smart routing name
Model goes offline	Hard-coded model name → request fails	Routic switches to a replacement automatically
New model added	Must update code to use it	Automatically benefits from better models
Provider outage	All requests to that provider fail	Routic falls back to another provider's equivalent

In short: canonical names give you control; smart routing gives you resilience.

Available routing names

Routing name	Capability	Routing strategy	Currently includes
`auto/reasoning`	Complex reasoning, math, logic	Prefers the strongest reasoning model	deepseek-r1, qwq-32b, etc.
`auto/chat`	Everyday chat, summarization, content generation	Prefers the best cost-performance ratio	deepseek-v3 series, etc.

The models behind each routing name are updated continuously as the catalog grows — no code changes needed on your end.

Usage

Replace the model parameter with a routing name. The request format is identical:

curl -X POST "https://api.routic.ai/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxxxxxxx" \
  -d '{
    "model": "auto/reasoning",
    "messages": [
      { "role": "user", "content": "Analyze the time complexity of this code and suggest optimizations." }
    ]
  }'

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.routic.ai/v1",
    api_key="sk-xxxxxxxx",
)

# Reasoning task → automatically picks the strongest reasoning model
response = client.chat.completions.create(
    model="auto/reasoning",
    messages=[{"role": "user", "content": "Prove that √2 is irrational."}],
)

# General chat → automatically picks the best value model
response = client.chat.completions.create(
    model="auto/chat",
    messages=[{"role": "user", "content": "Write a meeting invitation email."}],
)

Routing name vs canonical model name

Dimension	Canonical model name	Smart routing name
Usage	`"model": "deepseek-r1"`	`"model": "auto/reasoning"`
Determinism	Always calls the specified model	Routic picks the best available
Best for	Reproducible results, model-specific dependencies	No model preference, want resilience and value
Failover	None — error if model unavailable	Automatic fallback to equivalent models
Response model	`response.model` returns the name you specified	`response.model` returns the actual model used

Best practices

Use smart routing in production — automatic failover keeps your service running when models go offline.
Use canonical names when you need reproducibility — e.g., automated tests, benchmarks.
Start with smart routing, switch to canonical if needed — both work interchangeably; no structural code changes required.
Check response.model to see which model was actually called — smart routing responses include the underlying model name.