EvoLink Smart Router
Automatic model selection powered by EvoLink's proprietary routing agent.
Use evolink/auto as your default model ID. EvoLink's agent automatically picks a better-fit model for each request, so your team can spend less time on model tuning, reduce spend across mixed workloads, and maintain comparable quality.
Why teams trust EvoLink Smart Router
Self-built routing layer
Built around EvoLink's own routing layer instead of a fixed-model wrapper.
Cost-quality balance by request type
Different request types can be routed toward a more suitable quality and cost profile. The routing agent itself is free, and you only pay for the model actually used.
Actual model visibility in every response
You can inspect which model was used instead of treating routing like a blind black box.
How it works
Send evolink/auto as the model ID. EvoLink Smart Router evaluates each request and routes it to a better-fit model for the task.
Simple rewriting, editing, or polishing requests can route to a faster, lower-cost model.
Requests that need stronger reasoning or structured analysis can route to a stronger model.
When one agent handles rewriting, analysis, and other requests with different complexity, one default model ID can route them automatically.
{
"model": "gpt-5.4",
"usage": {
"prompt_tokens": 7,
"completion_tokens": 32,
"total_tokens": 39
}
}The model field in the response shows which model was actually used.
Start with one model ID
Replace the model field with evolink/auto and keep the rest of your request shape the same.
- No workflow rewrite
- No manual model switching
- Just set model to evolink/auto
curl https://api.evolink.ai/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "evolink/auto",
"messages": [
{
"role": "user",
"content": "Review this draft and rewrite it in a clearer tone."
}
]
}'What changes after switching to EvoLink Smart Router
Without EvoLink Smart Router
- Manual model choice for each new task
- One familiar model gets overused across all requests
- Mixed workloads push spend up
- Routing logic lives inside application code
With EvoLink Smart Router
- One default model ID across mixed requests
- Automatic right-fit model selection by request
- More balanced spend by default
- Less routing logic to maintain
Where EvoLink Smart Router fits agent workflows
Use one default model ID when your agent workflow moves across summarization, rewriting, analysis, and other mixed task types.
Agent workflows across mixed task types
Keep one router in front of rewriting, reasoning, and generation instead of tuning models at every step.
Document agents with variable query complexity
Questions against the same corpus can vary in difficulty, so one fixed model is rarely the best default.
Automation pipelines with summarization, rewriting, and analysis
Use one entry point across pipeline steps that need different cost and reasoning profiles.
General-purpose default model for new builds
Start with evolink/auto before you have enough workload data to hand-tune routing logic in application code.
FAQ
What is EvoLink Smart Router for agent workflows?
EvoLink Smart Router is EvoLink's self-built routing agent that automatically selects a better-fit model for mixed agent workloads.
When should I use EvoLink Smart Router instead of manually choosing models?
Use it when your workflow handles mixed tasks and you do not want to choose a model every time.
When should I use a fixed model instead of automatic model selection?
Use a fixed model when you need strict control for benchmarking, compliance, or controlled comparisons.
Does EvoLink Smart Router charge an extra routing fee?
No. The routing agent itself is free, and requests are billed only against the model that was actually used.
Can I use evolink/auto as the default model ID for mixed workloads?
Yes. Replace the model field with evolink/auto and keep the rest of the request shape the same.
How do I know which model was actually used for a request?
The actual model used is returned in the response.
Is EvoLink Smart Router a black box?
Routing logic is abstracted, but model transparency is exposed in the response.