Too many model choices
AI applications rarely send one type of request. A single workflow may include classification, rewriting, summarization, structured extraction, reasoning, and agent planning.
NewEvoLink Smart Router
AI Model Routing for Text Requests and Agent Workflows
Use evolink/auto as one model ID to route supported text and agent requests through EvoLink's unified API, with the selected model visible in response.model so you can compare results and control cost-quality tradeoffs as your workload changes.
Unified APIOne model IDresponse.model visibleDocs, pricing, and fixed-model paths
Smart Router in action
OpenAI-compatible request
Routing Result Preview
/v1/chat/completionsInput model
evolink/auto
Selected profile
Standard / Reasoning
Routed model
actual-routed-model
Visible in
response.model
The routed model stays inspectable in the API response.
Input
evolink/auto
Route
Profile
Return
model
Why AI apps and agent workflows need model routing
Hardcoded model costs
Hardcoding every request to one frontier model can increase cost. Building your own model-switching logic inside application code can add maintenance overhead.
Manual routing overhead
EvoLink Smart Router gives supported text and agent workflows one routing entry point while keeping the actual routed model visible in the response.
How evolink/auto routes supported requests
Automatic by default. Observable when you need control.
1. Request
Your application sends a normal OpenAI-compatible request with model set to evolink/auto.
2. Task Analysis
EvoLink evaluates the task type and complexity for supported text requests.
3. Profile Selection
The router maps the request to a routing profile such as Fast, Standard, or Reasoning.
4. Model Hit
The request is sent to a suitable candidate model.
5. Visible Response
The actual routed model is returned in the response, so your team can log, compare, and evaluate routing behavior.
Call evolink/auto and inspect response.model
EvoLink Smart Router is a real, callable API capability. Set the model field to evolink/auto in a standard OpenAI-compatible chat completion request.
cURL request
curl https://api.evolink.ai/v1/chat/completions \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "evolink/auto",
"messages": [
{
"role": "user",
"content": "Classify this support ticket and suggest the next action."
}
]
}'Response
{
"model": "actual-routed-model",
"choices": [
{
"message": {
"role": "assistant",
"content": "..."
}
}
]
}Model ID
evolink/auto
Endpoint
/v1/chat/completions
Format
OpenAI-compatible
Visibility
response.model
The routed model is visible in the response, so you can track which model was selected for each request.
Read Smart Router QuickstartDecide where Smart Router fits your workflow
Use this panel as a compact decision tool: profile the work, choose routing or fixed models, check real use cases, then compare approaches only if needed.
Use routing profiles to understand the task layer, not a fixed model list.
| Routing Profile | Best for | Example tasks |
|---|---|---|
| Fast | Simple, high-volume tasks | Rewriting, classification, formatting |
| Standard | General text processing | Summarization, structured extraction, support analysis |
| Reasoning | Complex reasoning and planning | Multi-step analysis, agent planning, decision support |
| Coding / Agentic Coding | Coding-heavy workflows, if supported | Code review, debugging, refactor planning |
Fast
Simple, high-volume tasks
Rewriting, classification, formatting
Standard
General text processing
Summarization, structured extraction, support analysis
Reasoning
Complex reasoning and planning
Multi-step analysis, agent planning, decision support
Coding / Agentic Coding
Coding-heavy workflows, if supported
Code review, debugging, refactor planning
The exact candidate model pool may change as upstream availability, pricing, and performance change. EvoLink returns the actual routed model in the response so teams can log and evaluate routing behavior over time.
Exact availability may change. Check Docs or inspect response.model for actual routed results.
Use this table to decide whether routing or a fixed model fits your workflow.
| Scenario | Smart Router | Fixed Model |
|---|---|---|
| Mixed agent tasks | Good fit | Can create manual switching overhead |
| Early product exploration | Good fit | Useful after baseline is known |
| Simple rewriting / extraction | Good fit | Fine if cost is already acceptable |
| Strict benchmark | Not ideal | Better fit |
| Deterministic QA flow | Test carefully | Better fit |
| Required model-specific feature | Not ideal | Required |
Smart Router is useful when your workload contains mixed text tasks and you want a default routing entry point. Fixed models are better when you need strict benchmarking, deterministic QA, or a model-specific capability.
Use cases show where routing creates leverage without forcing every path to stay routed.
Agent workflows with mixed task complexity
Agents often move between rewriting, extraction, summarization, and reasoning. Smart Router can reduce the need to manually hardcode a model for every task type.
Document and knowledge-base agents
User questions may range from simple lookup to multi-step analysis. Routing can help teams start with one default model ID while observing which models are selected.
Internal automation pipelines
Workflows that combine formatting, classification, reporting, and analysis can use routing as a default before pinning fixed models for stable paths.
Early-stage AI products
When teams do not yet have enough workload data to choose the best fixed model per feature, evolink/auto can be used as a starting point.
Use this table to place EvoLink among common model routing approaches.
| Approach | Best for | How EvoLink differs |
|---|---|---|
| OpenRouter Auto | OpenRouter users who want automatic model choice | EvoLink Smart Router is part of EvoLink's unified model delivery layer |
| NotDiamond | Teams optimizing model selection across candidate models | EvoLink combines routing with unified API access, billing, usage, and model delivery |
| Portkey Routing | Teams needing gateway rules, fallbacks, guardrails, and policies | EvoLink focuses on model access and routing for supported text / agent workflows |
| Self-built router | Teams with engineering resources to maintain routing logic | EvoLink reduces the need to maintain a routing layer in application code |
EvoLink Smart Router is not a standalone router product. It is a routing capability inside EvoLink's unified model delivery infrastructure.
Read model routing alternativesValidate Smart Router in 3 steps
Keep production adoption staged: confirm the call, observe the routed result, then compare before scaling.
1. Run evolink/auto
Make sure evolink/auto works in your account before adding production traffic.
1. Run evolink/auto
Make sure evolink/auto works in your account before adding production traffic.
- Confirm evolink/auto works in your account
2. Inspect response.model
Log the fields that let your team understand routing behavior over time.
2. Inspect response.model
Log the fields that let your team understand routing behavior over time.
- Log response.model for each request
- Log latency and token usage
- Log feature name or route name
3. Compare before scaling
Validate routing against a fixed-model baseline, then decide what should stay routed.
3. Compare before scaling
Validate routing against a fixed-model baseline, then decide what should stay routed.
- Compare Smart Router with one fixed-model baseline
- Decide which workflows can stay routed
- Pin fixed models for strict QA or deterministic flows
- Review pricing and usage before scaling traffic
Smart Router is not magic automation. It is a routing strategy that should be tested, observed, and gradually adopted.
Start validating evolink/auto
Use these paths to move from routing proof to docs, pricing review, API keys, or fixed model selection.
Frequently asked questions about AI model routing
What is AI model routing?
AI model routing is the process of sending a request to a suitable model based on task type, complexity, cost, latency, or other routing signals.
How do I use evolink/auto?
Set the model field to evolink/auto in a supported chat completion request through EvoLink's unified API.
Can I see which model was selected?
Yes. The actual routed model is visible in the response, so teams can log and evaluate routing behavior.
Is Smart Router always cheaper?
No. Smart Router is not a guarantee that every request will be cheaper. Its value is helping mixed workloads control cost-quality tradeoffs and avoid overusing one large model by default.
When should I use a fixed model instead?
Use a fixed model for strict benchmarks, deterministic QA, model-specific features, or workflows where you already know the best model.
Does Smart Router work for image or video models?
EvoLink Smart Router is currently positioned for supported text and agent workflows. For image and video generation, use fixed model IDs and compare pricing, latency, and output requirements directly.
How is EvoLink different from OpenRouter Auto or other routers?
EvoLink Smart Router is a routing capability inside EvoLink's unified model delivery infrastructure. It is designed to work alongside EvoLink's unified API access, billing, usage, and model delivery paths.
How should I test Smart Router before production?
Start with low-risk traffic, log response.model, latency, token usage, and feature name, then compare results against one fixed-model baseline before scaling.