Skip to main content
POST
/
v1
/
chat
/
completions
Smart Model Routing
curl --request POST \
  --url https://direct.evolink.ai/v1/chat/completions \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "evolink/auto",
  "messages": [
    {
      "role": "user",
      "content": "Introduce the history of artificial intelligence"
    }
  ],
  "temperature": 0.7,
  "top_p": 0.9,
  "top_k": 40,
  "stream": false
}
'
{
  "id": "chatcmpl-20260308112637503180122ABCD1234",
  "model": "gpt-5.4",
  "object": "chat.completion",
  "created": 1741428397,
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "The history of artificial intelligence dates back to the 1950s..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 15,
    "completion_tokens": 120,
    "total_tokens": 135
  }
}

​
Smart Model Routing

EvoLink Auto is an intelligent model routing feature that automatically selects a suitable AI model based on your request content, without manual model specification.

​
Key Benefits

  • Smart Matching: Automatically analyzes request content and selects a suitable model
  • Cost Optimization: Prioritizes cost-effective models while maintaining quality
  • Load Balancing: Automatically distributes requests across multiple models for improved stability
  • Transparency: Returns the actual model name used in the response for tracking and optimization

​
How It Works

The system selects the best-fit model from the model pool based on request complexity, length, and type.

​
Supported Models

EvoLink Auto intelligently routes between mainstream AI models including GPT-4, GPT-3.5, Claude, Gemini, and more.

​
Limitations

  • Not suitable for scenarios requiring specific model capabilities (e.g., GPT-4 vision features)
  • Does not guarantee the same model for every request

​
Use Cases

Ideal for scenarios where you’re unsure which model to use, or want the system to automatically optimize model selection.
Simply set the model parameter to evolink/auto, and the system will automatically select a suitable model for you.
BaseURL: The default BaseURL is https://direct.evolink.ai, which has better support for text models and long-lived connections. https://api.evolink.ai is the primary endpoint for multimodal services and serves as a fallback address for text models.

Authorizations

​
Authorization
string
header
required

All endpoints require Bearer Token authentication

Get your API Key:

Visit the API Key Management Page to get your API Key

Add the following to your request headers:

Authorization: Bearer YOUR_API_KEY

Body

application/json
​
model
enum<string>
default:evolink/auto
required

Use smart routing

Available options:
evolink/auto
Example:

"evolink/auto"

​
messages
object[]
required

List of conversation messages

Minimum array length: 1
Example:
[
  {
    "role": "user",
    "content": "Introduce the history of artificial intelligence"
  }
]
​
temperature
number

Sampling temperature, controls the randomness of the output

Notes:

  • Lower values (e.g., 0.2): More deterministic and focused output
  • Higher values (e.g., 1.5): More random and creative output
Required range: 0 <= x <= 2
Example:

0.7

​
top_p
number

Nucleus Sampling parameter

Notes:

  • Controls sampling from the top tokens by cumulative probability
  • For example, 0.9 means sampling from tokens whose cumulative probability reaches 90%
  • Default: 1.0 (considers all tokens)

Recommendation: Do not adjust both temperature and top_p simultaneously

Required range: 0 <= x <= 1
Example:

0.9

​
top_k
integer

Top-K sampling parameter

Notes:

  • For example, 10 means only the top 10 highest-probability tokens are considered during each sampling step
  • Smaller values make the output more focused
  • No limit by default
Required range: x >= 1
Example:

40

​
stream
boolean
default:false

Whether to return the response in streaming mode

  • true: Stream response, returning content in real-time chunks
  • false: Wait for the complete response before returning
Example:

false

Response

Request successful

​
id
string

Unique identifier for the chat completion

Example:

"chatcmpl-20260308112637503180122ABCD1234"

​
model
string

The model actually used

Example:

"gpt-5.4"

​
object
enum<string>

Response type

Available options:
chat.completion
Example:

"chat.completion"

​
created
integer

Creation timestamp

Example:

1741428397

​
choices
object[]

List of generated choices

​
usage
object

Token usage statistics