Tool Calling
OpenAI Format - Responses
- Generic Responses API reference for all compatible models
- Server-side context storage: use previous_response_id for multi-turn conversations without manually passing history
- Supports streaming and non-streaming responses
- Multimodal input: supports text + image + file mixed input
- Tool calling: supports Function Calling and built-in tools
- Structured output: supports JSON Schema format
POST
Authorizations
All APIs require Bearer Token authentication
Add to request header:
Authorization: Bearer YOUR_API_KEY
Body
application/json
Model name
Example:
"gpt-5.4"
Text input or message list for generating a response
Whether to stream the response
Sampling temperature (0-2)
Required range:
0 <= x <= 2Maximum number of output tokens to generate
Nucleus sampling parameter (0-1)
Required range:
0 <= x <= 1ID of the previous response, used for multi-turn conversations
System-level instructions
Available tools list
Text output format configuration
Response
Response API success
Unique response ID
Example:
"resp_abc123"
Object type
Available options:
response Creation timestamp
Example:
1757165031
Model used
The status of the response
Available options:
completed, failed, in_progress, incomplete Example:
"completed"
List of output items
Token usage statistics