Image-Text Chat Completion

Description

Create model responses based on given chat dialogue information, supporting image or video-related information in the conversation history. Images and videos cannot be provided simultaneously:

• If images are provided, up to 6 images can be accepted. Exceeding this limit will result in an error response.
• If a video is provided, only 1 video can be accepted. Exceeding this limit will result in an error response.
• Providing both images and videos will result in an error response.

Request URL

[POST] https://api.sensenova.cn/v1/llm/chat-completions

Request Header

No special headers required. Refer to API Authentication

Request Body

Note: For each request, the total number of tokens input by the user (i.e., the total number of tokens in all content) + the maximum number of tokens the user expects the model to generate (i.e., the value of max_new_tokens) must be <= the model's maximum context length (for different models' context length support, refer to Model List).

Name	Type	Required	Default	Optional Values	Description
model	string	Yes	-	Refer to Model List	Model ID
messages	object[]	Yes	-	-	Dialogue context input to the model, each object in the array represents a piece of chat context
max_new_tokens	int	No	1024	[1,16384]	Maximum number of tokens expected to be generated by the model (different models support different context lengths, so the maximum value varies, refer to Model List)
repetition_penalty	float	No	1.0	(0,2]	Repetition penalty coefficient, 1 means no penalty, greater than 1 tends to generate non-repetitive tokens, less than 1 tends to generate repetitive tokens, recommended range is [1,1.2]
stream	boolean	No	false	Enable: true Disable: false	Whether to use streaming transmission. If enabled, data will be returned as data-only SSE (server-sent events) with intermediate results and end with data: [DONE]
temperature	float	No	0.8	(0,2]	Temperature sampling parameter, values greater than 1 tend to generate more diverse responses, values less than 1 tend to generate more stable responses (supports up to six decimal places)
top_p	float	No	0.7	(0,1)	Nucleus sampling parameter, when decoding and generating tokens, sampling is done from the smallest token set whose cumulative probability is greater than or equal to top_p
user	string	No	-	-	External user ID, application developers can pass in the application system user ID for tracking

• messages section parameters:

Name	Type	Required	Default	Optional Values	Description
role	string	Yes	-	Enum values user assistant system	Role of the message author, enum values. Note that the last item in the array must be user
content	object[]	Yes	-	-	Content of the message

• content text section, can be plural, parameters are as follows:

Name	Type	Required	Default	Optional Values	Description
type	string	Yes	text	-	Message content type
text	string	Yes	-	-	Text content

• content image part, can be plural, choose one of the following formats, different image information can choose different parameter formats, parameters are as follows:

Name	Type	Required	Default	Optional Values	Description
type	string	Yes	image_url	-	Message content type
image_url	string	Yes	-	-	Image URL

Name	Type	Required	Default	Optional Values	Description
type	string	Yes	image_file_id	-	Message content type
image_file_id	string	Yes	-	-	File ID of the image uploaded via the file management interface

Name	Type	Required	Default	Optional Values	Description
type	string	Yes	image_base64	-	Message content type
image_base64	string	Yes	-	-	Base64 encoded content of the image

• content video part, can only be singular, choose one of the following formats, cannot coexist with image part, parameters are as follows:

Name	Type	Required	Default	Options	Description
type	string	Yes	video_url	-	Message content type
video_url	string	Yes	-	-	Video file URL

Name	Type	Required	Default	Options	Description
type	string	Yes	video_file_id	-	Message content type
video_file_id	string	Yes	-	-	File ID of the video uploaded via file management interface

Request Example

curl --request POST "https://api.sensenova.cn/v1/llm/chat-completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_TOKEN" \
  -d '{
        "model": "string",
        "messages": [
          {
            "role": "string", 
            "content": [
              {
                "type": "image_url", 
                "image_url": "string"
              },
              {
                "type": "text",
                "text": "string"
              }
            ]
          }
        ],
        "max_new_tokens": 1024, 
        "repetition_penalty": 1.0, 
        "stream": false,
        "temperature": 0.8,
        "top_p": 0.7,
        "user": "string"        
  }'

Response

Name	Type	Description
data	object	Generated content

• data section parameters:

Name	Type	Description
id	string	Message ID
choices	object[]	List of generated responses
usage	object	Token usage

• choices section parameters:

Name	Type	Description
message	string	Generated response content for non-streaming requests
finish_reason	string	Reason for stopping generation, enum values Stopped by end token: stop Stopped by reaching maximum generation length: length Stopped by triggering sensitive words: sensitive Stopped by reaching model context length limit: context
index	int	Index of the generated response
role	string	Role of the response message
delta	string	Generated response content for streaming requests

• usage section parameters:

Name	Type	Description
prompt_tokens	int	Number of tokens in user input
knowledge_tokens	int	Number of tokens for knowledge base content input to the model (non-zero only if knowledge base is used and content is retrieved)
completion_tokens	int	Number of tokens for generated message
total_tokens	int	Total number of tokens

Response Example

• Streaming

data:{"data":{"id":"123456789012345","choices":[{"delta":"This","finish_reason":"","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":1,"total_tokens":7}},"status":{"code":0, "message": "ok"}}

data:{"data":{"id":"123456789012345","choices":[{"delta":"is","finish_reason":"","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":2,"total_tokens":8}},"status":{"code":0, "message": "ok"}}

data:{"data":{"id":"123456789012345","choices":[{"delta":"a","finish_reason":"","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":3,"total_tokens":9}},"status":{"code":0, "message": "ok"}}

data:{"data":{"id":"123456789012345","choices":[{"delta":"test","finish_reason":"","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":4,"total_tokens":10}},"status":{"code":0, "message": "ok"}}

data:{"data":{"id":"123456789012345","choices":[{"delta":"!","finish_reason":"","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":5,"total_tokens":11}},"status":{"code":0, "message": "ok"}}

data:{"data":{"id":"123456789012345","choices":[{"delta":"","finish_reason":"stop","index": 0,"role":"assistant"}],"usage":{"prompt_tokens":6,"knowledge_tokens": 0,"completion_tokens":6,"total_tokens":13}},"status":{"code":0, "message": "ok"}}

data:[DONE]

• Non-Streaming

{
  "data": {
    "id": "4b44cd86cd2c000", 
    "choices": [ 
      {
        "message": "This is a test!",
        "finish_reason": "stop",
        "index": 0,
        "role": "string"       
      }
    ],
    "usage": {
      "prompt_tokens": 6,
      "knowledge_tokens": 0,
      "completion_tokens": 6,
      "total_tokens": 12
    }   
  }
}

Error Information

Refer to Error Codes