Skip to main content
POST
/
bedrock
/
model
/
{modelId}
/
invoke-with-response-stream
Invoke model with streaming (Bedrock format)
curl --request POST \
  --url http://localhost:8080/bedrock/model/{modelId}/invoke-with-response-stream \
  --header 'Content-Type: application/json' \
  --data '
{
  "prompt": "<string>",
  "max_tokens": 123,
  "max_tokens_to_sample": 123,
  "temperature": 123,
  "top_p": 123,
  "top_k": 123,
  "stop": [
    "<string>"
  ],
  "stop_sequences": [
    "<string>"
  ],
  "messages": [
    {}
  ],
  "system": "<string>",
  "anthropic_version": "<string>"
}
'
{}

Path Parameters

​
modelId
string
required

Model ID (e.g., anthropic.claude-3-sonnet-20240229-v1:0)

Body

application/json

Raw model invocation request. The body format depends on the model provider. For Anthropic models, use Anthropic format. For other models, use their native format.

​
prompt
string

Text prompt to complete

​
max_tokens
integer
​
max_tokens_to_sample
integer

Anthropic-style max tokens

​
temperature
number
​
top_p
number
​
top_k
integer
​
stop
string[]
​
stop_sequences
string[]

Anthropic-style stop sequences

​
messages
object[]

For Claude 3 models

​
system

System prompt (string or array of strings)

​
anthropic_version
string

Response

Successful streaming response

AWS event stream format