2024/05/30 - Amazon Bedrock Runtime - 2 new api methods
Changes This release adds Converse and ConverseStream APIs to Bedrock Runtime
Sends messages to the specified Amazon Bedrock model. Converse provides a consistent interface that works with all models that support messages. This allows you to write code once and use it with different models. Should a model have unique inference parameters, you can also pass those unique parameters to the model. For more information, see Run inference in the Bedrock User Guide.
This operation requires permission for the bedrock:InvokeModel action.
See also: AWS API Documentation
Request Syntax
client.converse(
modelId='string',
messages=[
{
'role': 'user'|'assistant',
'content': [
{
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
},
'toolUse': {
'toolUseId': 'string',
'name': 'string',
'input': {...}|[...]|123|123.4|'string'|True|None
},
'toolResult': {
'toolUseId': 'string',
'content': [
{
'json': {...}|[...]|123|123.4|'string'|True|None,
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
}
},
],
'status': 'success'|'error'
}
},
]
},
],
system=[
{
'text': 'string'
},
],
inferenceConfig={
'maxTokens': 123,
'temperature': ...,
'topP': ...,
'stopSequences': [
'string',
]
},
toolConfig={
'tools': [
{
'toolSpec': {
'name': 'string',
'description': 'string',
'inputSchema': {
'json': {...}|[...]|123|123.4|'string'|True|None
}
}
},
],
'toolChoice': {
'auto': {}
,
'any': {}
,
'tool': {
'name': 'string'
}
}
},
additionalModelRequestFields={...}|[...]|123|123.4|'string'|True|None,
additionalModelResponseFieldPaths=[
'string',
]
)
string
[REQUIRED]
The identifier for the model that you want to call.
The modelId to provide depends on the type of model that you use:
If you use a base model, specify the model ID or its ARN. For a list of model IDs for base models, see Amazon Bedrock base model IDs (on-demand throughput) in the Amazon Bedrock User Guide.
If you use a provisioned model, specify the ARN of the Provisioned Throughput. For more information, see Run inference using a Provisioned Throughput in the Amazon Bedrock User Guide.
If you use a custom model, first purchase Provisioned Throughput for it. Then specify the ARN of the resulting provisioned model. For more information, see Use a custom model in Amazon Bedrock in the Amazon Bedrock User Guide.
list
[REQUIRED]
The messages that you want to send to the model.
(dict) --
A message in the Message field. Use to send a message in a call to Converse.
role (string) -- [REQUIRED]
The role that the message plays in the message.
content (list) -- [REQUIRED]
The message content.
(dict) --
A block of content for a message.
text (string) --
Text to include in the message.
image (dict) --
Image to include in the message.
format (string) -- [REQUIRED]
The format of the image.
source (dict) -- [REQUIRED]
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse (dict) --
Information about a tool use request from a model.
toolUseId (string) -- [REQUIRED]
The ID for the tool request.
name (string) -- [REQUIRED]
The name of the tool that the model wants to use.
input (:ref:`document<document>`) -- [REQUIRED]
The input to pass to the tool.
toolResult (dict) --
The result for a tool request that a model makes.
toolUseId (string) -- [REQUIRED]
The ID of the tool request that this is the result for.
content (list) -- [REQUIRED]
The content for tool result content block.
(dict) --
The tool result content block.
json (:ref:`document<document>`) --
A tool result that is JSON format data.
text (string) --
A tool result that is text.
image (dict) --
A tool result that is an image.
format (string) -- [REQUIRED]
The format of the image.
source (dict) -- [REQUIRED]
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
status (string) --
The status for the tool result content block.
list
A system prompt to pass to the model.
(dict) --
A system content block
text (string) --
A system prompt for the model.
dict
Inference parameters to pass to the model. Converse supports a base set of inference parameters. If you need to pass additional parameters that the model supports, use the additionalModelRequestFields request field.
maxTokens (integer) --
The maximum number of tokens to allow in the generated response. The default value is the maximum allowed value for the model that you are using. For more information, see Inference parameters for foundatio{ "messages": [ { "role": "user", "content": [ { "text": "what's the weather in Queens, NY and Austin, TX?" } ] }, { "role": "assistant", "content": [ { "toolUse": { "toolUseId": "1", "name": "get_weather", "input": { "city": "Queens", "state": "NY" } } }, { "toolUse": { "toolUseId": "2", "name": "get_weather", "input": { "city": "Austin", "state": "TX" } } } ] }, { "role": "user", "content": [ { "toolResult": { "toolUseId": "2", "content": [ { "json": { "weather": "40" } } ] } }, { "text": "..." }, { "toolResult": { "toolUseId": "1", "content": [ { "text": "result text" } ] } } ] } ], "toolConfig": { "tools": [ { "name": "get_weather", "description": "Get weather", "inputSchema": { "type": "object", "properties": { "city": { "type": "string", "description": "City of location" }, "state": { "type": "string", "description": "State of location" } }, "required": ["city", "state"] } } ] } } n models.
temperature (float) --
The likelihood of the model selecting higher-probability options while generating a response. A lower value makes the model more likely to choose higher-probability options, while a higher value makes the model more likely to choose lower-probability options.
The default value is the default value for the model that you are using. For more information, see Inference parameters for foundation models.
topP (float) --
The percentage of most-likely candidates that the model considers for the next token. For example, if you choose a value of 0.8 for topP, the model selects from the top 80% of the probability distribution of tokens that could be next in the sequence.
The default value is the default value for the model that you are using. For more information, see Inference parameters for foundation models.
stopSequences (list) --
A list of stop sequences. A stop sequence is a sequence of characters that causes the model to stop generating the response.
(string) --
dict
Configuration information for the tools that the model can use when generating a response.
tools (list) -- [REQUIRED]
An array of tools that you want to pass to a model.
(dict) --
Information about a tool that you can use with the Converse API.
toolSpec (dict) --
The specfication for the tool.
name (string) -- [REQUIRED]
The name for the tool.
description (string) --
The description for the tool.
inputSchema (dict) -- [REQUIRED]
The input schema for the tool in JSON format.
json (:ref:`document<document>`) --
The JSON schema for the tool. For more information, see JSON Schema Reference.
toolChoice (dict) --
If supported by model, forces the model to request a tool.
auto (dict) --
The Model automatically decides if a tool should be called or to whether to generate text instead.
any (dict) --
The model must request at least one tool (no text is generated).
tool (dict) --
The Model must request the specified tool.
name (string) -- [REQUIRED]
The name of the tool that the model must request.
:ref:`document<document>`
Additional inference parameters that the model supports, beyond the base set of inference parameters that Converse supports in the inferenceConfig field. For more information, see Model parameters.
list
Additional model parameters field paths to return in the response. Converse returns the requested fields as a JSON Pointer object in the additionalModelResultFields field. The following is example JSON for additionalModelResponseFieldPaths.
[ "/stop_sequence" ]
For information about the JSON Pointer syntax, see the Internet Engineering Task Force (IETF) documentation.
Converse rejects an empty JSON Pointer or incorrectly structured JSON Pointer with a 400 error code. if the JSON Pointer is valid, but the requested field is not in the model response, it is ignored by Converse.
(string) --
dict
Response Syntax
{
'output': {
'message': {
'role': 'user'|'assistant',
'content': [
{
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
},
'toolUse': {
'toolUseId': 'string',
'name': 'string',
'input': {...}|[...]|123|123.4|'string'|True|None
},
'toolResult': {
'toolUseId': 'string',
'content': [
{
'json': {...}|[...]|123|123.4|'string'|True|None,
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
}
},
],
'status': 'success'|'error'
}
},
]
}
},
'stopReason': 'end_turn'|'tool_use'|'max_tokens'|'stop_sequence'|'content_filtered',
'usage': {
'inputTokens': 123,
'outputTokens': 123,
'totalTokens': 123
},
'metrics': {
'latencyMs': 123
},
'additionalModelResponseFields': {...}|[...]|123|123.4|'string'|True|None
}
Response Structure
(dict) --
output (dict) --
The result from the call to Converse.
message (dict) --
The message that the model generates.
role (string) --
The role that the message plays in the message.
content (list) --
The message content.
(dict) --
A block of content for a message.
text (string) --
Text to include in the message.
image (dict) --
Image to include in the message.
format (string) --
The format of the image.
source (dict) --
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse (dict) --
Information about a tool use request from a model.
toolUseId (string) --
The ID for the tool request.
name (string) --
The name of the tool that the model wants to use.
input (:ref:`document<document>`) --
The input to pass to the tool.
toolResult (dict) --
The result for a tool request that a model makes.
toolUseId (string) --
The ID of the tool request that this is the result for.
content (list) --
The content for tool result content block.
(dict) --
The tool result content block.
json (:ref:`document<document>`) --
A tool result that is JSON format data.
text (string) --
A tool result that is text.
image (dict) --
A tool result that is an image.
format (string) --
The format of the image.
source (dict) --
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
status (string) --
The status for the tool result content block.
stopReason (string) --
The reason why the model stopped generating output.
usage (dict) --
The total number of tokens used in the call to Converse. The total includes the tokens input to the model and the tokens generated by the model.
inputTokens (integer) --
The number of tokens sent in the request to the model.
outputTokens (integer) --
The number of tokens that the model generated for the request.
totalTokens (integer) --
The total of input tokens and tokens generated by the model.
metrics (dict) --
Metrics for the call to Converse.
latencyMs (integer) --
The latency of the call to Converse, in milliseconds.
additionalModelResponseFields (:ref:`document<document>`) --
Additional fields in the response that are unique to the model.
Sends messages to the specified Amazon Bedrock model and returns the response in a stream. ConverseStream provides a consistent API that works with all Amazon Bedrock models that support messages. This allows you to write code once and use it with different models. Should a model have unique inference parameters, you can also pass those unique parameters to the model. For more information, see Run inference in the Bedrock User Guide.
To find out if a model supports streaming, call GetFoundationModel and check the responseStreamingSupported field in the response.
For example code, see Invoke model with streaming code example in the Amazon Bedrock User Guide.
This operation requires permission for the bedrock:InvokeModelWithResponseStream action.
See also: AWS API Documentation
Request Syntax
client.converse_stream(
modelId='string',
messages=[
{
'role': 'user'|'assistant',
'content': [
{
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
},
'toolUse': {
'toolUseId': 'string',
'name': 'string',
'input': {...}|[...]|123|123.4|'string'|True|None
},
'toolResult': {
'toolUseId': 'string',
'content': [
{
'json': {...}|[...]|123|123.4|'string'|True|None,
'text': 'string',
'image': {
'format': 'png'|'jpeg'|'gif'|'webp',
'source': {
'bytes': b'bytes'
}
}
},
],
'status': 'success'|'error'
}
},
]
},
],
system=[
{
'text': 'string'
},
],
inferenceConfig={
'maxTokens': 123,
'temperature': ...,
'topP': ...,
'stopSequences': [
'string',
]
},
toolConfig={
'tools': [
{
'toolSpec': {
'name': 'string',
'description': 'string',
'inputSchema': {
'json': {...}|[...]|123|123.4|'string'|True|None
}
}
},
],
'toolChoice': {
'auto': {}
,
'any': {}
,
'tool': {
'name': 'string'
}
}
},
additionalModelRequestFields={...}|[...]|123|123.4|'string'|True|None,
additionalModelResponseFieldPaths=[
'string',
]
)
string
[REQUIRED]
The ID for the model.
The modelId to provide depends on the type of model that you use:
If you use a base model, specify the model ID or its ARN. For a list of model IDs for base models, see Amazon Bedrock base model IDs (on-demand throughput) in the Amazon Bedrock User Guide.
If you use a provisioned model, specify the ARN of the Provisioned Throughput. For more information, see Run inference using a Provisioned Throughput in the Amazon Bedrock User Guide.
If you use a custom model, first purchase Provisioned Throughput for it. Then specify the ARN of the resulting provisioned model. For more information, see Use a custom model in Amazon Bedrock in the Amazon Bedrock User Guide.
list
[REQUIRED]
The messages that you want to send to the model.
(dict) --
A message in the Message field. Use to send a message in a call to Converse.
role (string) -- [REQUIRED]
The role that the message plays in the message.
content (list) -- [REQUIRED]
The message content.
(dict) --
A block of content for a message.
text (string) --
Text to include in the message.
image (dict) --
Image to include in the message.
format (string) -- [REQUIRED]
The format of the image.
source (dict) -- [REQUIRED]
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse (dict) --
Information about a tool use request from a model.
toolUseId (string) -- [REQUIRED]
The ID for the tool request.
name (string) -- [REQUIRED]
The name of the tool that the model wants to use.
input (:ref:`document<document>`) -- [REQUIRED]
The input to pass to the tool.
toolResult (dict) --
The result for a tool request that a model makes.
toolUseId (string) -- [REQUIRED]
The ID of the tool request that this is the result for.
content (list) -- [REQUIRED]
The content for tool result content block.
(dict) --
The tool result content block.
json (:ref:`document<document>`) --
A tool result that is JSON format data.
text (string) --
A tool result that is text.
image (dict) --
A tool result that is an image.
format (string) -- [REQUIRED]
The format of the image.
source (dict) -- [REQUIRED]
The source for the image.
bytes (bytes) --
The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
status (string) --
The status for the tool result content block.
list
A system prompt to send to the model.
(dict) --
A system content block
text (string) --
A system prompt for the model.
dict
Inference parameters to pass to the model. ConverseStream supports a base set of inference parameters. If you need to pass additional parameters that the model supports, use the additionalModelRequestFields request field.
maxTokens (integer) --
The maximum number of tokens to allow in the generated response. The default value is the maximum allowed value for the model that you are using. For more information, see Inference parameters for foundatio{ "messages": [ { "role": "user", "content": [ { "text": "what's the weather in Queens, NY and Austin, TX?" } ] }, { "role": "assistant", "content": [ { "toolUse": { "toolUseId": "1", "name": "get_weather", "input": { "city": "Queens", "state": "NY" } } }, { "toolUse": { "toolUseId": "2", "name": "get_weather", "input": { "city": "Austin", "state": "TX" } } } ] }, { "role": "user", "content": [ { "toolResult": { "toolUseId": "2", "content": [ { "json": { "weather": "40" } } ] } }, { "text": "..." }, { "toolResult": { "toolUseId": "1", "content": [ { "text": "result text" } ] } } ] } ], "toolConfig": { "tools": [ { "name": "get_weather", "description": "Get weather", "inputSchema": { "type": "object", "properties": { "city": { "type": "string", "description": "City of location" }, "state": { "type": "string", "description": "State of location" } }, "required": ["city", "state"] } } ] } } n models.
temperature (float) --
The likelihood of the model selecting higher-probability options while generating a response. A lower value makes the model more likely to choose higher-probability options, while a higher value makes the model more likely to choose lower-probability options.
The default value is the default value for the model that you are using. For more information, see Inference parameters for foundation models.
topP (float) --
The percentage of most-likely candidates that the model considers for the next token. For example, if you choose a value of 0.8 for topP, the model selects from the top 80% of the probability distribution of tokens that could be next in the sequence.
The default value is the default value for the model that you are using. For more information, see Inference parameters for foundation models.
stopSequences (list) --
A list of stop sequences. A stop sequence is a sequence of characters that causes the model to stop generating the response.
(string) --
dict
Configuration information for the tools that the model can use when generating a response.
tools (list) -- [REQUIRED]
An array of tools that you want to pass to a model.
(dict) --
Information about a tool that you can use with the Converse API.
toolSpec (dict) --
The specfication for the tool.
name (string) -- [REQUIRED]
The name for the tool.
description (string) --
The description for the tool.
inputSchema (dict) -- [REQUIRED]
The input schema for the tool in JSON format.
json (:ref:`document<document>`) --
The JSON schema for the tool. For more information, see JSON Schema Reference.
toolChoice (dict) --
If supported by model, forces the model to request a tool.
auto (:class:`.EventStream`) --
The Model automatically decides if a tool should be called or to whether to generate text instead.
any (:class:`.EventStream`) --
The model must request at least one tool (no text is generated).
tool (dict) --
The Model must request the specified tool.
name (string) -- [REQUIRED]
The name of the tool that the model must request.
:ref:`document<document>`
Additional inference parameters that the model supports, beyond the base set of inference parameters that ConverseStream supports in the inferenceConfig field.
list
Additional model parameters field paths to return in the response. ConverseStream returns the requested fields as a JSON Pointer object in the additionalModelResultFields field. The following is example JSON for additionalModelResponseFieldPaths.
[ "/stop_sequence" ]
For information about the JSON Pointer syntax, see the Internet Engineering Task Force (IETF) documentation.
ConverseStream rejects an empty JSON Pointer or incorrectly structured JSON Pointer with a 400 error code. if the JSON Pointer is valid, but the requested field is not in the model response, it is ignored by ConverseStream.
(string) --
dict
The response of this operation contains an :class:`.EventStream` member. When iterated the :class:`.EventStream` will yield events based on the structure below, where only one of the top level keys will be present for any given event.
Response Syntax
{
'stream': EventStream({
'messageStart': {
'role': 'user'|'assistant'
},
'contentBlockStart': {
'start': {
'toolUse': {
'toolUseId': 'string',
'name': 'string'
}
},
'contentBlockIndex': 123
},
'contentBlockDelta': {
'delta': {
'text': 'string',
'toolUse': {
'input': 'string'
}
},
'contentBlockIndex': 123
},
'contentBlockStop': {
'contentBlockIndex': 123
},
'messageStop': {
'stopReason': 'end_turn'|'tool_use'|'max_tokens'|'stop_sequence'|'content_filtered',
'additionalModelResponseFields': {...}|[...]|123|123.4|'string'|True|None
},
'metadata': {
'usage': {
'inputTokens': 123,
'outputTokens': 123,
'totalTokens': 123
},
'metrics': {
'latencyMs': 123
}
},
'internalServerException': {
'message': 'string'
},
'modelStreamErrorException': {
'message': 'string',
'originalStatusCode': 123,
'originalMessage': 'string'
},
'validationException': {
'message': 'string'
},
'throttlingException': {
'message': 'string'
}
})
}
Response Structure
(dict) --
stream (:class:`.EventStream`) --
The output stream that the model generated.
messageStart (dict) --
Message start information.
role (string) --
The role for the message.
contentBlockStart (dict) --
Start information for a content block.
start (dict) --
Start information about a content block start event.
toolUse (dict) --
Information about a tool that the model is requesting to use.
toolUseId (string) --
The ID for the tool request.
name (string) --
The name of the tool that the model is requesting to use.
contentBlockIndex (integer) --
The index for a content block start event.
contentBlockDelta (dict) --
The messages output content block delta.
delta (dict) --
The delta for a content block delta event.
text (string) --
The content text.
toolUse (dict) --
Information about a tool that the model is requesting to use.
input (string) --
The input for a requested tool.
contentBlockIndex (integer) --
The block index for a content block delta event.
contentBlockStop (dict) --
Stop information for a content block.
contentBlockIndex (integer) --
The index for a content block.
messageStop (dict) --
Message stop information.
stopReason (string) --
The reason why the model stopped generating output.
additionalModelResponseFields (:ref:`document<document>`) --
The additional model response fields.
metadata (dict) --
Metadata for the converse output stream.
usage (dict) --
Usage information for the conversation stream event.
inputTokens (integer) --
The number of tokens sent in the request to the model.
outputTokens (integer) --
The number of tokens that the model generated for the request.
totalTokens (integer) --
The total of input tokens and tokens generated by the model.
metrics (dict) --
The metrics for the conversation stream metadata event.
latencyMs (integer) --
The latency for the streaming request, in milliseconds.
internalServerException (dict) --
An internal server error occurred. Retry your request.
message (string) --
modelStreamErrorException (dict) --
A streaming error occurred. Retry your request.
message (string) --
originalStatusCode (integer) --
The original status code.
originalMessage (string) --
The original message.
validationException (dict) --
Input validation failed. Check your request parameters and retry the request.
message (string) --
throttlingException (dict) --
The number of requests exceeds the limit. Resubmit your request later.
message (string) --