Class: AWS.BedrockRuntime
- Inherits:
-
AWS.Service
- Object
- AWS.Service
- AWS.BedrockRuntime
- Identifier:
- bedrockruntime
- API Version:
- 2023-09-30
- Defined in:
- (unknown)
Overview
Constructs a service interface object. Each API operation is exposed as a function on service.
Service Description
Describes the API operations for running inference using Amazon Bedrock models.
Sending a Request Using BedrockRuntime
var bedrockruntime = new AWS.BedrockRuntime();
bedrockruntime.converse(params, function (err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Locking the API Version
In order to ensure that the BedrockRuntime object uses this specific API, you can
construct the object by passing the apiVersion
option to the constructor:
var bedrockruntime = new AWS.BedrockRuntime({apiVersion: '2023-09-30'});
You can also set the API version globally in AWS.config.apiVersions
using
the bedrockruntime service identifier:
AWS.config.apiVersions = {
bedrockruntime: '2023-09-30',
// other service API versions
};
var bedrockruntime = new AWS.BedrockRuntime();
Version:
-
2023-09-30
Constructor Summary collapse
-
new AWS.BedrockRuntime(options = {}) ⇒ Object
constructor
Constructs a service object.
Property Summary collapse
-
endpoint ⇒ AWS.Endpoint
readwrite
An Endpoint object representing the endpoint URL for service requests.
Properties inherited from AWS.Service
Method Summary collapse
-
converse(params = {}, callback) ⇒ AWS.Request
Sends messages to the specified Amazon Bedrock model.
-
converseStream(params = {}, callback) ⇒ AWS.Request
Sends messages to the specified Amazon Bedrock model and returns the response in a stream.
-
invokeModel(params = {}, callback) ⇒ AWS.Request
Invokes the specified Amazon Bedrock model to run inference using the prompt and inference parameters provided in the request body.
-
invokeModelWithResponseStream(params = {}, callback) ⇒ AWS.Request
Invoke the specified Amazon Bedrock model to run inference using the prompt and inference parameters provided in the request body.
-
waitFor(state, params = {}, callback) ⇒ AWS.Request
Waits for a given BedrockRuntime resource.
Methods inherited from AWS.Service
makeRequest, makeUnauthenticatedRequest, setupRequestListeners, defineService
Constructor Details
new AWS.BedrockRuntime(options = {}) ⇒ Object
Constructs a service object. This object has one method for each API operation.
Examples:
Constructing a BedrockRuntime object
var bedrockruntime = new AWS.BedrockRuntime({apiVersion: '2023-09-30'});
Options Hash (options):
-
params
(map)
—
An optional map of parameters to bind to every request sent by this service object. For more information on bound parameters, see "Working with Services" in the Getting Started Guide.
-
endpoint
(String|AWS.Endpoint)
—
The endpoint URI to send requests to. The default endpoint is built from the configured
region
. The endpoint should be a string like'https://{service}.{region}.amazonaws.com'
or an Endpoint object. -
accessKeyId
(String)
—
your AWS access key ID.
-
secretAccessKey
(String)
—
your AWS secret access key.
-
sessionToken
(AWS.Credentials)
—
the optional AWS session token to sign requests with.
-
credentials
(AWS.Credentials)
—
the AWS credentials to sign requests with. You can either specify this object, or specify the accessKeyId and secretAccessKey options directly.
-
credentialProvider
(AWS.CredentialProviderChain)
—
the provider chain used to resolve credentials if no static
credentials
property is set. -
region
(String)
—
the region to send service requests to. See AWS.BedrockRuntime.region for more information.
-
maxRetries
(Integer)
—
the maximum amount of retries to attempt with a request. See AWS.BedrockRuntime.maxRetries for more information.
-
maxRedirects
(Integer)
—
the maximum amount of redirects to follow with a request. See AWS.BedrockRuntime.maxRedirects for more information.
-
sslEnabled
(Boolean)
—
whether to enable SSL for requests.
-
paramValidation
(Boolean|map)
—
whether input parameters should be validated against the operation description before sending the request. Defaults to true. Pass a map to enable any of the following specific validation features:
- min [Boolean] — Validates that a value meets the min
constraint. This is enabled by default when paramValidation is set
to
true
. - max [Boolean] — Validates that a value meets the max constraint.
- pattern [Boolean] — Validates that a string value matches a regular expression.
- enum [Boolean] — Validates that a string value matches one of the allowable enum values.
- min [Boolean] — Validates that a value meets the min
constraint. This is enabled by default when paramValidation is set
to
-
computeChecksums
(Boolean)
—
whether to compute checksums for payload bodies when the service accepts it (currently supported in S3 only)
-
convertResponseTypes
(Boolean)
—
whether types are converted when parsing response data. Currently only supported for JSON based services. Turning this off may improve performance on large response payloads. Defaults to
true
. -
correctClockSkew
(Boolean)
—
whether to apply a clock skew correction and retry requests that fail because of an skewed client clock. Defaults to
false
. -
s3ForcePathStyle
(Boolean)
—
whether to force path style URLs for S3 objects.
-
s3BucketEndpoint
(Boolean)
—
whether the provided endpoint addresses an individual bucket (false if it addresses the root API endpoint). Note that setting this configuration option requires an
endpoint
to be provided explicitly to the service constructor. -
s3DisableBodySigning
(Boolean)
—
whether S3 body signing should be disabled when using signature version
v4
. Body signing can only be disabled when using https. Defaults totrue
. -
s3UsEast1RegionalEndpoint
('legacy'|'regional')
—
when region is set to 'us-east-1', whether to send s3 request to global endpoints or 'us-east-1' regional endpoints. This config is only applicable to S3 client. Defaults to
legacy
-
s3UseArnRegion
(Boolean)
—
whether to override the request region with the region inferred from requested resource's ARN. Only available for S3 buckets Defaults to
true
-
retryDelayOptions
(map)
—
A set of options to configure the retry delay on retryable errors. Currently supported options are:
- base [Integer] — The base number of milliseconds to use in the exponential backoff for operation retries. Defaults to 100 ms for all services except DynamoDB, where it defaults to 50ms.
- customBackoff [function] — A custom function that accepts a
retry count and error and returns the amount of time to delay in
milliseconds. If the result is a non-zero negative value, no further
retry attempts will be made. The
base
option will be ignored if this option is supplied. The function is only called for retryable errors.
-
httpOptions
(map)
—
A set of options to pass to the low-level HTTP request. Currently supported options are:
- proxy [String] — the URL to proxy requests through
- agent [http.Agent, https.Agent] — the Agent object to perform
HTTP requests with. Used for connection pooling. Defaults to the global
agent (
http.globalAgent
) for non-SSL connections. Note that for SSL connections, a special Agent object is used in order to enable peer certificate verification. This feature is only available in the Node.js environment. - connectTimeout [Integer] — Sets the socket to timeout after
failing to establish a connection with the server after
connectTimeout
milliseconds. This timeout has no effect once a socket connection has been established. - timeout [Integer] — Sets the socket to timeout after timeout milliseconds of inactivity on the socket. Defaults to two minutes (120000).
- xhrAsync [Boolean] — Whether the SDK will send asynchronous HTTP requests. Used in the browser environment only. Set to false to send requests synchronously. Defaults to true (async on).
- xhrWithCredentials [Boolean] — Sets the "withCredentials" property of an XMLHttpRequest object. Used in the browser environment only. Defaults to false.
-
apiVersion
(String, Date)
—
a String in YYYY-MM-DD format (or a date) that represents the latest possible API version that can be used in all services (unless overridden by
apiVersions
). Specify 'latest' to use the latest possible version. -
apiVersions
(map<String, String|Date>)
—
a map of service identifiers (the lowercase service class name) with the API version to use when instantiating a service. Specify 'latest' for each individual that can use the latest available version.
-
logger
(#write, #log)
—
an object that responds to .write() (like a stream) or .log() (like the console object) in order to log information about requests
-
systemClockOffset
(Number)
—
an offset value in milliseconds to apply to all signing times. Use this to compensate for clock skew when your system may be out of sync with the service time. Note that this configuration option can only be applied to the global
AWS.config
object and cannot be overridden in service-specific configuration. Defaults to 0 milliseconds. -
signatureVersion
(String)
—
the signature version to sign requests with (overriding the API configuration). Possible values are: 'v2', 'v3', 'v4'.
-
signatureCache
(Boolean)
—
whether the signature to sign requests with (overriding the API configuration) is cached. Only applies to the signature version 'v4'. Defaults to
true
. -
dynamoDbCrc32
(Boolean)
—
whether to validate the CRC32 checksum of HTTP response bodies returned by DynamoDB. Default:
true
. -
useAccelerateEndpoint
(Boolean)
—
Whether to use the S3 Transfer Acceleration endpoint with the S3 service. Default:
false
. -
clientSideMonitoring
(Boolean)
—
whether to collect and publish this client's performance metrics of all its API requests.
-
endpointDiscoveryEnabled
(Boolean|undefined)
—
whether to call operations with endpoints given by service dynamically. Setting this
-
endpointCacheSize
(Number)
—
the size of the global cache storing endpoints from endpoint discovery operations. Once endpoint cache is created, updating this setting cannot change existing cache size. Defaults to 1000
-
hostPrefixEnabled
(Boolean)
—
whether to marshal request parameters to the prefix of hostname. Defaults to
true
. -
stsRegionalEndpoints
('legacy'|'regional')
—
whether to send sts request to global endpoints or regional endpoints. Defaults to 'legacy'.
-
useFipsEndpoint
(Boolean)
—
Enables FIPS compatible endpoints. Defaults to
false
. -
useDualstackEndpoint
(Boolean)
—
Enables IPv6 dualstack endpoint. Defaults to
false
.
Property Details
Method Details
converse(params = {}, callback) ⇒ AWS.Request
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.
Service Reference:
Examples:
Calling the converse operation
var params = {
messages: [ /* required */
{
content: [ /* required */
{
image: {
format: png | jpeg | gif | webp, /* required */
source: { /* required */
bytes: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */
}
},
text: 'STRING_VALUE',
toolResult: {
content: [ /* required */
{
image: {
format: png | jpeg | gif | webp, /* required */
source: { /* required */
bytes: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */
}
},
json: {
},
text: 'STRING_VALUE'
},
/* more items */
],
toolUseId: 'STRING_VALUE', /* required */
status: success | error
},
toolUse: {
input: { /* required */
},
name: 'STRING_VALUE', /* required */
toolUseId: 'STRING_VALUE' /* required */
}
},
/* more items */
],
role: user | assistant /* required */
},
/* more items */
],
modelId: 'STRING_VALUE', /* required */
additionalModelRequestFields: {
},
additionalModelResponseFieldPaths: [
'STRING_VALUE',
/* more items */
],
inferenceConfig: {
maxTokens: 'NUMBER_VALUE',
stopSequences: [
'STRING_VALUE',
/* more items */
],
temperature: 'NUMBER_VALUE',
topP: 'NUMBER_VALUE'
},
system: [
{
text: 'STRING_VALUE'
},
/* more items */
],
toolConfig: {
tools: [ /* required */
{
toolSpec: {
inputSchema: { /* required */
json: {
}
},
name: 'STRING_VALUE', /* required */
description: 'STRING_VALUE'
}
},
/* more items */
],
toolChoice: {
any: {
},
auto: {
},
tool: {
name: 'STRING_VALUE' /* required */
}
}
}
};
bedrockruntime.converse(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Parameters:
-
params
(Object)
(defaults to: {})
—
modelId
— (String
)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.
-
messages
— (Array<map>
)The messages that you want to send to the model.
role
— required — (String
)The role that the message plays in the message.
Possible values include:"user"
"assistant"
content
— required — (Array<map>
)The message content.
text
— (String
)Text to include in the message.
image
— (map
)Image to include in the message.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse
— (map
)Information about a tool use request from a model.
toolUseId
— required — (String
)The ID for the tool request.
name
— required — (String
)The name of the tool that the model wants to use.
input
— required — (map
)The input to pass to the tool.
toolResult
— (map
)The result for a tool request that a model makes.
toolUseId
— required — (String
)The ID of the tool request that this is the result for.
content
— required — (Array<map>
)The content for tool result content block.
json
— (map
)A tool result that is JSON format data.
text
— (String
)A tool result that is text.
image
— (map
)A tool result that is an image.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)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.
Note: This field is only supported Anthropic Claude 3 models.Possible values include:"success"
"error"
system
— (Array<map>
)A system prompt to pass to the model.
text
— (String
)A system prompt for the model.
inferenceConfig
— (map
)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 theadditionalModelRequestFields
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
— (Array<String>
)A list of stop sequences. A stop sequence is a sequence of characters that causes the model to stop generating the response.
toolConfig
— (map
)Configuration information for the tools that the model can use when generating a response.
Note: This field is only supported by Anthropic Claude 3, Cohere Command R, Cohere Command R+, and Mistral Large models.tools
— required — (Array<map>
)An array of tools that you want to pass to a model.
toolSpec
— (map
)The specfication for the tool.
name
— required — (String
)The name for the tool.
description
— (String
)The description for the tool.
inputSchema
— required — (map
)The input schema for the tool in JSON format.
json
— (map
)The JSON schema for the tool. For more information, see JSON Schema Reference.
toolChoice
— (map
)If supported by model, forces the model to request a tool.
auto
— (map
)The Model automatically decides if a tool should be called or to whether to generate text instead.
any
— (map
)The model must request at least one tool (no text is generated).
tool
— (map
)The Model must request the specified tool.
name
— required — (String
)The name of the tool that the model must request.
additionalModelRequestFields
— (map
)Additional inference parameters that the model supports, beyond the base set of inference parameters that
Converse
supports in theinferenceConfig
field. For more information, see Model parameters.additionalModelResponseFieldPaths
— (Array<String>
)Additional model parameters field paths to return in the response.
Converse
returns the requested fields as a JSON Pointer object in theadditionalModelResultFields
field. The following is example JSON foradditionalModelResponseFieldPaths
.[ "/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 a400
error code. if the JSON Pointer is valid, but the requested field is not in the model response, it is ignored byConverse
.
Callback (callback):
-
function(err, data) { ... }
Called when a response from the service is returned. If a callback is not supplied, you must call AWS.Request.send() on the returned request object to initiate the request.
Context (this):
-
(AWS.Response)
—
the response object containing error, data properties, and the original request object.
Parameters:
-
err
(Error)
—
the error object returned from the request. Set to
null
if the request is successful. -
data
(Object)
—
the de-serialized data returned from the request. Set to
null
if a request error occurs. Thedata
object has the following properties:output
— (map
)The result from the call to
Converse
.message
— (map
)The message that the model generates.
role
— required — (String
)The role that the message plays in the message.
Possible values include:"user"
"assistant"
content
— required — (Array<map>
)The message content.
text
— (String
)Text to include in the message.
image
— (map
)Image to include in the message.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse
— (map
)Information about a tool use request from a model.
toolUseId
— required — (String
)The ID for the tool request.
name
— required — (String
)The name of the tool that the model wants to use.
input
— required — (map
)The input to pass to the tool.
toolResult
— (map
)The result for a tool request that a model makes.
toolUseId
— required — (String
)The ID of the tool request that this is the result for.
content
— required — (Array<map>
)The content for tool result content block.
json
— (map
)A tool result that is JSON format data.
text
— (String
)A tool result that is text.
image
— (map
)A tool result that is an image.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)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.
Note: This field is only supported Anthropic Claude 3 models.Possible values include:"success"
"error"
stopReason
— (String
)The reason why the model stopped generating output.
Possible values include:"end_turn"
"tool_use"
"max_tokens"
"stop_sequence"
"content_filtered"
usage
— (map
)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
— required — (Integer
)The number of tokens sent in the request to the model.
outputTokens
— required — (Integer
)The number of tokens that the model generated for the request.
totalTokens
— required — (Integer
)The total of input tokens and tokens generated by the model.
metrics
— (map
)Metrics for the call to
Converse
.latencyMs
— required — (Integer
)The latency of the call to
Converse
, in milliseconds.
additionalModelResponseFields
— (map
)Additional fields in the response that are unique to the model.
-
(AWS.Response)
—
Returns:
converseStream(params = {}, callback) ⇒ AWS.Request
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.
Service Reference:
Examples:
Node.js EventStream Example
// In Node.js, events are streamed and can be read as they arrive.
bedrockruntime.converseStream({/** params **/}, function(err, data) {
if (err) {
// handle error
return console.error(err);
}
var eventStream = data.stream;
eventStream.on('data', function(event) {
// Check the top-level field to determine which event this is.
if (event.messageStart) {
// handle messageStart event
} else if (event.contentBlockStart) {
// handle contentBlockStart event
} else if (event.contentBlockDelta) {
// handle contentBlockDelta event
} else if (event.contentBlockStop) {
// handle contentBlockStop event
} else if (event.messageStop) {
// handle messageStop event
} else if (event.metadata) {
// handle metadata event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
}
});
eventStream.on('error', function(err) { /** Handle error events **/});
eventStream.on('end', function() { /** Finished reading all events **/});
});
Browser EventStream Example
// In browsers, events aren't processed until the response is fully buffered.
// Events will be accessible as an array.
bedrockruntime.converseStream({/** params **/}, function(err, data) {
if (err) {
// handle error
return console.error(err);
}
var events = data.stream;
for (var event of events) {
// Check the top-level field to determine which event this is.
if (event.messageStart) {
// handle messageStart event
} else if (event.contentBlockStart) {
// handle contentBlockStart event
} else if (event.contentBlockDelta) {
// handle contentBlockDelta event
} else if (event.contentBlockStop) {
// handle contentBlockStop event
} else if (event.messageStop) {
// handle messageStop event
} else if (event.metadata) {
// handle metadata event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
}
}
});
Async Iterator EventStream Example (Experimental)
// In Node.js v10.x, Readable streams have experimental support for async iteration.
// Instead of listening to the event stream's 'data' event, you can use a for...await loop.
async function example() {
try {
const result = await bedrockruntime.converseStream({/** params **/}).promise();
const events = result.stream;
for await (const event of events) {
// Check the top-level field to determine which event this is.
if (event.messageStart) {
// handle messageStart event
} else if (event.contentBlockStart) {
// handle contentBlockStart event
} else if (event.contentBlockDelta) {
// handle contentBlockDelta event
} else if (event.contentBlockStop) {
// handle contentBlockStop event
} else if (event.messageStop) {
// handle messageStop event
} else if (event.metadata) {
// handle metadata event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
}
}
} catch (err) {
// handle error
}
}
Calling the converseStream operation
var params = {
messages: [ /* required */
{
content: [ /* required */
{
image: {
format: png | jpeg | gif | webp, /* required */
source: { /* required */
bytes: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */
}
},
text: 'STRING_VALUE',
toolResult: {
content: [ /* required */
{
image: {
format: png | jpeg | gif | webp, /* required */
source: { /* required */
bytes: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */
}
},
json: {
},
text: 'STRING_VALUE'
},
/* more items */
],
toolUseId: 'STRING_VALUE', /* required */
status: success | error
},
toolUse: {
input: { /* required */
},
name: 'STRING_VALUE', /* required */
toolUseId: 'STRING_VALUE' /* required */
}
},
/* more items */
],
role: user | assistant /* required */
},
/* more items */
],
modelId: 'STRING_VALUE', /* required */
additionalModelRequestFields: {
},
additionalModelResponseFieldPaths: [
'STRING_VALUE',
/* more items */
],
inferenceConfig: {
maxTokens: 'NUMBER_VALUE',
stopSequences: [
'STRING_VALUE',
/* more items */
],
temperature: 'NUMBER_VALUE',
topP: 'NUMBER_VALUE'
},
system: [
{
text: 'STRING_VALUE'
},
/* more items */
],
toolConfig: {
tools: [ /* required */
{
toolSpec: {
inputSchema: { /* required */
json: {
}
},
name: 'STRING_VALUE', /* required */
description: 'STRING_VALUE'
}
},
/* more items */
],
toolChoice: {
any: {
},
auto: {
},
tool: {
name: 'STRING_VALUE' /* required */
}
}
}
};
bedrockruntime.converseStream(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Parameters:
-
params
(Object)
(defaults to: {})
—
modelId
— (String
)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.
-
messages
— (Array<map>
)The messages that you want to send to the model.
role
— required — (String
)The role that the message plays in the message.
Possible values include:"user"
"assistant"
content
— required — (Array<map>
)The message content.
text
— (String
)Text to include in the message.
image
— (map
)Image to include in the message.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)The raw image bytes for the image. If you use an AWS SDK, you don't need to base64 encode the image bytes.
toolUse
— (map
)Information about a tool use request from a model.
toolUseId
— required — (String
)The ID for the tool request.
name
— required — (String
)The name of the tool that the model wants to use.
input
— required — (map
)The input to pass to the tool.
toolResult
— (map
)The result for a tool request that a model makes.
toolUseId
— required — (String
)The ID of the tool request that this is the result for.
content
— required — (Array<map>
)The content for tool result content block.
json
— (map
)A tool result that is JSON format data.
text
— (String
)A tool result that is text.
image
— (map
)A tool result that is an image.
Note: This field is only supported by Anthropic Claude 3 models.format
— required — (String
)The format of the image.
Possible values include:"png"
"jpeg"
"gif"
"webp"
source
— required — (map
)The source for the image.
bytes
— (Buffer, Typed Array, Blob, String
)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.
Note: This field is only supported Anthropic Claude 3 models.Possible values include:"success"
"error"
system
— (Array<map>
)A system prompt to send to the model.
text
— (String
)A system prompt for the model.
inferenceConfig
— (map
)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 theadditionalModelRequestFields
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
— (Array<String>
)A list of stop sequences. A stop sequence is a sequence of characters that causes the model to stop generating the response.
toolConfig
— (map
)Configuration information for the tools that the model can use when generating a response.
Note: This field is only supported by Anthropic Claude 3 models.tools
— required — (Array<map>
)An array of tools that you want to pass to a model.
toolSpec
— (map
)The specfication for the tool.
name
— required — (String
)The name for the tool.
description
— (String
)The description for the tool.
inputSchema
— required — (map
)The input schema for the tool in JSON format.
json
— (map
)The JSON schema for the tool. For more information, see JSON Schema Reference.
toolChoice
— (map
)If supported by model, forces the model to request a tool.
auto
— (map
)The Model automatically decides if a tool should be called or to whether to generate text instead.
any
— (map
)The model must request at least one tool (no text is generated).
tool
— (map
)The Model must request the specified tool.
name
— required — (String
)The name of the tool that the model must request.
additionalModelRequestFields
— (map
)Additional inference parameters that the model supports, beyond the base set of inference parameters that
ConverseStream
supports in theinferenceConfig
field.additionalModelResponseFieldPaths
— (Array<String>
)Additional model parameters field paths to return in the response.
ConverseStream
returns the requested fields as a JSON Pointer object in theadditionalModelResultFields
field. The following is example JSON foradditionalModelResponseFieldPaths
.[ "/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 a400
error code. if the JSON Pointer is valid, but the requested field is not in the model response, it is ignored byConverseStream
.
Callback (callback):
-
function(err, data) { ... }
Called when a response from the service is returned. If a callback is not supplied, you must call AWS.Request.send() on the returned request object to initiate the request.
Context (this):
-
(AWS.Response)
—
the response object containing error, data properties, and the original request object.
Parameters:
-
err
(Error)
—
the error object returned from the request. Set to
null
if the request is successful. -
data
(Object)
—
the de-serialized data returned from the request. Set to
null
if a request error occurs. Thedata
object has the following properties:stream
— (ReadableStream<Events> | Array<Events>
)The output stream that the model generated.
stream is an object-mode Readable stream in Node.js v0.10.x and higher. Attach a listener to thedata
event to receive events. stream is an array of events in browsers. The possible events that may be returned are listed below. Note that the top-level field in each event matches the event name.messageStart
— (map
)Message start information.
role
— required — (String
)The role for the message.
Possible values include:"user"
"assistant"
contentBlockStart
— (map
)Start information for a content block.
start
— required — (map
)Start information about a content block start event.
toolUse
— (map
)Information about a tool that the model is requesting to use.
toolUseId
— required — (String
)The ID for the tool request.
name
— required — (String
)The name of the tool that the model is requesting to use.
contentBlockIndex
— required — (Integer
)The index for a content block start event.
contentBlockDelta
— (map
)The messages output content block delta.
delta
— required — (map
)The delta for a content block delta event.
text
— (String
)The content text.
toolUse
— (map
)Information about a tool that the model is requesting to use.
input
— required — (String
)The input for a requested tool.
contentBlockIndex
— required — (Integer
)The block index for a content block delta event.
contentBlockStop
— (map
)Stop information for a content block.
contentBlockIndex
— required — (Integer
)The index for a content block.
messageStop
— (map
)Message stop information.
stopReason
— required — (String
)The reason why the model stopped generating output.
Possible values include:"end_turn"
"tool_use"
"max_tokens"
"stop_sequence"
"content_filtered"
additionalModelResponseFields
— (map
)The additional model response fields.
metadata
— (map
)Metadata for the converse output stream.
usage
— required — (map
)Usage information for the conversation stream event.
inputTokens
— required — (Integer
)The number of tokens sent in the request to the model.
outputTokens
— required — (Integer
)The number of tokens that the model generated for the request.
totalTokens
— required — (Integer
)The total of input tokens and tokens generated by the model.
metrics
— required — (map
)The metrics for the conversation stream metadata event.
latencyMs
— required — (Integer
)The latency for the streaming request, in milliseconds.
internalServerException
— (map
)An internal server error occurred. Retry your request.
message
— (String
)
modelStreamErrorException
— (map
)A streaming error occurred. Retry your request.
message
— (String
)originalStatusCode
— (Integer
)The original status code.
originalMessage
— (String
)The original message.
validationException
— (map
)Input validation failed. Check your request parameters and retry the request.
message
— (String
)
throttlingException
— (map
)The number of requests exceeds the limit. Resubmit your request later.
message
— (String
)
-
(AWS.Response)
—
Returns:
invokeModel(params = {}, callback) ⇒ AWS.Request
Invokes the specified Amazon Bedrock model to run inference using the prompt and inference parameters provided in the request body. You use model inference to generate text, images, and embeddings.
For example code, see Invoke model code examples in the Amazon Bedrock User Guide.
This operation requires permission for the bedrock:InvokeModel
action.
Service Reference:
Examples:
Calling the invokeModel operation
var params = {
body: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */, /* required */
modelId: 'STRING_VALUE', /* required */
accept: 'STRING_VALUE',
contentType: 'STRING_VALUE',
guardrailIdentifier: 'STRING_VALUE',
guardrailVersion: 'STRING_VALUE',
trace: ENABLED | DISABLED
};
bedrockruntime.invokeModel(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Parameters:
-
params
(Object)
(defaults to: {})
—
body
— (Buffer, Typed Array, Blob, String
)The prompt and inference parameters in the format specified in the
contentType
in the header. To see the format and content of the request and response bodies for different models, refer to Inference parameters. For more information, see Run inference in the Bedrock User Guide.contentType
— (String
)The MIME type of the input data in the request. The default value is
application/json
.accept
— (String
)The desired MIME type of the inference body in the response. The default value is
application/json
.modelId
— (String
)The unique identifier of the model to invoke to run inference.
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.
-
trace
— (String
)Specifies whether to enable or disable the Bedrock trace. If enabled, you can see the full Bedrock trace.
Possible values include:"ENABLED"
"DISABLED"
guardrailIdentifier
— (String
)The unique identifier of the guardrail that you want to use. If you don't provide a value, no guardrail is applied to the invocation.
An error will be thrown in the following situations.
-
You don't provide a guardrail identifier but you specify the
amazon-bedrock-guardrailConfig
field in the request body. -
You enable the guardrail but the
contentType
isn'tapplication/json
. -
You provide a guardrail identifier, but
guardrailVersion
isn't specified.
-
guardrailVersion
— (String
)The version number for the guardrail. The value can also be
DRAFT
.
Callback (callback):
-
function(err, data) { ... }
Called when a response from the service is returned. If a callback is not supplied, you must call AWS.Request.send() on the returned request object to initiate the request.
Context (this):
-
(AWS.Response)
—
the response object containing error, data properties, and the original request object.
Parameters:
-
err
(Error)
—
the error object returned from the request. Set to
null
if the request is successful. -
data
(Object)
—
the de-serialized data returned from the request. Set to
null
if a request error occurs. Thedata
object has the following properties:body
— (Buffer(Node.js), Typed Array(Browser)
)Inference response from the model in the format specified in the
contentType
header. To see the format and content of the request and response bodies for different models, refer to Inference parameters.contentType
— (String
)The MIME type of the inference result.
-
(AWS.Response)
—
Returns:
invokeModelWithResponseStream(params = {}, callback) ⇒ AWS.Request
Invoke the specified Amazon Bedrock model to run inference using the prompt and inference parameters provided in the request body. The response is returned in a stream.
To see if a model supports streaming, call GetFoundationModel and check the responseStreamingSupported
field in the response.
InvokeModelWithResponseStream
. For example code, see Invoke model with streaming code example in the Amazon Bedrock User Guide.
This operation requires permissions to perform the bedrock:InvokeModelWithResponseStream
action.
Service Reference:
Examples:
Node.js EventStream Example
// In Node.js, events are streamed and can be read as they arrive.
bedrockruntime.invokeModelWithResponseStream({/** params **/}, function(err, data) {
if (err) {
// handle error
return console.error(err);
}
var eventStream = data.body;
eventStream.on('data', function(event) {
// Check the top-level field to determine which event this is.
if (event.chunk) {
// handle chunk event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
} else if (event.modelTimeoutException) {
// handle modelTimeoutException event
}
});
eventStream.on('error', function(err) { /** Handle error events **/});
eventStream.on('end', function() { /** Finished reading all events **/});
});
Browser EventStream Example
// In browsers, events aren't processed until the response is fully buffered.
// Events will be accessible as an array.
bedrockruntime.invokeModelWithResponseStream({/** params **/}, function(err, data) {
if (err) {
// handle error
return console.error(err);
}
var events = data.body;
for (var event of events) {
// Check the top-level field to determine which event this is.
if (event.chunk) {
// handle chunk event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
} else if (event.modelTimeoutException) {
// handle modelTimeoutException event
}
}
});
Async Iterator EventStream Example (Experimental)
// In Node.js v10.x, Readable streams have experimental support for async iteration.
// Instead of listening to the event stream's 'data' event, you can use a for...await loop.
async function example() {
try {
const result = await bedrockruntime.invokeModelWithResponseStream({/** params **/}).promise();
const events = result.body;
for await (const event of events) {
// Check the top-level field to determine which event this is.
if (event.chunk) {
// handle chunk event
} else if (event.internalServerException) {
// handle internalServerException event
} else if (event.modelStreamErrorException) {
// handle modelStreamErrorException event
} else if (event.validationException) {
// handle validationException event
} else if (event.throttlingException) {
// handle throttlingException event
} else if (event.modelTimeoutException) {
// handle modelTimeoutException event
}
}
} catch (err) {
// handle error
}
}
Calling the invokeModelWithResponseStream operation
var params = {
body: Buffer.from('...') || 'STRING_VALUE' /* Strings will be Base-64 encoded on your behalf */, /* required */
modelId: 'STRING_VALUE', /* required */
accept: 'STRING_VALUE',
contentType: 'STRING_VALUE',
guardrailIdentifier: 'STRING_VALUE',
guardrailVersion: 'STRING_VALUE',
trace: ENABLED | DISABLED
};
bedrockruntime.invokeModelWithResponseStream(params, function(err, data) {
if (err) console.log(err, err.stack); // an error occurred
else console.log(data); // successful response
});
Parameters:
-
params
(Object)
(defaults to: {})
—
body
— (Buffer, Typed Array, Blob, String
)The prompt and inference parameters in the format specified in the
contentType
in the header. To see the format and content of the request and response bodies for different models, refer to Inference parameters. For more information, see Run inference in the Bedrock User Guide.contentType
— (String
)The MIME type of the input data in the request. The default value is
application/json
.accept
— (String
)The desired MIME type of the inference body in the response. The default value is
application/json
.modelId
— (String
)The unique identifier of the model to invoke to run inference.
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.
-
trace
— (String
)Specifies whether to enable or disable the Bedrock trace. If enabled, you can see the full Bedrock trace.
Possible values include:"ENABLED"
"DISABLED"
guardrailIdentifier
— (String
)The unique identifier of the guardrail that you want to use. If you don't provide a value, no guardrail is applied to the invocation.
An error is thrown in the following situations.
-
You don't provide a guardrail identifier but you specify the
amazon-bedrock-guardrailConfig
field in the request body. -
You enable the guardrail but the
contentType
isn'tapplication/json
. -
You provide a guardrail identifier, but
guardrailVersion
isn't specified.
-
guardrailVersion
— (String
)The version number for the guardrail. The value can also be
DRAFT
.
Callback (callback):
-
function(err, data) { ... }
Called when a response from the service is returned. If a callback is not supplied, you must call AWS.Request.send() on the returned request object to initiate the request.
Context (this):
-
(AWS.Response)
—
the response object containing error, data properties, and the original request object.
Parameters:
-
err
(Error)
—
the error object returned from the request. Set to
null
if the request is successful. -
data
(Object)
—
the de-serialized data returned from the request. Set to
null
if a request error occurs. Thedata
object has the following properties:body
— (ReadableStream<Events> | Array<Events>
)Inference response from the model in the format specified by the
body is an object-mode Readable stream in Node.js v0.10.x and higher. Attach a listener to thecontentType
header. To see the format and content of this field for different models, refer to Inference parameters.data
event to receive events. body is an array of events in browsers. The possible events that may be returned are listed below. Note that the top-level field in each event matches the event name.chunk
— (map
)Content included in the response.
bytes
— (Buffer, Typed Array, Blob, String
)Base64-encoded bytes of payload data.
internalServerException
— (map
)An internal server error occurred. Retry your request.
message
— (String
)
modelStreamErrorException
— (map
)An error occurred while streaming the response. Retry your request.
message
— (String
)originalStatusCode
— (Integer
)The original status code.
originalMessage
— (String
)The original message.
validationException
— (map
)Input validation failed. Check your request parameters and retry the request.
message
— (String
)
throttlingException
— (map
)The number or frequency of requests exceeds the limit. Resubmit your request later.
message
— (String
)
modelTimeoutException
— (map
)The request took too long to process. Processing time exceeded the model timeout length.
message
— (String
)
contentType
— (String
)The MIME type of the inference result.
-
(AWS.Response)
—
Returns:
waitFor(state, params = {}, callback) ⇒ AWS.Request
Waits for a given BedrockRuntime resource. The final callback or 'complete' event will be fired only when the resource is either in its final state or the waiter has timed out and stopped polling for the final state.
Parameters:
-
state
(String)
—
the resource state to wait for. Available states for this service are listed in "Waiter Resource States" below.
-
params
(map)
(defaults to: {})
—
a list of parameters for the given state. See each waiter resource state for required parameters.
Callback (callback):
-
function(err, data) { ... }
Callback containing error and data information. See the respective resource state for the expected error or data information.
If the waiter times out its requests, it will return a
ResourceNotReady
error.
Returns: