Anthropic Claude on AWS Messages

Overview

You can use this Snap to generate message responses using the specified Anthropic Claude model and model parameters. Additionally, you can use the Snap to create chat bots or virtual assistant applications.


Anthropic Claude on AWS Messages Overview

Prerequisites

None.

Known issues

None.

Limitations

  • When you select JSON mode with Claude models, they may produce malformed JSON, causing parsing errors.

    Workaround: Ensure your prompt clearly asks for a valid JSON response, such as: Respond with a valid JSON object.

Snap views

View Description Examples of upstream and downstream Snaps
Input This Snap supports a maximum of one binary or document input view. When the input type is a document, you must provide a field to specify the path to the input prompt. The Snap requires a prompt, which can be generated either by the Amazon Bedrock Prompt Generator or any user-desired prompt intended for submission to the Anthropic Claude Messages API.
Output This Snap has at the most one document output view. The Snap provides the result generated by the Anthropic Claude Messages API. Mapper
Error

Error handling is a generic way to handle errors without losing data or failing the Snap execution. You can handle the errors that the Snap might encounter when running the pipeline by choosing one of the following options from the When errors occur list under the Views tab. The available options are:

  • Stop Pipeline Execution Stops the current pipeline execution when an error occurs.
  • Discard Error Data and Continue Ignores the error, discards that record, and continues with the remaining records.
  • Route Error Data to Error View Routes the error data to an error view without stopping the Snap execution.

Learn more about Error handling in Pipelines.

Snap settings

Note:
  • Suggestion icon (): Indicates a list that is dynamically populated based on the configuration.
  • Expression icon (): Indicates whether the value is an expression (if enabled) or a static value (if disabled). Learn more about Using Expressions in SnapLogic.
  • Add icon (Plus Icon): Indicates that you can add fields in the field set.
  • Remove icon (Minus Icon): Indicates that you can remove fields from the field set.
Field / Field set Type Description
Label String

Required. Specify a unique name for the Snap. Modify this to be more appropriate, especially if more than one of the same Snaps is in the pipeline.

Default value: Anthropic Claude on AWS Messages

Example: Create customer support responses
Model name String/Expression/Suggestion

Required. Specify the model name to use for the chat completion. Learn more about the list of Anthropic Claude models from AWS Bedrock that are compatible with the Messages API.

Default value: N/A

Example: anthropic.claude-3-haiku
Use message payload Checkbox

Select this checkbox to generate responses using the messages specified in the Message payload field.

Note:
  • When you select this checkbox, the Snap hides the Prompt and System prompt fields and the Message payload field is displayed.
  • When the input view is Binary, this field is hidden.

Default status: Deselected
Message payload String/Expression

Appears when you select the Use message payload checkbox.

Required. Specify the prompt to send to the chat completions endpoint as the user message. The expected data type for this field is a list of objects (a list of messages). You can generate this list with the Amazon Bedrock Prompt Generator Snap.

For example,
[
    {
        "content": "You are a helpful assistant",
        "sl_role": "SYSTEM"
    },
    {
        "content": "Who won the world series in 2020?",
        "sl_role": "USER",
        "name": "Snap-User"
    },
    {
        "content": "The Los Angeles Dodgers won the World Series in 2020",
        "sl_role": "ASSISTANT"
    },
    {
        "content": "Where was it played?",
        "sl_role": "USER",
        "name": "Snap-User2"
    }
]
Note:
  • If a message contains an unsupported role (for example, SYSTEM), it is passed as is to the endpoint. This can potentially result in call failure because the role is not supported by the Amazon Bedrock Prompt Generator Snap.
  • If a message being processed contains any field not supported (for example, name field), it is ignored and excluded in the final message sent to the API because the endpoint does not support it. The supported fields are only content and role.

Default value: N/A

Example: $messages
Prompt String/Expression

Appears when you select Document as the Input type under the Views tab.

Required. Specify the prompt to send to the Anthropic Claude Messages endpoint as the user message.

Default value: N/A

Example: $msg
Model parameters Configure the parameters to tune the model runtime.
Maximum tokens Integer/Expression

Required. Specify the maximum number of tokens to generate the message responses.

Minimum value: 0

Maximum value: 4096

Default value: 200

Example: 100
Temperature Decimal/Expression

Specify the sampling temperature to use a decimal value between 0 and 1. If left blank, the endpoint uses the default value 1.

Default value: N/A

Example: 0.2
Top P Decimal/Expression

Specify the nucleus sampling value, a decimal value between 0 and 1. If left blank, the endpoint uses its default value set to 0.999.

Default value: N/A

Example: 0.2
Top K Integer/Expression

Specify a value to limit the number of high-probability tokens considered for each generation step to control the randomness of the output. It is disabled by default.

Minimum value: 0

Maximum value: 500

Default value: N/A

Example: 300
Stop sequences String/Expression

Specify a sequence of texts or tokens to stop the model from generating further output. Learn more.

Note:
  • You can configure up to 8191 stop sequences when generating the text. These stop sequences tell the model to halt further text generation if any of the specified sequences are encountered.
  • The returned text does not contain the stop sequence.

Default value: N/A

Example: pay, ["amazing"], ["September", "paycheck"]
Advanced prompt configuration Configure the advanced prompt settings.
JSON mode Checkbox/Expression

Select this checkbox to enable the model to generate strings that can be parsed into valid JSON objects. The output includes the parsed JSON object in a field named json_output that contains the data.

Note:
  • This field does not support input document values from upstream Snaps.
  • When the output from the model is an invalid JSON, the Snap fails indicating that it failed to parse the JSON in the output. However, the Snap provides the full output from the LLM Model in the error view with the error message.
  • When the output from the model indicates that there are not enough tokens, the Snap fails with a reason that it failed to parse the JSON in the output. However, the Snap provides the full output from the LLM Model in the error view.
  • When you select this checkbox and specify a Message payload, the Snap automatically adds an Assistant message at the end of the message payload to ensure the Claude model returns JSON in its response. Therefore, if you use JSON mode and a Message payload, ensure your message list ends with a User message, not an Assistant message, to avoid errors.

Default status: Deselected
System Prompt String/Expression

Specify the prompt (inital instruction). This prompt prepares for the conversation by defining Claude's role, personality, tone, and other relevant details to understand and respond to the user's input.

Note:
  • If you leave this field blank, empty or null, the Snap processes the request without using any system prompt.
  • This field supports input document values from upstream Snap.
  • The output represents the result generated by the Anthropic Claude on AWS Messages Response.

Default value: N/A

Example:
  • Explain the answer to a 6-year-old child.
  • Explain your role as an AI assistant.
Advanced response configurations Configure the response settings to customize the responses and optimize output processing.
Simplify response Checkbox/Expression Select this checkbox to receive a simplified response format that retains only the most commonly used fields and standardizes the output for compatibility with other models. This option supports only a single choice response. Here's an example of a simplified output format.
{
  "role": <string/null>,
  "content": <string/JSON(for JSON mode)>,
  "tool_calls": <array of tool call information object>, // optional
  "finish_reason": <string>,
  "usage": {
    "prompt_tokens": <integer/null>,
    "output_tokens": <integer/null>,
    "total_tokens": <integer/null>,
  },
  "_sl_responses": <object/array of the raw responses for debug mode> //optional
  "original": {}
}
Important: This field does not support upstream values.

Default status: Deselected
Continuation requests Checkbox/Expression
Select this checkbox to enable continuation requests. When selected, the Snap automatically requests additional responses if the finish reason is Maximum tokens.
Important: This Snap uses the same schema as the Anthropic Claude response. However, when multiple responses are merged through Continuation requests, certain fields may not merge correctly, such as id. This is because of the structure of the responses, where specific fields are not designed to be combined across multiple entries.
The following example represents the format of the output when you select the Continuation requests checkbox:
{
  "id": <id>,  // include if has only one repsonse
  "type": "message",
  "role": "assistant",
  "model": <model>,
  "content": [
    {
      "type": "text",
      "text": <response1> + <response2> + ... <response n>,
      
      // including, if type is "tool_use"
      "id": <id>,
      "name": <name>,
      "input": <input>
    }
  ],
  "stop_reason": <stop_reason from latest response>
  "stop_sequence": <stop_sequence from latest response>,
  "usage": {
    "input_tokens": <sum of input_tokens>,
    "output_tokens": <sum of output_tokens>
  },
  
  // When JSON mode is enabled
  "json_output": <parse json from output>,
  // When debug mode is enabled
  "_sl_responses": [<raw response1>, <raw response2>, ... ,<raw response n>]
}
Important: This field does not support upstream values.

Default status: Deselected
Continuation requests limit Integer/Expression

Appears when you select Continuation requests checkbox.

Required. Select the maximum number of continuation requests to be made.

Important: This field does not support upstream values.

Minimum value: 1

Maximum value: 20

Default value: N/A

Example: 3
Debug mode Checkbox/Expression

Appears when you select Simplify response or Continuation requests checkbox.

Select this checkbox to enable debug mode. This mode provides the raw response in the _sl_response field and is recommended for debugging purposes only. If Continuation requests is enabled, the _sl_responses field will contain an array of raw response objects from each individual request.

Important: This field does not support upstream values.

Default status: Deselected
Snap execution Dropdown list
Select one of the three modes in which the Snap executes. Available options are:
  • Validate & Execute: Performs limited execution of the Snap and generates a data preview during pipeline validation. Subsequently, performs full execution of the Snap (unlimited records) during pipeline runtime.
  • Execute only: Performs full execution of the Snap during pipeline execution without generating preview data.
  • Disabled: Disables the Snap and all Snaps that are downstream from it.

Default value: Validate & Execute

Example: Execute only

Troubleshooting

Unable to generate JSON response

The tokens are not enough.

Modify the settings and try again.

Unable to parse JSON content string

JSON output is malformed.

Modify the settings and try again.

Continuation requests limit error.

The Continuation requests limit value is invalid.

Provide a valid value for Continuation requests limit that is between 1-20.

Unable to parse JSON content string

This error occurs because of a limitation in the Claude models, which sometimes generates incorrectly formatted JSON responses.

To avoid this, explicitly request a valid JSON response in your prompt. For example, Respond with a valid JSON object.

Examples