Description
The Claude component interfaces with AI capabilities provided by Anthropic.
※To use the Claude component, an API Key from Anthropic is required in advance. For instructions on obtaining an API Key from Anthropic, please refer to Claude API Key Setup.
Component Properties
| Property Name | Description | Example |
|---|---|---|
| Basic Properties | Specifies the basic properties under the "Claude" category | |
| Platform | Choose from Anthropic, Bredrock, Vertex for the platform using Claude features | |
| Model | Specify the model name to be used according to the Claude feature and platform | "claude-3-5-sonnet-20240620" |
| System Message | Set the initial message for the AI's role and behavioral guidelines | |
| Sample Image File | Sample image file for setting the context of the conversation | |
| User Message Input Example | Sample user message for setting the context of the conversation | |
| Assistant Response Output Example | Sample assistant response for setting the context of the conversation | |
| Image File | The image file to be used in actual conversations | |
| Max Tokens | The maximum number of tokens for responses. Although the maximum value varies by model, the maximum value for all models is 4096 | |
| Temperature | (0—1.0) Controls the diversity of outputs. Lower values result in more predictable outputs, higher values produce more random outputs | |
| TopP | (0—1.0) Selects tokens based on cumulative probability. Values closer to 1.0 produce more diverse outputs | |
| TopK | (Integer) Limits the number of most likely tokens considered. Smaller values result in more focused outputs, larger values in more diverse outputs | |
| Stop Sequences | Specifies certain strings to stop text generation. It's an array of JSON strings that halts generation when the specified sequence is reached. | ["END", "STOP", "."] |
| Timeout | API timeout | |
| API Key | Enter when Anthropic is selected on the platform | |
| Access Key | Enter when Bedrock is selected on the platform. Access key obtained from AWS | |
| Secret Access Key | Enter when Bedrock is selected on the platform. Secret access key obtained from AWS | |
| Region | Choose when Bedrock is selected on the platform. Select the region available for the Claude model to be used | |
| Project ID | Enter when Vertex is selected on the platform. The ID of the GCP project to be used | |
| Location | Enter when Vertex is selected on the platform. The location selected determines the available models | |
| Secret Key JSON | Enter when Vertex is selected on the platform. The secret key JSON file needed for authentication of the service account created in Google Cloud |
※ For more details on Anthropic-related models, refer to the Anthropic User Guide. It includes descriptions of the models, maximum tokens, and usage for Anthropic, Bedrock, and Vertex.
※ For details on Anthropic API timeout, refer to the Anthropic API Reference.
※ For any inquiries about the links related to Anthropic, please contact Anthropic directly.
※ For details on regions and permissions to use Claude with Amazon Bedrock, refer to Amazon Bedrock.
※ For any inquiries about the links related to Amazon Bedrock, please contact Amazon directly.
※ For details on locations related to Vertex, refer to Generative AI on Vertex AI. Be sure to refer to the English page as the Japanese page may be outdated. After enabling the Vertex AI API in the Before you begin section, you need to enable the Claude model card you intend to use.
※ For any inquiries about the links related to Vertex, please contact Google directly.
Output Payload
| Output Payload | Description |
|---|---|
cv.Payload | Stores the response from the Claude API as is |
Output Component Local Variables
| Component Local Variable | Description |
|---|---|
| id | Process ID |
| model | Model name used |
| stop_reason | Reason for stopping |
| stop_sequence | Sequence at which stopping occurred |
| input_tokens | Input tokens |
| output_tokens | Output tokens |