Skip to main content

BedrockAgent


Resource Icon​

Resource Overview​

A service that creates and manages AI agents running on AWS Bedrock.

Associated Resources​

Parent Resources​

Resource Configuration​

Basic Settings​

  • foundation_model : Option to specify the base LLM (foundation model) to use in Bedrock - amazon.titan-text-express-v1, amazon.nova-pro-v1, amazon.nova-lite-v1, amazon.nova-micro-v1, anthropic.claude-sonnet-4-20250514-v1:0, anthropic.claude-sonnet-4-5-20250929-v1:0, anthropic.claude-haiku-4-5-20251001-v1:0, anthropic.claude-3-sonnet-20240229-v1:0, anthropic.claude-3-opus-20240229-v1:0, anthropic.claude-3-haiku-20240307-v1:0, anthropic.claude-3-5-sonnet-20240620-v1:0, anthropic.claude-3-5-sonnet-20241022-v2:0, anthropic.claude-3-5-haiku-20241022-v1:0, anthropic.claude-3-7-sonnet-20250219-v1:0, deepseek.r1-v1, openai.gpt-oss-120b-1, openai.gpt-oss-20b-1, qwen.qwen3-coder-30b-a3b-v1, qwen.qwen3-32b-v1
  • description : Description of the agent
  • idle_session_ttl_in_seconds : Time (in seconds) before an idle session is automatically terminated
  • instruction : Default instructions the agent should follow (system prompt)
  • customer_encryption_key_arn : Configure Bedrock resource data to be encrypted using a user-specified KMS key (ARN)
  • target_bedrock_knowledge_base : ARN of the Bedrock agent to associate with the knowledge base
  • knowledge_base_association_description : Description of the association between the knowledge base and the agent
  • knowledge_base_state : Current state of the knowledge base - ENABLED, DISABLED

Guardrail Configuration (guardrail_configuration)​

  • guardrail_configuration.guardrail_use_method : Method for applying a guardrail - already_exists, create_new
  • guardrail_configuration.guardrail_identifier : Unique identifier (ID) of the guardrail to use
  • guardrail_configuration.target_bedrock_guardrail : Bedrock guardrail resource to associate
  • guardrail_configuration.guardrail_version : Version of the guardrail resource to use

Prompt Override Configuration (prompt_override_configuration)​

Prompt Settings (prompt_override_configuration.prompt_configurations)​

  • prompt_override_configuration.prompt_configurations.prompt_creation_mode : Whether prompts are automatically generated or manually specified - DEFAULT, OVERRIDDEN

  • prompt_override_configuration.prompt_configurations.prompt_type : Type of prompt - PRE_PROCESSING, ORCHESTRATION, POST_PROCESSING, KNOWLEDGE_BASE_RESPONSE_GENERATION

  • prompt_override_configuration.prompt_configurations.base_prompt_template : Defines the base sentence structure (template) of the prompt

  • prompt_override_configuration.prompt_configurations.prompt_state : Current state of the prompt - ENABLED, DISABLED

  • prompt_override_configuration.prompt_configurations.parser_mode : Method for interpreting the model output format - DEFAULT, OVERRIDDEN

  • prompt_override_configuration.prompt_configurations.max_length : Maximum length of the model response when executing the prompt

  • prompt_override_configuration.prompt_configurations.stop_sequences : String sequences that instruct the model to stop text generation

  • prompt_override_configuration.prompt_configurations.temperature : Option to control creativity (randomness) of generated responses

  • prompt_override_configuration.prompt_configurations.top_p : Option to limit word selection based on cumulative probability

  • prompt_override_configuration.prompt_configurations.top_k : Option to limit candidates to the top K highest-probability words

  • prompt_override_configuration.override_lambda : AWS Lambda function to invoke to customize default prompt behavior

References​