Platform Metadata
Platform-specific options. The available settings vary depending on the selected platform and model.
| Platform | Type | Description |
|---|---|---|
| — | None | No platform-specific settings |
| AWS | BedrockClaude | For Claude models. Supports reasoning and tool execution |
| AWS | BedrockClaudeThinking (Deprecated) | Legacy mode for explicit thinking tokens (deprecated) |
| AWS | BedrockNova | For Nova models. Supports web search and code execution |
| AWS | BedrockKnowledgeBase (Deprecated) | Direct integration with Bedrock Knowledge Base (deprecated) |
| AWS | BedrockAgentCoreRuntime | Advanced agent workflows via AgentCore Runtime |
| Azure | AzureInterfaceOpenAIChatCompletions | For Azure OpenAI Chat Completions API |
| GoogleGemini | For Gemini models. Supports thinking process and Google Search | |
| OpenAI | OpenAIResponse | For Response API. Combines web search, file search, and code execution |
| OpenAI | OpenAIChatCompletions | For Chat Completions API |
BedrockClaude
Settings for Claude models. Supports reasoning and tool execution.
| Field | Description |
|---|---|
| Adaptive Thinking | Enables thinking process before responding |
| Thinking Level | Sets the depth of thinking (max / high / medium / low) |
| Max Tokens | Sets the maximum response length (4,096–64,000) |
| Max Tool Iterations | Sets the upper limit for tool iteration count (1–20) |
| Current Time Tool | Allows the AI to reference the current date and time |
| AgentCore Code Interpreter Tool | Enables code execution |
| ┗ Code Interpreter ID | The AgentCore Code Interpreter resource ID (required) |
| ┗ Code Interpreter Tool Description | Customizes the tool description passed to the AI (optional) |
| ┗ Code Interpreter Prompt ID | Specifies the system prompt ID for tool execution (optional, defaults to built-in) |
| Knowledge Base Search Tool | Configures Knowledge Bases that Claude can search as tools (multiple can be configured) |
| ┗ Knowledge Base ID | The Bedrock Knowledge Base ID (required) |
| ┗ Tool Description | Description passed to Claude explaining what information this KB contains (required) |
| ┗ Search Result Count | Number of results to retrieve (1–100, required) |
| ┗ Attribute Filter | Filter conditions in AWS Bedrock Knowledge Base JSON format (optional) |
| ┗ Reranking Model ARN | ARN of the model used for reranking (optional) |
| ┗ Reranking Result Count | Number of results to return after reranking. Must be smaller than Search Result Count (optional) |
| ┗ Document Link | Specifies how to link to source documents. Unset (ALL): default, OFF: no links, specific string: use metadata key as URL (optional) |
| ┗ KB Search Prompt ID | Specifies the system prompt ID for tool execution (optional, defaults to built-in AIS:tool/bedrock-knowledge-base) |
For detailed setup instructions, see the following pages.
BedrockClaudeThinking (Deprecated)
Legacy mode for explicitly specifying thinking tokens. Migration to BedrockClaude's Adaptive Thinking is recommended.
| Field | Description |
|---|---|
| Budget Token | Number of tokens allocated for the thinking process (required) |
| Max Token | Sets the maximum response length (required) |
BedrockNova
Settings for Nova models. Supports web search and code execution.
| Field | Description |
|---|---|
| Enable Web Search | Enables web search |
| Enable Code Interpreter | Enables code execution |
| Reasoning Setting | Sets the depth of reasoning (low / medium / high) |
BedrockKnowledgeBase(Deprecated)
Legacy mode for direct integration with Bedrock Knowledge Base. Migration to Grounding's Bedrock Knowledge Base Retrieval is recommended.
| Field | Description |
|---|---|
| Bedrock Knowledge Base ID | Specifies the Bedrock Knowledge Base ID (required) |
BedrockAgentCoreRuntime
Settings for advanced agent workflows using AgentCore Runtime.
| Field | Description |
|---|---|
| Agent Runtime ARN | Specifies the AgentCore runtime ARN (required) |
| RAG Prompt ID | Specifies the prompt ID to use (required) |
OpenAIResponse
Settings for OpenAI's Response API. Supports combining web search, file search, and code execution.
| Field | Description |
|---|---|
| Enable Web Search | Enables web search |
| Enable File Search | Enables file search |
| Enable Code Interpreter | Enables code execution |
| Reasoning Level | Sets the depth of reasoning (low / medium / high) |
OpenAIChatCompletions
Settings for OpenAI's Chat Completions API.
| Field | Description |
|---|---|
| Reasoning Level | Sets the depth of reasoning (none / low / medium / high) |
AzureInterfaceOpenAIChatCompletions
Settings for Azure OpenAI's Chat Completions API.
| Field | Description |
|---|---|
| Reasoning Level | Sets the depth of reasoning (none / low / medium / high) |
GoogleGemini
Settings for Google Gemini models. Supports thinking process and Google Search.
| Field | Description |
|---|---|
| Thinking Process Enabled | Enables the AI's thinking process |
| Thinking Budget | Sets the token budget for the thinking process |
| Thinking Level | Sets the depth of thinking (LOW / MEDIUM / HIGH) |
| Google Search Enabled | Enables Google Search |
| URL Context Enabled | Allows the AI to reference URL content |