Dapr Conversation Task (Type: 20)

The Dapr Conversation Task (type: "20") invokes an LLM/AI provider through Dapr's Conversation building block, giving workflows a provider-agnostic way to run prompts (OpenAI, Anthropic, AWS Bedrock, …) over the same Dapr sidecar used by the other Dapr task types.

The provider and its credentials live in a domain-owned Dapr conversation component (e.g. openai); the task only references that component name (componentName) and the messages (inputs). Required config: componentName. Optional: inputs (array of role/content messages), parameters (provider string parameters such as model, maxTokens), metadata, contextId, temperature, scrubPII, and timeoutSeconds (default 30). To add the runtime component, see vnext-helm-charts #28.

🚧 Full English translation is pending. See the Turkish page for the complete configuration tables, message structure, mapping example, and response reference.