| Step Details | |
| Introduced in Version | 9.13 |
| Last Modified in Version | 9.27 |
| Location | AI > Common |
This step submits prompts to a configured Large Language Model (LLM) provider and returns the generated response. The Chat Completion step in the AI.Common module acts as a unified interface, allowing users to work with different AI providers (such as OpenAI, Azure Foundry, AWS Bedrock, and others) through a single step.
Users can provide prompts directly, select saved prompts, or pass prompts dynamically from the Flow. Additionally, users can configure or override the provider and model used to process the request. This step is useful for automating tasks such as chat responses, summarization, content generation, and structured data extraction within a Flow.
Prerequisites
This step requires installing the AI.Common module along with any supported provider modules (for example: OpenAI, Azure Foundry, AWS Bedrock, Anthropic, Google Gemini). These modules must be added as Project Dependencies before they are available for use.
Each provider must be configured in System > Settings with the required authentication details (such as API keys, endpoints, or deployment names) before use.
When overriding providers, only those added as dependencies to the Project will be available for selection.
Properties
Settings
| Property | Description | Data Type |
|---|---|---|
| Override API Key (v9.13+) | Allows users to override the API key configured at the system level for the selected provider. | Boolean |
| Override Default Provider (v9.27+) | Enables selection of a specific AI provider configured in the system. | Boolean |
| Provider | Specifies which AI provider to use for the request. Only providers added as project dependencies are available. | --- |
| Get Provider from Flow | Allows the provider to be passed dynamically from the Flow. Disables Override Default Provider when enabled. | Boolean |
| Select Model | Enables model selection for the chosen provider. | Boolean |
| Model | Specifies the model to use for generating the response. | --- |
| Get Model from Flow | Allows the model name to be passed dynamically from the Flow. Disables Select Model when enabled. | Boolean |
| System Prompt | Defines the system-level instructions that guide the LLM’s behavior, tone, and response style. | String |
| Select System Prompt | Allows selection of a predefined system prompt. Disables manual System Prompt input when enabled. | Boolean |
| Get System Prompt from Flow | Allows the system prompt to be passed dynamically from the Flow. Disables Select System Prompt when enabled. | Boolean |
| Use TextMerge System Prompt | Enables the use of a TextMerge-based system prompt. Disables Get System Prompt from Flow when enabled. | Boolean |
| TextMerge System Prompt | Displays the "Pick or Create TextMerge System Prompt" option when Use TextMerge System Prompt is enabled. | --- |
| Prompt | Specifies the user input sent to the LLM to generate a response. | String |
| Select Prompt | Allows selection of a prompt from the AI Prompt Manager. Disables direct Prompt input when enabled. | Boolean |
| Get Prompt from Flow | Allows the prompt to be passed dynamically from the Flow. Disables Select Prompt when enabled. | Boolean |
| Use TextMerge Prompt | Enables the use of a TextMerge-based prompt. Disables Get Prompt from Flow when enabled. | Boolean |
| TextMerge Prompt | Displays the "Pick or Create TextMerge Prompt" option when Use TextMerge Prompt is enabled. | --- |
Inputs
| Property | Description | Data Type |
|---|---|---|
| Message History | Provides prior conversation context to generate more accurate and contextual responses. | String |
| Response Schema (v9.15+) | Defines the structured JSON format the response must follow, enabling consistent and machine-readable outputs. | Any |
Outputs
| Property | Description | Data Type |
|---|---|---|
| Chat Response | The response generated by the selected AI provider based on the input prompt. | String |
| Reason (Failed outcome) | The reason for failure if the request does not complete successfully. | String |
About the Chat Response Structure:
The Chat Response output includes the generated result along with supporting metadata, such as the prompt used. The ChatResponse field contains the generated output from the LLM.
If a Response Schema is defined, the response will follow the specified structured format, ensuring consistent and predictable outputs that can be reliably used in downstream workflow steps.
Example: { "ChatResponse": "...generated text...", "Prompt": "List top 3 destinations to visit in NYC" }
This structure ensures that both the generated response and the input used to produce it are available for validation, debugging, and reuse within the workflow.
Step Changes
| Description | Version | Release Date | Developer Task |
|---|---|---|---|
| Removed the "Decisions.AI" prefix from LLM provider names in the provider picker and updated the default Max Iterations value to 5. | 9.27 | June 2026 | [DT-047294] |
| Updated Properties panel UI and restricted provider selection to project dependencies. | 9.27 | June 2026 | [DT-046761] |
| Added Response Schema input for structured output. | 9.15 | September 2025 | [DT-045124] |
| Introduced the Chat Completion step in AI.Common. | 9.13 | July 2025 | [DT-044658] |