You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
# Customer intent: As an integration solution developer, I want to learn about the benefits and support that Azure Logic Apps provides to build flexible, adaptable, and responsive workflows that complete tasks using AI agents, models, and other AI capabilities for my integration and automation scenarios.
12
12
---
13
13
@@ -54,8 +54,8 @@ This conceptual guide describes key concepts, differences between agent and non-
54
54
55
55
For more information, see the following documentation:
56
56
57
-
-[What is Azure OpenAI Service](/azure/ai-services/openai/overview)?
58
-
-[What is an AI agent](/azure/ai-services/agents/overview#what-is-an-ai-agent)?
57
+
-[What is Azure OpenAI Service?](/azure/ai-services/openai/overview)
58
+
-[What is an AI agent?](/azure/ai-services/agents/overview#what-is-an-ai-agent)
59
59
-[Azure OpenAI Service models](/azure/ai-services/openai/concepts/models)
60
60
61
61
## Key concepts
@@ -64,10 +64,10 @@ The following table provides basic introductions to key concepts:
64
64
65
65
| Concept | Description |
66
66
|---------|-------------|
67
-
| **Agent** | A prebuilt action that uses a structured iterative process to solve complex, multi-step problems. This agent can interact with humans by using a chat interface and work with services, systems, apps, and data by invoking tools that you create with prebuilt actions in Azure Logic Apps. An agent can process information, make choices, and complete tasks by using a deployed model in Azure OpenAI Service. <br><br>**Note**: An agent workflow can include multiple agents in a sequence. You can't add an agent inline as a tool in another agent. However, an agent can use another agent by meeting the following requirements: <br><br>- The secondary agent exists in another agent workflow within the same Standard logic app. <br><br>- The secondary agent workflow must start with a trigger that creates a callable endpoint, such as a [**Request** trigger](/azure/connectors/connectors-native-reqres). <br><br>- The "parent" agent must have a tool that runs the action named **Call workflow in this logic app**. You can use this action to run the secondary agent workflow. <br><br>For more information, see [What is an AI agent](/azure/ai-services/agents/overview#what-is-an-ai-agent)? |
67
+
| **Agent** | A prebuilt action that uses a structured iterative process to solve complex, multi-step problems. This agent can interact with humans by using a chat interface and work with services, systems, apps, and data by invoking tools that you create with prebuilt actions in Azure Logic Apps. An agent can process information, make choices, and complete tasks by using a deployed model in Azure OpenAI Service. <br><br>**Note**: An agent workflow can include multiple agents in a sequence. You can't add an agent inline as a tool in another agent. However, an agent can use another agent by meeting the following requirements: <br><br>- The secondary agent exists in another agent workflow within the same Standard logic app. <br><br>- The secondary agent workflow must start with a trigger that creates a callable endpoint, such as a [**Request** trigger](/azure/connectors/connectors-native-reqres). <br><br>- The "parent" agent must have a tool that runs the action named **Call workflow in this logic app**. You can use this action to run the secondary agent workflow. <br><br>For more information, see [What is an AI agent?](/azure/ai-services/agents/overview#what-is-an-ai-agent) |
68
68
|**Large language model (LLM)**| A program that is trained to recognize patterns and perform jobs without human intervention. <br><br> For more information, see [Azure OpenAI Service models](/azure/ai-services/openai/concepts/models). |
69
-
|**Agentic loop**| Another name for an agent that uses a structured iterative process to solve complex, multi-step problems. The agent accomplishes this goal by iteratively following these steps: <br><br>1. **Perceive**: Collectand process information from available data sources. <br><br>2. **Reason**: Understand tasks and formulate solutions by using generative AI models. <br><br>3. **Act**: Complete tasks by using available information and tools. |
70
-
|**Tool**| A tool contains one or more actions that perform a task for an agent. For example, a tool can send email, work with data sources, perform calculations or conversions, interact with APIs, and so on. <br><br>For more information, see the following documentation: <br><br>- [Limitations and known issues](create-agent-workflows.md#limitations-and-known-issues) <br>- [Create a tool to get weather forecast](create-agent-workflows.md#create-a-tool-to-get-weather-forecast).|
69
+
|**Agentic loop**| Another name for an agent that uses a structured iterative process to solve complex, multi-step problems. The agent accomplishes this goal by iteratively following these steps: <br><br>1. **Think**: Collect, process, and analyze available information and inputs, such as text, images, audio, sensor data, and so on, from specific data sources. Apply reason, logic, or learning models to understand requests, create plans or solutions, and choose the best action to answer or fulfill requests with help from generative AI models. <br><br>2. **Act**: Based on the choices made and available tools, complete tasks in the digital or real world. <br><br>3. **Learn** (Optional): Adapt its own behavior over time by using feedback or other information. |
70
+
|**Tool**| A tool contains one or more actions that perform a task for an agent. For example, a tool can send email, work with data sources, perform calculations or conversions, interact with APIs, and so on. <br><br>For more information, see the following documentation: <br><br>- [Limitations and known issues](create-agent-workflows.md#limitations-and-known-issues) <br>- [Create a tool to get weather forecast](create-agent-workflows.md#create-a-tool-to-get-weather-forecast)|
71
71
| **Agent parameter** | A parameter that you create on a tool or in an action parameter, based on the use case for the agent parameter. You create agent parameters so the agent can pass model-only outputs as parameter inputs for an action in a tool. You don't need agent parameters for values from non-model sources. For more information, see [Create agent parameters](create-agent-workflows.md#create-agent-parameters-for-the-get-forecast-action). <br><br>Agent parameters differ from traditional parameters in the following ways: <br><br>- Agent parameters apply only to the tool where you define them. This restriction means you can't share agent parameters with other tools. In comparison, you can share traditional parameters globally with operations and control flow structures in a workflow. <br><br>- Agent parameters don't have resolved values when the workflow starts to run. An agent parameter receives a value only if the agent invokes the tool by using specific arguments. These arguments become the agent parameters for invoking the tool. <br><br>- An agent can invoke the same tool multiple times with different agent parameter values, even when that tool exists in the same loop iteration. For example, a tool can check the weather in both Seattle and London. |
72
72
|**Channels**| The input and output channels in the agent's chat interface. An agent has a **Channels** tab that controls whether you can exchange messages with the agent in the chat interface. <br><br>For more information, see the following documentation: <br><br>- [Limitations for agent chat interface - Channels tab](create-agent-workflows.md#limitations-and-known-issues) <br>- [Set up chat interactions for the agent](create-agent-workflows.md#set-up-chat-interactions-for-the-agent)|
73
73
|**Context**| An agent maintains chat history by keeping a maximum number of tokens or messages as context and passing that context into the model for the next chat interaction. Each model has different *context length* limits. For more information, see [Limitations for chat history context length (token limit)](create-agent-workflows.md#limitations-and-known-issues). |
Copy file name to clipboardExpand all lines: articles/logic-apps/ai-resources.md
+10-1Lines changed: 10 additions & 1 deletion
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -6,7 +6,7 @@ ms.suite: integration
6
6
ms.reviewers: estfan, azla
7
7
ms.topic: conceptual
8
8
ms.collection: ce-skilling-ai-copilot
9
-
ms.date: 05/19/2025
9
+
ms.date: 05/21/2025
10
10
#CustomerIntent: I want a guide that introduces starting points, building blocks, examples, samples, and other resources to help me learn about using AI in my integration solutions using Standard and Consumption workflows in Azure Logic Apps.
11
11
---
12
12
@@ -41,6 +41,15 @@ For more information, see the following resources:
41
41
| Resource type | Link |
42
42
|---------------|------|
43
43
|**Documentation**| - [Workflows with AI agents and models in Azure Logic Apps](agent-workflows-concepts.md) <br><br>- [Create workflows that use AI agents and models to complete tasks in Azure Logic Apps](create-agent-workflows.md)|
44
+
|**Blog article**|[Announcing agent loop: Build AI Agents in Azure Logic Apps](https://techcommunity.microsoft.com/blog/integrationsonazureblog/%F0%9F%93%A2announcing-agent-loop-build-ai-agents-in-azure-logic-apps-%F0%9F%A4%96/4415052)|
45
+
|**Blog article**|[Agent Loop Demos for Azure Logic Apps](https://techcommunity.microsoft.com/blog/integrationsonazureblog/%F0%9F%A4%96-agent-loop-demos-%F0%9F%A4%96/4414770)|
46
+
|**Video overview**|[Getting Ready for Agents in Azure Logic Apps](https://youtu.be/EMXLB4thdhg)|
![prmerger-automator[bot]](https://avatars.githubusercontent.com/u/22479449?v=4&size=40){kind=avatar}
0 commit comments