This article explains how to run agents within the AgentRunner application. Agents are workflows created in the node editor that can complete complex tasks, such as utilizing LLM nodes or making API calls.
Audience
This article is intended for all AgentRunner users, including owners, developers, and operators, all of whom have permission to run agents.
Prerequisites
Access to an AgentRunner project containing at least one agent.
Familiarity with the AgentRunner node editor and the concept of agents.
Running an Agent
You can run an agent from two locations within the AgentRunner application:
Project View: Locate the agent you want to run and click the play icon located under the name of the agent. ❗ ![Play Icon Location] //Recommend screenshot showing the play icon under a agent name in the project view. Alt text: "Play icon location in project view."
Node Editor: Open the agent in the node editor. Click the "Run Agent" button located in the right top corner. ❗ ![Run Agent Button Location in Node Editor] //Recommend screenshot showing the "Run Agent" button in the node editor. Alt text: "Run Agent button in the node editor."
Using the Run Agent Modal
Clicking the play icon or the "Run Agent" button will open the Run Agent modal in a full-screen view. This modal is divided into two main sections:
Left Side: Inputs - Displays the input fields required to run the agent.
Right Side: Outputs - Displays the results returned by the agent after execution.
Selecting an Agent Version
At the top of the Run Agent modal, there is a dropdown menu that allows you to select which version of the agent you want to run. Each version can be loaded and run independently.
❗![Version Selector Location] //Recommend screenshot showing the version selector dropdown. Alt text: "Version selector in the Run Agent modal."
Closing the Run Agent Modal
To close the Run Agent modal, either:
Click the "X" button located in the top right corner of the modal.
Press the "ESC" key on your keyboard.
Note: Pressing ESC when the Run Agent modal is not in focus will close the modal and you will not receive those responses, but you can still find whatever is returned in the logs.
Inputting Values
The left side of the Run Agent modal displays the input fields defined in the node editor.
Input Field Names: Each input field is labeled with a name that corresponds to the input name defined in the node editor. This makes it easy to identify which field represents which input.
Input Field Expansion: Each input field has an icon (arrows pointing in two different corners) that you can click to expand the input field. This is especially useful for longer text inputs, as it allows you to scroll within the text. ❗ ![Expand Input Field Icon] //Recommend screenshot showing the expand input field icon. Alt text: "Expand input field icon."
Input Types: Currently, only text inputs are accepted. You cannot input other datatypes. In the future, image and voice inputs will be supported.
Entering Input Values: Enter the appropriate text value into each input field.
Running the Agent
Once all required input fields have been filled, click the "Run Agent" button.
Note: If you click "Run Agent" without filling in all the required input fields, the run will fail and an error message will be displayed. You can then complete the inputs and click "Run Agent" again.
Viewing Outputs
After the agent has been run, the results will be displayed on the right-hand side of the Run Agent modal.
Output Structure: Each agent can have a maximum of one output node, which can contain an unlimited number of key-value pairs.
Output Field Names: The names of the output values can be edited in the node editor, making them easy to recognize in the Run Agent modal.
Output Display: Each output value will be displayed in a separate field.
Output Formatting: You can view the outputs as either a JSON object or a list of fields.
JSON Object: Displays the output as a standard JSON object. Useful for developers who need to copy the JSON result.
List of Fields: Displays the output as a list of key-value pairs. When viewing as a list, you can further choose to format the output as either Markdown or JSON. Useful for content creators or others who need a specific text format.
Copying Output: A copy button is located in the right bottom of each output field. Clicking this button will copy the text within that output field to your clipboard. ❗ ![Copy Button Location] //Recommend screenshot showing the copy button in an output field. Alt text: "Copy button location in output field."
Post-Execution Information
After an agent is run:
Results are displayed on the screen.
Details of the execution are saved in the logs.
Expected Outcome
After following these steps, you should be able to successfully run a agent, input the required values, and view the results in the Run Agent modal.
Troubleshooting
If you encounter unexpected results or errors when running an agent:
Check the node editor to ensure that all inputs and outputs are correctly connected.
Consult the detailed, searchable logs for more information. Developers and owners can view and debug these logs.
Additional Resources
Tutorial video (available upon login)
Feedback
You can provide feedback on the "Run Agent" feature or suggest improvements through the AI chatbot within the AgentRunner application.
