# How to Use Polling Events The polling event feature allows you to periodically retrieve events from a remote API and automatically execute commands. ## Configuration Method ### 1. Enabling Polling Events To enable the polling event feature, configure the following settings: - **Polling Event URL**: API endpoint for retrieving events - Example: `https://api.example.com/events` - **Polling Interval (seconds)**: Interval between polling requests - Recommended: 30 seconds or more - **Polling Event Bearer Token**: Token for API authentication ### 2. Input in the Configuration Screen The following items can be entered in the configuration screen: ``` Polling Event URL: https://api.example.com Polling Interval (seconds): 30 Polling Event Bearer Token: •••••••••••••••• ``` ## Operation Specifications ### Polling Process - When enabled, GET requests are sent to the polling event URL at the specified interval - Commands are automatically executed based on the retrieved response - Each event has a unique ID to prevent duplicate execution ### API Response Format The polling endpoint must return an event list with the following schema: ```typescript // Base event schema id: z.string(), // Unique event ID createdAt: z.string(), // Event creation date and time type: z.literal('performSearch'), tag: z.string(), // Search tag query: z.string(), // Search query ``` ### Scheduling Functionality This mechanism does not have direct scheduling functionality, but scheduling can be achieved by updating the event list externally. ### Supported Event Types Currently supported event types: #### performSearch Event to execute a search - tag: Tag for the search target (when executing an LLM, the LLM will be executed with this tag) - query: Search query string (when executing an LLM, this effectively functions as a prompt) **LLM Execution Feature** - If an LLM can be executed with the tag, the query effectively functions as a prompt - If TabTool is installed, it's possible to open any page and analyze its contents ## Implementation Examples ### API Endpoint Example ```json { "events": [ { "id": "event-001",// use uuid "type": "performSearch", "tag": "claude-4-0", "query": "Please open https://example.com, retrieve its contents, and send it to a specific URL. Please close the tab after completion.", "createdAt": "2024-01-15T10:32:00Z" }, { "id": "event-002", "type": "performSearch", "tag": "gemini-flash-2.5", "query": "Please analyze the contents of the specified web page and create a summary. Use markdownTool to save it. Please close the tab after completion.", "createdAt": "2024-01-15T10:33:00Z" }, { "id": "event-003", "type": "performSearch", "tag": "gpt-5", "query": "Please regularly check news sites and extract important information. Please close the tab after completion.", "createdAt": "2024-01-15T10:34:00Z" } ] } ``` ## Use Cases ### Leveraging Google Chrome Operations Various use cases are possible by leveraging the advantages of being able to operate Google Chrome: #### Processing Pages Requiring Login - Retrieve content from pages requiring login - Analyze the retrieved data - Send analysis results to a specific endpoint with query parameters #### Flexible Prompt Execution - Ability to flexibly send prompts enables complex processing - Sufficiently supports regular automated processing #### Tab Management Optimization - Utilize TabTool's "closeTab" function to solve the problem of too many tabs - Recommended to include "Please close the tab after completion" in prompts ### Specific Application Examples **Regular Data Collection** - Retrieve data from dashboards requiring login - Analyze the retrieved data and generate reports - Send results to specified API endpoints **Website Monitoring** - Regularly check specific websites for changes - Notify when important updates occur - Analyze changes and create summaries **Automated Workflows** - Process multiple web pages sequentially - Integrate data from each page - Send final results to external systems ### Preventing Duplicate Execution - Each event ID must be unique - Event IDs that have been executed are recorded internally and will not be executed again - If an event with the same ID is received again, execution is skipped ## Precautions - It is recommended to set the polling interval to 30 seconds or more - API endpoints should implement appropriate authentication (Bearer token) - Response format should follow the specified schema - Be mindful of API rate limits when sending a large number of events at once ## Troubleshooting ### Common Issues **Polling Does Not Start** - Verify the URL is correctly configured - Check if the Bearer token is valid **Events Are Not Executed** - Verify the response format matches the schema - Check if event IDs are not duplicated **Authentication Errors** - Verify the Bearer token is correctly configured - Check the token's expiration date ## Summary The polling event feature is a powerful function that retrieves events from a remote API periodically and executes them automatically. Key points: - **Configuration**: Set URL, interval, and Bearer token - **Event Types**: performSearch enables LLM execution - **Scheduling**: Achieved through integration with external systems - **Chrome Operations**: Can process pages requiring login - **Tab Management**: Control tab proliferation with the closeTab function - **Duplicate Prevention**: Prevent duplicate execution with unique IDs By utilizing this feature, you can streamline various tasks such as regular data collection, website monitoring, and automated workflows. ## Comprehension Check **What is the basic mechanism of polling events?** **Answer:** **A feature that periodically retrieves events from a remote API and automatically executes commands** - It sends GET requests to the polling event URL at specified intervals and automatically executes commands based on the retrieved response. **What parameters are important when executing an LLM with a performSearch event?** **Answer:** **tag and query** - tag specifies the type of LLM (e.g., claude-4-0, gemini-flash-2.5, gpt-5), and query is a string that functions as a prompt. **What mechanism prevents duplicate execution?** **Answer:** **Unique ID system** - Each event ID must be unique, and event IDs that have been executed are recorded internally and will not be executed again. **Why is it recommended to use TabTool's "closeTab" function?** **Answer:** **To prevent the problem of too many tabs** - By including "Please close the tab after completion" in prompts, you can control tab proliferation. **How can scheduling functionality be achieved?** **Answer:** **By regularly adding events from external systems** - There is no direct scheduling functionality, but scheduling can be achieved by updating the event list externally. **What are appropriate examples of using the polling event feature?** **Answer:** **Regular data collection, website monitoring, automated workflows** - Data retrieval from dashboards requiring login, checking websites for changes, and sequential processing of multiple pages are all possible.

Techniques