AI Action

In automation, an action is one of the operations to be executed after a trigger is activated. Teable’s AI Action enables dynamic parsing and content generation of table data through the intelligent combination of natural language instructions (Prompt) and structured data (field values).

​

Configuration Process

1. Input Prompt
   • Input prompt
   • Click the ”+” button in the top right corner of the input box to select data from the table. You can insert values from fields in the table. You can also use data generated by other nodes as the data source.

2. Model Selection

Choose from specialized model types to match your task requirements:

• Vision Models 👁

  • Vision model supports advanced attachment recognition (e.g., text extraction, visual content analysis)
  • Compatible with common file formats: Word/Excel/PDF (auto-parses embedded text/images/charts)

• Reasoning Models ⚛️

  • Reasoning models output both processed messages and complete reasoning process, ideal for tasks requiring transparent decision-making (e.g., data validation, complex problem analysis).

4. Temperature Setting

   • Temperature is a parameter that controls the creativity level of AI-generated text
   • By adjusting the temperature, you can influence the AI model’s probability distribution
   • Temperature ranges:
     • Low (0.2-0.5): Produces more deterministic results (e.g., content organization, data analysis)
     • Medium (0.6-0.8): Balances accuracy and readability (e.g., report writing, email responses)
     • High (0.9-1.2): Generates more creative expressions (e.g., copywriting, story creation)

5. Output Type Selection

   • Available in both Text and JSON formats

   • JSON format output provides structured data(e.g., extracted key fields like amount/date from invoices), enabling easy process of downstream automation processes.

​

Testing

• After configuration, click the “Test” button
• The system will display the test results of the action. You can review the output and adjust the prompt as needed

​

Used Credits

• View credit consumption in the test output panel
• In Cloud version, usedCredits reflects credits consumed per model call. Calculated from input/output tokens using model-specific conversion rates (visible during model selection)