GPT Chat Task (LLM Task)

Message Data Extraction

The task automatically extracts and processes certain structured data from the assistant's responses:

# HTML-style Tags

When the assistant includes content within HTML-style tags (e.g., <thinking>my thought process</thinking>), the task will:

• Extract the content and add it to message_data_extracted.{tag_name} in the response data
• Remove the tagged content from the message

# JSON Code Blocks

When the assistant includes JSON code blocks (enclosed in ```json), the task will:

• Parse the JSON content and add it to message_data_extracted.json in the response data
• Keep the JSON blocks in the original message

The extracted data is available in the task result under payload._gpt_chat_details.message_data_extracted. For example:

This feature can be disabled by setting the extract_data_from_message parameter to false, which may be necessary when working with responses that contain actual HTML content.