Extract entities

The Extract Entities action is a powerful tool that automatically pulls specific, structured information from a user’s past conversation history. It uses LLM (Large Language Model) capabilities to intelligently extract data - such as addresses, names, or dates - based on what the user has already said, and saves them into variables for later use. This allows for seamless automation of data capture without needing the user to re-enter information.

Configuration Fields

• Conversation ID: A unique identifier for the conversation from which data will be extracted. Using the {{conversation_id}} variable ensures that the current conversation’s context is used.

• Entities: This is where you define what information you want to extract from the user’s conversation.

  • Entity name: The name of the Entity that you want to extract (e.g., user_address, meeting_date, user_email) this isn’t a part work flow variables.
  • Entity format: The expected format of the data to extract such as "Date", "Email", "Phone Number", or a custom-defined type.

• Storage: Extracted entities are stored inside the journey’s workflow variables.
  The workflow variables is a dictionary, and within it, extracted_entities is another dictionary that holds all the extracted key–value pairs.

💡 Example: If the user previously said “My address is in Delhi” and you’ve configured address as an entity, the Extract Entities action will analyze the conversation history and automatically save “Delhi” into the address variable.

you can use extract-entities later in your flow as we’ll

The Extracted entities are available in workflow variables and can be referenced using syntax. e.g.

  updated_address = variables["extracted_entities"]["address"]