Text output

4 Situations:

• "system_question" attribute of "#single_turn_interact#" node: it can only be a declarative description and must ask question,
• "inform_user_before_asking" attribute of the "#single_turn_interact#" node: it can only be a declarative description and cannot ask question
• The "response" part of the "response_scripts" attribute of the "#start#" node: it can only be a declarative description and cannot ask question
• "#inform_user#" node's "inform_content" attribute: it can only be a declarative description and cannot ask questions In these four cases, there must be no description of the actual execution of the action (actions that need to be actually executed must be solved with the "#activity#..." node, and sometimes the "#trigger#" mechanism is also needed), and there should be no duplicate content (similar meanings here are also considered duplicates)

2 Expression Forms:

• Direct Output to the user (different from the prompt word mode below, here cannot require LLM to actually perform an action), such as:
  • "What do you need?"
  • "Please wait"
  • "You are so rich!"
  • "You had a lot of worries before, but don't worry."
  • "Don't worry, we still need to learn more about the situation before we can give you specific suggestions."
• A descriptive statement used to generate output content (i.e. Prompt Mode). At this time, it needs to be enclosed by a pair of greater than and less signs. It is a subject-less statement starting with a verb (generally requiring LLM to perform an action related to content output), such as:
  • "<Ask users what they need>"
  • "<Please ask the user to wait a moment.>"
  • "<Sigh that the user is very rich>"
  • "<Summarize previous users' concerns and use relevant reasons to convince users not to worry>"
  • "<First comfort the user, and then express that we need to know more about the situation before we can give specific suggestions>" In this "prompt mode": use the "you" or the "system_role" of "#start#" node to refer to the AI ​​robot in the dialogue, and use the "user" or "user_role" of the "#start#" node to refer to the other party (i.e. the user) in the dialogue; use other pronouns that can easily cause confusion with caution

Replacement of Text Output Content

• The above 4 situations (2 expression forms): If there is a string like "...{...}..." in the content, the system will replace the "{...}" with the string value of the corresponding InfoItem before outputting it.
• The above 4 situations (2 expression forms): For Python ChatTree, if there is "|" in the content, it will be replaced with "/"; Xmind ChatTree will not

Polling Output of Text Content

The above 4 situations (2 expression forms):

• For Xmind ChatTree, text can be separated into several parts by "|" (each part can be in either of the above 2 expression forms), such as "...|<...>|...". Each time the system executes to the relevant node, it will sequentially extract a part of it and output it, so as to achieve the effect of non-repetitive output content during repeated executions.
• Python ChatTrees can use a list to contain multiple output contents (each part can be in any of the 2 expression forms mentioned above), such as ["...","<...>","..."]. Each time the system executes to the relevant node, it will sequentially extract an element of the list for output, so as to achieve the effect of non-repetitive output content during repeated executions.

Content Requirements

• As "system_question" of "#single_turn_interact#" node, you need to pay attention to the completeness of the description of the scene in the question content. For example, "How many kilometers does it run in a year?" can be optimized to "How many kilometers does the vehicle run in a year?"
• Do not use modal particles at the beginning of sentences such as "um", "oh", "ok" etc
• If there is a string of letters and/or numbers, it is best to enclose it in square brackets or single quotes to ensure the correct TTS effect