Self Checking
The term "self-checking" refers to the introduction of a target check mechanism based on the conversation chain.
It requires setting the targets of the conversation beforehand, then judging whether these targets are met based on the content of the conversation.
If the target is reached, it will proceed to the execution of the next conversation.
If the target is not reached, the question-follow-up mode will be initiated, with more rounds of dialogue added to achieve the conversation goal.
The number of follow-up questions will be set by the user. It's recommended to use a self-checking chain when you have a measurable target for your conversation.
For example: the goal of this conversation is to gather the user's work experience, or to convey the corporate values to the user, and so on.
Its configuration is as follows:
Model
The model is the core of any AI process. The model receives message inputs, uses them as parameters in the prompts, and outputs results. You can use models from several different providers:
OpenAI: Provides models such as gpt-3.5-turbo, gpt-4, etc. By clicking on more operations, you can alter other complex parameters such as temperature, completion length, or flow.
We will continuously integrate models from other providers, so stay tuned.
System Prompt
The prompt is an AI cue, a method of using natural language to guide or inspire AI models to complete specific tasks.
You can take some courses to improve the effectiveness of the prompts.
Target
Input the goal of the conversation. For example: Gather all the educational background information of the user.
Check Prompt
This part is highly crucial, it's the inspection logic of self-checking. We will fill in the template by default, and you can choose not to make any adjustments.
The goal is {target}, and the content of the conversation is {dialogue}. Please determine if the dialogue has achieved the target. If so, only output "Yes" and do not provide any other explanations. If not, in order to achieve the goal, combine with the following dialogue and ask a question again.
{target} and {chat_history} are variables. {target} is the target entered by the user, and {chat_history} is all the dialogue content between the AI and the user.
How the self-checking chain works
Self-checking engages in a dialogue with the user, and once the user responds, the content of the conversation and the objective are submitted to the check prompt for an AI evaluation.
If the AI determines that the goal has been achieved, it moves on to the next dialogue.
If the AI decides the objective has not been met, it checks whether the number of follow-up questions has been exceeded.
If not exceeded, it will initiate multiple rounds of follow-up questions.
If the follow-up limit is exceeded, it proceeds to the next dialogue stage.
Last updated