Dynamic Prompting
A method of adapting requests in real time based on the model's answers. This approach lets you flexibly change the structure or content of a prompt depending on the current context of the conversation or the results of previous requests.
| Advantages | Drawbacks |
|---|---|
| Increases the adaptability and flexibility of work with the model | Can be hard to implement and manage |
| Lets you get more accurate and relevant answers | Requires more complex logic for handling answers |
| Effective for complex, multi-step tasks | Potentially increases system response time |
Dynamic Prompting example:
System: You are a dynamic assistant. Your task is to help the user put together a training plan. Adapt your questions and recommendations based on the user's answers.
Assistant:
Hello! I will help you put together a training plan. To start with, tell me what your main goal is: losing weight, building muscle mass or general fitness improvement?
Human: My main goal is building muscle mass.**
Assistant:**
Great, for building muscle mass it is important to combine strength training with the right nutrition. How many times a week are you ready to train?
Human: I can train 3 times a week.**
Assistant:**
Three times a week is a good frequency for building muscle mass. Now let's clarify the details:
- Do you have access to a gym or are you planning to work out at home?
- Do you have any experience with strength training?
- Are there any health limitations that need to be taken into account?
Answer these questions and I will be able to suggest a more specific training plan.