Excited to be here! I'm DevRay,
🔧 Has anyone tried dynamically switching LLM prompts based on user intent (e.g., symptom checker vs. career advice) within a single Groq session? Curious how you structure that without bloating token usage.
Looking forward to learning, sharing, and building with everyone!
— DevRay
Hi DevRay,
That’s a good question! I’ve only built systems like this that chain prompts together, but I haven’t tried doing this ins a single Groq API call.
We have a new Prompt Engineering Pattern Guide that @benank put together, and there’s examples of Chain of Thought, Guided CoT, ReAct, and Chain of Verification — I think with a combination of those prompt techniques you might get something that might work.
Please let me know if using any of them work!
Thanks yawnxyz for the quick reply!
That Prompt Engineering Pattern Guide sounds incredibly helpful — I’ll definitely check it out. I think combining Chain of Thought with either Guided CoT or ReAct could be a great way to approach it in a single Groq API call.
I’ll experiment with those techniques and let you know how it goes!
Appreciate the guidance!
Best,
DevRay
I’ve always wanted some time to mix and match and experiment with these techniques to see if they work in tandem.
Good luck and please report back any findings!