Database Integration
Hi,
How to integrate a database with the LLM to pick up a relevant question to user profile?
How to integrate a database with the LLM to pick up a relevant question to user profile?
1 answers ( 0 marked as helpful)
Hi Ahmed,
Thank you for your question!
In our project, the database query is built based on the user's selected settings (level, position industry compnay). However, if you want the LLM to select relevant questions based on the user’s profile or bio, here’s an approach I’d recommend:
1. Label your questions clearly in the database using metadata - such as skill, difficulty, category, and industry - so they can be filtered programmatically.
2. Introduce a separate LLM (or lightweight classifier) whose role is to analyze the user’s bio and determine what types of questions would be most appropriate. This "decider" model should be familiar with your labeling system so it can generate a structured query to retrieve the most relevant questions from the database.
This way, you're combining structured filtering with LLM reasoning to deliver more personalized and targeted questions to each user.
I hope this helps! If you have any more questions, feel free to reach out.
Best,
Petar
Thank you for your question!
In our project, the database query is built based on the user's selected settings (level, position industry compnay). However, if you want the LLM to select relevant questions based on the user’s profile or bio, here’s an approach I’d recommend:
1. Label your questions clearly in the database using metadata - such as skill, difficulty, category, and industry - so they can be filtered programmatically.
2. Introduce a separate LLM (or lightweight classifier) whose role is to analyze the user’s bio and determine what types of questions would be most appropriate. This "decider" model should be familiar with your labeling system so it can generate a structured query to retrieve the most relevant questions from the database.
This way, you're combining structured filtering with LLM reasoning to deliver more personalized and targeted questions to each user.
I hope this helps! If you have any more questions, feel free to reach out.
Best,
Petar