Notifications
Clear all
Topic starter
The agent does not query the vector database and relies solely on its internal knowledge for responses.
Hello! I'm working on setting up a RAG AI Agent, but it seems the agent isn't consulting the vector database and is instead using its own knowledge base to answer. I've noticed that when I ask the same question, sometimes it accesses the database, and other times it doesn't. What steps can I take to ensure the system consistently checks the database first?
Information on your callin.io setup
- callin.io version:1.67.0
- Database (default: SQLite): Postgres PGvector
- callin.io EXECUTIONS_PROCESS setting (default: own, main):
- Running callin.io via (Docker, npm, callin.io cloud, desktop app):
- Operating system:
Posted : 20/12/2024 3:53 am
Hi there!
Thanks for posting and welcome to the community!
This is the interesting part of prompt engineering.
In your instructions, you're telling it to “use the provided embeddings,” which is a bit unclear to the AI.
callin.io utilizes Langchain’s tool calling feature for vector store retrieval. Therefore, the AI needs to understand that the “embeddings” it should access is the tool it needs to invoke.
So, in summary: you need to make your Agent instructions and the tool description more detailed.
Something like this:
You will need to iterate over this quite a bit to get the prompt perfect so it functions as you intend the agent to behave.
I would suggest checking out YouTube to learn more about building RAG systems with callin.io, as there is a lot of content available to help you learn.
Posted : 20/12/2024 10:31 am
Topic starter
Thank you so much for your response. I'm still getting the hang of things. I'll give it a shot and reach out again if I run into any issues. Thanks again. Apologies for the delayed reply.
Posted : 23/12/2024 6:52 am
This thread was automatically closed 7 days following the last response. New replies are no longer permitted.
Posted : 30/12/2024 6:53 am