@User @Sahil

@Samesh I'm sorry to hear that you're experiencing issues with the agent not effectively using the Knowledge Base. To ensure the agent utilizes the Knowledge Base more effectively and reduces hallucinations, you can try the following steps: 1. **Explicitly Reference the Knowledge Base**: In your prompts, explicitly mention that the response should be based on the Knowledge Base. For example, "According to our Knowledge Base, can you tell me...?" 2. **Use Specific Keywords**: Include specific keywords or phrases from the Knowledge Base in your prompts to guide the agent towards the relevant information. 3. **Limit the Scope**: Narrow down the scope of the question to a specific topic that is covered in the Knowledge Base to make it easier for the agent to find and use the correct information. 4. **Feedback Loop**: Provide feedback to the agent when it does not use the Knowledge Base correctly. This can help in adjusting its responses in future interactions. 5. **Update the Knowledge Base**: Regularly update the Knowledge Base with accurate and relevant information to ensure the agent has the latest data to draw from. 6. **Restrict Responses**: Configure the agent to restrict its responses to only what is available in the Knowledge Base, if such a setting is available. If you've already tried these steps and are still facing issues, it might be beneficial to reach out to our support team for a more in-depth look into your specific case. Would you like me to assist you with contacting support or is there anything else I can help you with right now?

You can continue this conversation by mentioning me in the message.

Got it. Tagging @User for additional help.

And the issue is the last message from the agent (check transcript). It is just hallucinating and making up the response even though nothing like that was mentioned in my knowledge base anywhere

Like, we add the context from the KB via similarity search (based on the user response) + plus original response

due to which it causes some confusion

i would suggest you to use pinecone or any other vector db

and use it via tools calling

So would this work with just tools calling or would you suggest using a complete custom LLM? (and then retreive from vector db on my own server)

(Basically setup the entire RAG on my own server)

Because I don't think using tool calling will effectively add the retrieved KB content to the context?

if you could wait then you can use it

otherwise you can go with the RAG flow

I don't see any other possible APIs on VAPI

will be available pretty soon.

not right now

Appreciate the quick help Sahil

I will check this out

But just to understand, what exactly is the difference between the upcoming "custom kb" feature and the existing "kb" feature? Is it on how you do retrieval?

Or you are basically going to be storing the kb content in a vector db?

just like custom llm