Hey Vapi team, I’ve found ways of accessing the audio data but I also need control of the text data and responses for my RAG. I’ve seen a tutorial which uses make.com however I need more control than this. Any advice?

@bertiefa Hello there! It's great to hear that you've been able to access the audio data successfully. For more control over the text data and responses, especially for a robust application like a RAG (Retrieve and Generate model), you might want to consider a few options: 1. **Direct API Access**: If you're looking for granular control, directly using the Voice API (VAPI) to send and receive data might be the best approach. This way, you can handle the text data and responses within your own infrastructure. 2. **Custom Integration**: Depending on your requirements, you might want to build a custom integration that can handle complex logic and maintain state between interactions. This could involve server-side processing with your own logic layer that interacts with the VAPI. 3. **Webhooks**: You can use webhooks to receive real-time data from VAPI and process it according to your needs. This can be a powerful way to integrate VAPI with your RAG system. 4. **Middleware Services**: If make.com doesn't offer the control you need, you might want to look into other middleware services or platforms that offer more flexibility and customization options. 5. **Local Processing**: For maximum control, consider processing the text data locally on your servers. This allows you to apply custom algorithms or machine learning models to the data before sending responses back through VAPI. If you need specific guidance on how to implement these options or if you have any other questions, please feel free to ask. We're here to help you make the most out of VAPI for your project!

can you elaborate more what you have done till now and what your trying to achieve and what resources used to help your better.

Hey Shubham I'm also interested in improving RAG for my VAPI app, Right now I'm only using the Vapi WebApp to upload documents and add them to my agent. I feel like the agent is very unreliable, sometimes it finds info present in the uploaded documents, sometimes it doesn't. That makes me think I'd like to adapt the RAG process under the hood, as the standard way of the VAPI app doesn't seem working good enough for my use case. What would be A) the easiest way to customize the RAG of my agent and B) the approach where latency stays as minimal as possible while still being able to customize the RAG process?

use external rag provider and integrate it as tool.

ok great will try! But the latency will probably suffer a lot right?

This was my thinking. Having access to the actual RAG code would be helpful

yes somewhat comprise.