After sunsetting Trieve on Nov 1, what’s the recommended alternative for a knowledge base? I understand we have two options: building a custom solution or using the query tool with Gemini. Based on my experience so far, Gemini’s performance both in latency and accuracy is significantly worse compared to what Trieve provided. Hosting Trieve ourselves or creating our own knowledge base solution is not a trivial task (as shown by the number of SaaS companies focusing solely on this area), and it would be preferable if this were provided directly by Vapi. What’s the official stance on this?

Recommended Alternative: To closely match Trieve's automatic retrieval, set up a custom knowledge base using your own server. Here’s a minimal Node.js/Express example for the required

/kb/search

endpoint:

js
app.post('/kb/search', (req, res) => {
  const { message } = req.body;
  // Implement your search logic here
  res.json({ documents: [/* relevant docs */] });
});

Attach this endpoint to your Vapi assistant via the API for full control over retrieval logic. For details, see the [Custom Knowledge Base guide](https://docs.vapi.ai/knowledge-base/custom-knowledge-base). Source: - [Custom Knowledge Base](https://docs.vapi.ai/knowledge-base/custom-knowledge-base)

I understand the concern about Trieve’s sunset and the limitations you’re seeing with Gemini. I can help evaluate or implement a solution that balances accuracy and latency, whether that’s a custom knowledge base or optimizing the query tool. Could you clarify what your priority is: maintaining Trieve-level performance, or minimizing development overhead? @MarioC

@MarioC we are in the same position here. Are there any other Trieve level performance tools?

One of the mentioned knowledge base alternatives is PineCone: