Paul H.

Following up on your offer of support for LiteLLM Proxy integration. Simply put, we want to proxy the web research enrichment in order to use our own models in GCP.

It does, thank you!

Thanks, I'll take a look into the API documentation. As the more (ideal) straightforward path of passing in a custom base url for OpenAI doesn't seem viable at this time, I think that will be our path forward. I would love to know if there was a technical or security reason why clay doesn't offer a customizable interface to OpenAI, as it seems popular amongst a lot of tools in the space today.

We're utilizing LiteLLM Proxy to create budgets, rate limits, observability of LLM calls etc (eventually a/b testing) and also offer a wider selection of models than are available in the native extensions. LiteLLM Proxy is compatible with the OpenAI interface, so to achieve our goals for at scale deployment for our organization, it's a critical component to us that offers significantly more flexibility than our current options with clay.

Muhammad S. thanks! I was hoping that I could use the OpenAI integration with a native enrichment, e.g. the web researcher agent. The HTTP API is nice but it appears we would need to write our own agent / enrichment

The question (since I deleted the message) Is the Clay team going to support Vertex AI in favor of google gemini ai studio.