Hello all!
Do any of the out of the box LLM models within Foundry have the capability to use the web like the GPT-4o model does within the Chat GPT interface?
If not out of the box, are you aware of any models that enable this that could be brought into Foundry?
This is for a categorization effort where more up to date information from the web is helpful in improving accuracy, any pointers would be greatly appreciated!
Thanks in advance,
Paul Burns
To my knowledge it’s not there. We invested in building our own bing api proxy and made that available to AIP Logic as a function/tool.
+1 to have this native.
Any chance you would be willing to provide more details on this implementation?
It’s a function calling a webhook that calls the following API:
https://learn.microsoft.com/en-us/bing/search-apis/bing-web-search/quickstarts/rest/python
You’ll a subscription key.
You’re right, all of AIP’s Palantir-provided models are quite locked down, and don’t allow the LLMs to directly access the web, or any external systems, outside of your control. @nicornk’s post is a good overview of how you can give an LLM web access as a tool, subject to the limits and network egress policies you define.
Additionally, as of recently, you can also bring your own LLM, and register it as a model in AIP via a Function interface. This lets you integrate a search-capable model (e.g. Perplexity), using your own key.
Give it a spin!
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.