Tool use not triggered within Anything LLM

Hi,
first of all many thanks for this great llm server! It perfectly fits my needs for local AI use without heating up the internal gpu all the time! It works great with my python scripts for local e-mail-summaries.

I have the following issue:
I'd like to use FastFlowLM as my backend for Anything LLM (up to now, it's LM Studio, but there is no NPU support). In Anything LLM, I've configured an Agent with Web Search via Serper-API.

My problem: It seems that FastFlowLM doesn't trigger the tool use. I've tried with different search providers like DuckDuckGo (didn't work) and with different backends like LM Studio (worked).
I've also tried different LLMs within FastFlowLM, but none triggered the tool use. The result is a hanging chat request, that never ends.

The missing tool use might be an issue in the FastFlowLM OpenAI-API-implementation or is simply not available in FastFlowLM (yet? ;)).

In any case, it would be great, if tool use (at least web search) worked with FastFlowLM :)

Thanks!

Start by reproducing an AnythingLLM Agent web-search request against FastFlowLM's OpenAI API, using the Serper API setup described in the issue. Compare the request behavior with LM Studio and trace FastFlowLM's tool-use handling. Done means the request completes and invokes the configured web-search tool instead of hanging.