Local LLM Tools - difficultly accessing tools

[totalwar235]

I am attempting to use Odysseus for OCR on written notes and then transcribing them to a document in the library to a text file. I am using Odysseus through docker and have had to modify the docker build to increase the timeout limit.

while using OLLAMA end points the LLM generally does not have access to the tools or does not generate multiple message to activate the tool use. In the attached message I tried to get my LLM to just append a line to the library file

I am assuming this is an issue on my end, are there any potential settings I need to change to get tooling to work with OLLAMA?

[SirThinksAl0t]

This happened to me on certain models, once I changed to a better model I got it to work.
Example of models that work for me: gemma-4-26b-a4b-qat, gemma-4-31b.

[neil969]

This is similar to what happens to me all the time.. I end up in a circular argument where it says it can do something but then just keeps saying it is done but it is not .. ( see #5492 )
It happens pretty well regardless what LLM model you use ( My "normal" model) is Gemma4:31b but it happens with even more capable models from the Ollama "Cloud" too - sometimes it works with no problems and then suddenly it stops working , but may come back again later - I have no idea what is going on :)
I update the build (docker on Linux, so - git fetch - git pull - docker compose build blah, blah ) every few days, and it has been there though many multiples of builds, so I would love to know too.
I think Pewdiepie is working on an optimised model for Odysseus, so maybe that is the answer...