increase simulation model complexity

[TomMonks]

The feasibility test of this approach used a simple DES of a call centre. This limits the number of parameters available for the agent to generate. A step up in complexity would be to introduce treat-sim https://pypi.org/project/treat-sim/ as it provides more complex parameter sets and also time dependent arrivals (this is effectively a parameter file representing an arrival profile - so some thought needed here about generating these - with a smaller model like gemma3:27b this may need to be a separate server tool/resource and LLM context). treat-sim also has a full test suite and is more robust for piloting.

Side note: I have an R Simmer version called the treat-simmer package available on R-universe https://pythonhealthdatascience.r-universe.dev/packages It would be nice to have an MCP server available that uses an R model. mcptools may be a way forward here https://www.tidyverse.org/blog/2025/07/mcptools-0-1-0/

To do:

1. Add treat-sim

pip install treat-sim

Also add to environment.yml

2. Create an MCP wrapper

• Implement a simple MCP wrapper for treat-sim. Given the time dependent arrival rate the get_schema may need some consideration. A good stress test
• Suggested way forward - limit to simpler parameters in phase 1.

[TomMonks]

#2 should be setup before implementation