GSoC 2026 Proposal Draft Review: Mesa-LLM iteration to push to production

[yassinekolsi]

Hi everyone! 👋

I am preparing my GSoC 2026 application for the Mesa-LLM iteration to push to production project (175 hours).

Building on my previous PRs and the discussions around the repository, I've put together a detailed draft of my proposal. To tackle the core challenges, my architecture focuses on:

• A bounded two-phase execution model to handle the tick-loop latency constraint without stalling the simulation.
• A provider-agnostic ToolRegistry and adapter layer.
• A robust Ollama client wrapper with explicit health checks and context budget policies.

I’m tagging the listed mentors for this project (@jackiekazil, @colinfrisch) to make sure I'm on the right track, but I would honestly love thoughts from anyone in the community. Especially folks like @quaquel, @codebreaker32, or @falloficarus22 your feedback on my previous PRs was super helpful in shaping how I approached this

If anyone has a few spare minutes before the deadline, I'd be incredibly grateful for any quick feedback.

Link to Proposal (PDF): [https://drive.google.com/file/d/1fUWm_NYX2yFDPUdY9Ol7KD98FSfeS8ss/view?usp=sharing]
(Note: I used Google Drive so you can highlight text and leave comments directly on the PDF if that's easiest)

Thank you so much for your time and guidance