The purpose behind adding this idea to the ideas list of 2025 was to have an LLM agent that makes SymPy easier to use. Although I don't recall much being discussed before it's addition to the list (I might be wrong).
An idea definitely needs a mentor interested in mentoring a new contributor. I'm not sure who would mentor this project (not that it's a "bad" idea but without a mentor an idea will just be an idea.). Maybe, this is partly the reason why during the updation of this year's list, the idea was removed. There might be other reasons that I'm unaware of. And this idea specifically needs more discussion, I haven't seen it yet being discussed thoroughly to truly come down to a consensus if it would truly be nice to have something like this. It is a lot of typical work, turning the docs into LLM parseable descriptions, building tool schemas that are independent of the framework being used on the user's side (framework-agnostic), deciding whether the tool will allow multimodal inputs or not, implementing ways to trace LLM's "thinking" process, etc... Aasim On Tue, 27 Jan 2026 at 01:53, Srujan Patwardhan <[email protected]> wrote: > Hi, is this idea not being considered this time? > > On Saturday, 17 January 2026 at 16:15:46 UTC+5:30 Srujan Patwardhan wrote: > >> Hello SymPy maintainers, >> >> My name is Srujan Patwardhan. I’m a final-year Computer Science student >> at VIT Pune, and I’m also pursuing a B.S. in Data Science and Applications >> from IIT Madras. I’ve previously interned with organizations including >> Barclays and Spiro, and I’m currently working as an AI Engineer Intern at a >> product-based startup focused on Enterprise Intelligence and Business >> Dynamics. >> >> I reviewed the GSoC 2025 ideas list and came across the project “Official >> LLM Tool Agent for SymPy.” I noticed it wasn’t taken up last year, and I’m >> very interested in contributing to it for GSoC 2026. Recently, I’ve been >> working extensively on agentic architectures—particularly integrating LLMs >> with structured tool/function calling to enable reliable natural-language >> querying over complex systems. This project aligns closely with my >> interests and the work I’ve been doing. >> >> I had a couple of questions as I begin ramping up: >> >> 1. >> >> I plan to start by studying the SymPy documentation and identifying >> functions/modules with strong dependency relationships, so I can design >> representative workflows for early testing. Are there particular modules >> or >> function families you’d prefer the agent to prioritize (or that would be >> most valuable to target first)? >> 2. >> >> Is there a designated mentor (or maintainer) for this idea whom I >> should reach out to for initial guidance and discussion? >> >> I’d be excited to contribute to SymPy this year and would appreciate any >> direction on where to begin. Thank you for your time, and I look forward to >> hearing from you. >> >> Best regards, >> Srujan Patwardhan >> > -- > You received this message because you are subscribed to the Google Groups > "sympy" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion visit > https://groups.google.com/d/msgid/sympy/32ce8317-af90-465b-89c3-432a6c5728b4n%40googlegroups.com > <https://groups.google.com/d/msgid/sympy/32ce8317-af90-465b-89c3-432a6c5728b4n%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/sympy/CAO0%2BxCVc4-cpmDYLxG7cLr6DczD11BjD-AiLh%2Bq25FNVP4hshA%40mail.gmail.com.
