# Interested in GSoC — currently on #2004 (routing layer)

**URL:** <https://discourse.holoviz.org/t/interested-in-gsoc-currently-on-2004-routing-layer/10054>\
**Category:** Lumen\
**Tags:** lumen\
**Created:** [August 16, 2026, 3:58am UTC](https://discourse.holoviz.org/t/interested-in-gsoc-currently-on-2004-routing-layer/10054 "2026-08-16T03:58:42Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Vamshi1404](https://discourse.holoviz.org/user_avatar/discourse.holoviz.org/vamshi1404/32/7543_2.png) [@Vamshi1404](https://discourse.holoviz.org/u/Vamshi1404)\
**Post date:** [August 16, 2026, 3:58am UTC](https://discourse.holoviz.org/t/interested-in-gsoc-currently-on-2004-routing-layer/10054/1 "2026-08-16T03:58:42Z")

</div>

Hi all — interested in GSoC down the line, and currently working through #2004 (routing layer for llm.py).

Traced through the model\_kwargs/invoke flow and proposed an approach in the issue — now waiting on maintainer input on one open design question.

Background: I’ve been building a similar NL-to-SQL agent project (schema-aware, multi-model fallback), so this space is genuinely interesting to me. Happy to hear thoughts on #2004, or pointers to other good areas to explore in the meantime.

---

<div class="post-metadata">

**Author:** ![ghostiee-11](https://discourse.holoviz.org/user_avatar/discourse.holoviz.org/ghostiee-11/32/7469_2.png) [@ghostiee-11](https://discourse.holoviz.org/u/ghostiee-11)\
**Post date:** [September 14, 2026, 8:02pm UTC](https://discourse.holoviz.org/t/interested-in-gsoc-currently-on-2004-routing-layer/10054/2 "2026-09-14T20:02:07Z")

</div>

Thanks for taking this up [http://github.com/holoviz/lumen/pull/2015](https://PR) reviewed and merged 🙂