Hi @VeryMilkyJoe, @fendor 
I’m Sanskar Balpande, a student at IIT Madras, and I’m seriously interested in working on the “Language Server Support for cabal.project Files” project for GSoC 2025 — because every time I switch between packages or toggle flags, I lose valuable time reconstructing dependency graphs and overrides that could be surfaced instantly with proper tooling.
Quick background:
-
I built a truly undetectable AutoTyper for CodeTantra (the most secure exam application in India) (GitHub - SanskarXD69/CodeTantra-AutoTyper: Tired of tediously typing code into CodeTantra’s text boxes while “learning automation”? Meet my CodeTantra AutoTyper-the only bot smart enough to learn from CodeTantra so you don’t have to!)), perfectly simulating human-like typing quirks to bypass its restrictive editor. If I can accomplish what thousands failed to do even after trying everything, I’m ready to fix HLS’s handling of cabal.project.
-
Infact i am the only student among all IITs and NITs who has succeeded in making an AutoTyper for CodeTantra
-
I run vLLM on Proxmox VE containers with Agent Zero as the primary pipeline, so I’m quite comfortable with the finest Ai-Agents.
A few questions I’d love your input on:
- Which specific cabal.project editing scenarios (like flag toggling, dependency resolution across multiple packages, etc.) would deliver the most immediate value if supported first?
- What’s your preferred feedback loop — should I iterate via GitHub Discussions, or would RFC-style posts here on Discourse be more helpful for the community?
- Are there any canonical projects or configs we should use to stress-test things like completions and diagnostics in real-world cabal. project setups?
I’m currently planning to prototype a minimal LSP plugin and iterate with your feedback. This feature could save me—and the entire Haskell community—hours of context-switching every week.
Looking forward to your thoughts!
— Sanskar (sanskarbalpande@gmail.com)
→ github/SanskarXD69