[RFC] Technical Debt Audit & GSoC 2026 Implementation Roadmap

[ayushshukla1807]

Hey everyone,

With the GSoC pre-selection phase wrapping up, I wanted to share a quick summary of what I've been seeing under the hood of Montage. Over the last few days, I've managed to land a few surgical fixes (PRs #544 to #553), but the process really helped me map out where the project needs to go this summer.

Whether or not I'm selected, I think these items are high priority for the project's health:

1. Modernization & Safety: We've got a lot of legacy Py2 stuff (like six) that's just taking up space. I've started adding type hints in refactor: modernize utils.py with type hints and remove legacy Py2 code #551, but we should really push this across rdb.py to stop those annoying NoneType crashes. Also, we really need to kill that circular import TODO in loaders.py once and for all.

2. Data Integrity: I found some favorites leaking between campaigns. PR fix: add database-level UniqueConstraint to Favorite model #553 is a start, but we need a full Alembic setup to enforce these constraints at the DB level so we don't have to rely on application-side checks.

3. Infrastructure: The hardcoded debug login in public_endpoints is a bit of a security headache. I've got a proposal in feat: add /health check and modernize public endpoints #552 to standardize this. I also think we should beef up the /health check to monitor the Toolforge volume connections directly.

4. Juror Tools: The entry_lookup I added in feat: implement entry ID lookup by filename for jurors #550 has a lot of potential. I want to build that out into a proper metadata inspector so jurors can see uploader history without leaving the interface.

I've attached a quick architectural roadmap below. Really enjoyed getting into the weeds with this!

graph TD
    A[Montage Backend] --> B[Modernization]
    A --> C[Data Integrity]
    A --> D[Infrastructure]
    B --> B1[Type Hints - Python 3.14]
    B --> B2[Remove Legacy Py2/Six]
    C --> C1[Alembic Migrations]
    C --> C2[UniqueConstraints]
    D --> D1[Secure OAuth Config]
    D --> D2[Health Monitoring]

Loading

[lgelauff]

Thanks for your thoughts. I'm not sure if the approach of adding a lot of PR's is the best way to have a high level conversation. I

What exactly are you proposing the goal is here? This issue sounds like the AI talking more than you - it doesnt mention the why, and only looks at code quality by itself. For example, while an entry lookup might be an interesting side feature, is it something anyone has asked for before? What is the connection with requests? How does it help the juror/coordinator workflow?

[ayushshukla1807]

Fair point on the tone—I was just trying to organize my notes. The main goal is to fix the practical bottlenecks I ran into while doing the PRs (like the circular imports in loaders.py). The entry lookup was just something I realized was missing when I was debugging image IDs. If no one's asked for it, maybe it's not a priority, but it seemed like a low-hanging fruit for jurors.