I’ve been using Gemini to build an incident-report importer that cleans up CSV/Excel files and pushes them into Firestore. Right now it’s halfway there, but it still trips over a bunch of annoyances:

CSV/Excel parsing is fragile. It chokes on stray quotation marks and tiny format quirks. It’d be great if Gemini handled messy quoting and trimming automatically.

Column mapping is brittle. I need to split “Student Name” into first and last name, normalize years (“7th” → “Year 7”), and drop any unknowns. Having a simple schema-mapping feature would save tons of custom code.

Validation is manual. I keep wrestling with “Year Unknown” rows and missing “Positive Incident” in my points map. If I could declare “Allowed Years: Year 7–12” and “Category → points” up front, Gemini could flag or fix invalid rows for me.

Firebase upload setup is clunky. I still have to wire up firebaseadmin.initializeapp(), decide on collection names, write the doc-ID sanitization, then print success/failure logs. A one-click “export to Firestore” snippet would be a massive time-saver.

Diffs back into my repo. I’d love to point Gemini at my GitHub repo and have it apply these fixes directly to my existing scripts (instead of copy-pasting a whole new file).

If Gemini could natively:

Auto-clean and parse CSV/Excel

Let me visually map source columns to target schema

Enforce simple validation rules (allowed values, mandatory fields)

Scaffold the full Firebase-upload code with logging and retry logic

Push the changes directly back to my GitHub project

…it would turn a half-baked prototype into a production-ready pipeline. Right now I’m spending more time babysitting the imports than actually using the data. A little more schema awareness and built-in Firestore support would go a long way.