Cleaning Up Before Launch: The Unglamorous Work That Makes Open Source Matter

We were three days away from pushing AI Agents Salebot to GitLab when reality hit. Ninety-four files, nearly 30,000 lines of Python, 17 production modules—and absolutely none of it was ready for public consumption.

The project had grown organically over weeks. Every sprint left artifacts: local databases in data/, development notes in docs/archive/, Vosk speech recognition models sitting at several megabytes each. The .gitignore was a suggestion, not a rule. When you’re heads-down building features, you don’t think about what you’re accidentally committing.

But shipping means reckoning.

The first decision was philosophical. The codebase carried MIT licensing—permissive, forgiving, almost too open. For a bot handling API authentication and security logic, we needed teeth. GPL-3.0 became the choice: copyleft protection ensuring anyone building on our work must open-source their improvements. It’s a two-minute change in a LICENSE file, but it echoes everything we believe about what should be free.

Then came the brutal editing. Out went docs/archive/—internal notes nobody needed. Out went local databases and environment-specific logs. The Vosk models, precious as they were during development, couldn’t justify their megabyte weight in a distributed repository. We kept .env.example as a bootstrap template instead of committing actual credentials.

The repository structure revealed itself: src/ for modules, tests/ for pytest suites, scripts/ for utilities. Everything else was either documentation or configuration. Aggressive pruning made decisions clearer.

Initialization mattered. We used git init --initial-branch=main --object-format=sha1, explicitly choosing SHA-1 for GitLab compatibility. The first commit—hash 4ef013c—contained exactly what belonged: the entry point bot.py, all 17 Python modules with their async patterns intact, test suites, and nothing else. No mystery files. No “we’ll figure this out later.” No garbage.

Here’s the thing nobody tells you about open source: the unglamorous cleanup is where projects earn credibility. It’s not the feature count or the test coverage percentages. It’s knowing that when someone clones your repository, they get exactly what works—no extraneous models, no abandoned notes, no local configuration bleeding through.

We pushed main to GitLab expecting a smooth deployment. DNS hiccups happened (of course), but the repository itself was solid. Clean history, clear purpose, protected intent.

Why did the Java developer never finish their cleanup? They kept throwing exceptions. 😄