What it does

Guesslang takes a blob of source code and names the programming language. It supports 54 languages—from Assembly to YAML—and claims over 90% accuracy. You can pipe text to a CLI, call it from Python, or ask for probability scores when it’s uncertain.

The interesting bit

The model isn’t just a regex party. It’s a TensorFlow neural network trained on actual code, and Microsoft baked it into VS Code’s automatic language detection. That’s a lot of trust for a side project with under 900 GitHub stars.

Key highlights

• 54 supported languages, including oddballs like COBOL, DM, and Verilog
• CLI with --probabilities flag for ranked guesses
• Python API via Guess().language_name(source)
• Powers VS Code’s paste-to-detect feature (since v1.60)
• Training pipeline exists as separate guesslangtools repo

Caveats

• Requires Python 3.7+ and TensorFlow, which means Visual C++ redistributables on Windows
• The 90% accuracy claim lacks published benchmark details in the README
• No clear indication of model size or inference speed

Verdict

Worth a look if you’re building anything that needs to syntax-highlight unknown code or route snippets to the right parser. Skip it if you need guaranteed accuracy for security-critical classification—this is educated guessing, not certainty.