Step 2: Analyze your repo

Before generating any content, understand what context the script extracts — and why it matters.

What the analyzer does

The script scans your repository and builds a context summary that gets passed to the AI. It collects:

Field	Source	Used for
Project name	package.json, pyproject.toml, Cargo.toml, directory name	README title
Primary language	File extension counts	Install command format
Dependencies	requirements.txt, package.json, go.mod, etc.	Installation section
Entry points	main.py, index.js, app.go, etc.	Usage examples
License	LICENSE file contents	Contributing section
Key files	Up to 5 source files, truncated at 200 lines	All sections

Run the analysis

Sample projectLocal directoryGitHub repo

python readme_generator.py --sample --analyze

python readme_generator.py --input /path/to/your/project --analyze

python readme_generator.py --input owner/repo --analyze

The --analyze flag prints the full context summary without calling the AI or generating a README.

Review the output

Look for:

Is the project name correct? If not, the script couldn't find a config file to read it from. It falls back to the directory name. You can rename the output file later, or add a name field to your package.json / pyproject.toml.

Is the right language detected? Language is detected from the most common file extension. If a repo has more test files than source files in a secondary language, the detection may be off. This is used as a hint to the AI — it will not cause incorrect output, but may affect command formatting.

Are dependencies listed? Missing dependencies usually means the script did not find a recognized dependency file. Check that your requirements.txt, package.json, or equivalent is in the root directory.

Are entry points found? Entry points with recognized names (main.py, index.js, etc.) are read and included as code context. If your entry point has a custom name, the AI will still work — it will use other files for context.

Why context quality matters

The AI generates content based entirely on what you provide. A richer context leads to more accurate output:

Less context	More context
Generic install instructions	Commands specific to your stack
Placeholder usage examples	Examples based on your actual functions
Boilerplate descriptions	Accurate feature lists

If the analysis output looks thin, add a brief README.md stub with even a few sentences — the script reads existing README content and passes it as additional context.

---

Next step: Generate your README →