I built CodeIngest (like gitingest for local files) (github.com)

CodeIngest is a tiny CLI tool (think “gitingest” but for local files) that converts a codebase into a single AI-friendly text file you can feed into models like GPT or Claude. Install by moving or symlinking the script into /usr/local/bin, then run it against a directory or specific files (e.g., codeingest ., codeingest app.py utils.py). Options include mixing files and directories, specifying a custom output (-o output.txt), and adding ignore patterns (--ignore "*.test.js" "temp*"). The generated output combines a directory tree plus full file contents with clear, repeated separators (e.g., “================================================ FILE: src/app.py =================================================”) so models and ingestion pipelines can easily parse boundaries. That makes it handy for creating inputs for RAG, fine-tuning, automated code review prompts, or indexing into vector stores. Practical implications: simplifies chunking and context management, eases preprocessing for embedding or prompt engineering, and lets you exclude tests/temps to keep token usage down—important for large repos. Be mindful of token limits and sensitive data when ingesting entire projects.