Publishing Contextifier to PyPI

PyPI Distribution Motivation

I needed a library that converts files directly into LLM-ready chunks.

Distribution Pipeline

flowchart TD
    A[코드 작성] --> B[pyproject.toml 설정]
    B --> C[버전 범프]
    C --> D[GitHub Push]
    D --> E{GitHub Actions}
    E --> F[빌드]
    F --> G[PyPI 업로드]
    G --> H["pip install contextifier"]

    E -.->|workflow_dispatch| I[수동 배포 트리거]

pyproject.toml Configuration

[project]
name = "contextifier"
version = "0.2.2"
description = "다양한 문서 형식을 AI가 이해할 수 있는 텍스트로 변환"
authors = [{name = "CocoRoF"}]
requires-python = ">=3.9"

GitHub Actions Workflow

The workflow is configured to support manual deployment triggers via workflow_dispatch, which comes in handy for emergency patches.

Naming Mistake

I initially published the package under the name "Contextify", but since a package with that name already existed on npm, I renamed it to "Contextifier" to avoid confusion. The rename happened in v0.1.1 and required updating the GitHub repository name, documentation, and all workflow files.

Version History

v0.1.0 → v0.1.1 (rename) → v0.1.5 (OCR 개선) → v0.1.6 (차트 추출) → v0.2.0 (테이블 인터페이스) → v0.2.1 (문서화) → v0.2.2 (청킹 개선)

Seven version bumps in three weeks. I iterated quickly and incorporated user feedback along the way.

Lessons Learned

1. Choose your name carefully from the start — renaming later is costly
2. workflow_dispatch is invaluable — a manual trigger option in your CI/CD pipeline is a must
3. Follow semver properly — define clear criteria for what warrants a minor version bump