docparser/Roadmap
Roadmap
Where docparser stands today and where it's headed.
docparser is beta (0.2.x), published on PyPI as
rc-docparser. For the live state,
follow the repository.
Solid today
- Nine input formats —
.docx,.xlsx,.pdf,.html,.pptx,.epub,.txt,.md,.csv— normalized into a shared Markdown + JSON block schema. - Pluggable PDF backends (
builtin/pymupdf4llm/docling/marker), OCR, and table extraction. - Optional VLM captioning across OpenRouter, OpenAI, Gemini, a local server, or
a fully-local
transformersmodel, with hash-keyed caching. - A lean MIT core with opt-in extras, plus a
docparserCLI. - CI runs lint + mypy + tests on Python 3.10–3.12; tagged releases publish to PyPI via Trusted Publishing (OIDC, no stored token).
Direction
- Broader fidelity. Continued improvements to heading/structure detection and table handling across formats.
- More captioning options. Expanding provider/model coverage and caption schema as VLMs evolve.
- Corpus tooling. Richer corpus-level rollups for RAG and fine-tuning pipelines.
Contributing
git clone https://github.com/Research-Commons/docparser
cd docparser
python -m venv .venv && source .venv/bin/activate
pip install -e ".[all,dev]"
pytest -ra
ruff check src tests
mypyLicensed under MIT.