chore(deps): remove unused proprietary dep pymupdf-layout

[cbcoutinho]

Summary

Closes #725.

• Remove pymupdf-layout>=1.26.6 from pyproject.toml and uv.lock. The package is Artifex commercial-only proprietary (its wheel ships only a one-line COPYING: "Commercial license. See artifex.com for details.") and is incompatible with the project's AGPL build. It was declared as a runtime dependency but unused — not imported anywhere in nextcloud_mcp_server/ or tests/.
• Delete tools/parse-doc.py, an unused dev scratch script that was the only caller of pymupdf.layout.activate(). Per the explicit warning in nextcloud_mcp_server/document_processors/pymupdf.py:11-14, activating layout breaks pymupdf4llm.to_markdown(page_chunks=True) (upstream pymupdf4llm#323), so the script was inconsistent with main code anyway.

Test plan

• uv lock regenerates cleanly (also drops transitive networkx)
• uv run ruff check passes
• uv run pytest tests/unit/ — 564 passed
• License-check CI green once chore(licenses): add dependency license policy + CI check #724 lands

---

This PR was generated with the help of AI, and reviewed by a Human

[claude]

Claude finished @cbcoutinho's task in 1m 11s —— View job

---

PR Review: chore(deps): remove unused proprietary dep pymupdf-layout

Tasks:

• Gather context and read changed files
• Search for remaining references to removed packages
• Review pyproject.toml changes
• Review uv.lock changes
• Review deleted tools/parse-doc.py
• Post final review

---

Review Summary

This is a clean, well-scoped dependency removal. The changes are correct and complete. No issues found.

pyproject.toml

The single line removal of "pymupdf-layout>=1.26.6" from the dependencies list is correct. The package was indeed unused — no import of pymupdf.layout or pymupdf_layout exists anywhere in nextcloud_mcp_server/ or tests/.

pymupdf and pymupdf4llm remain as runtime deps, which are actively used in nextcloud_mcp_server/document_processors/pymupdf.py.

uv.lock

Three blocks correctly removed:

1. networkx 3.6 — was only a transitive dep of pymupdf-layout
2. pymupdf-layout 1.26.6 entry with its 5 platform wheels
3. The pymupdf-layout entries in nextcloud-mcp-server's dependencies and requires-dist sections

Other transitive deps of pymupdf-layout (onnxruntime, numpy, pyyaml) are correctly retained since they're still needed by other packages (e.g. fastembed).

tools/parse-doc.py

The deleted script was the sole caller of pymupdf.layout.activate(). The deletion is appropriate:

• It was an unused dev scratch script (not referenced anywhere else in the repo)
• It directly contradicted the explicit warning at nextcloud_mcp_server/document_processors/pymupdf.py:11-14, which instructs not to call pymupdf.layout.activate() as it breaks pymupdf4llm.to_markdown(page_chunks=True)
• The tools/ directory still contains rag_eval_cli.py, so the directory itself is not empty

Cross-checks

• No remaining references to pymupdf-layout, pymupdf.layout, or parse-doc exist in the codebase (only CHANGELOG.md and the documentary comment in pymupdf.py:11, both of which are intentional)
• networkx has no other callers — its removal from the lock file is correct

License Concern (mentioned in PR)

The PR correctly identifies the license incompatibility: pymupdf-layout ships a commercial-only COPYING note, which is incompatible with the project's AGPL-3.0-only license. The pending license-check CI (#724) will formally enforce this going forward.

Verdict: LGTM. This is a straightforward and correct cleanup. The PR description is accurate and the implementation matches the stated intent.