Summary

Implements #88 — import a document that is already indexed in PageIndex Cloud into a local OpenKB knowledge base, with no local PDF and no re-processing:

openkb add --from-pageindex-cloud <DOC_ID>

• Fetches structure + OCR'd page content from PageIndex Cloud by doc_id (get_document / get_page_content), bypassing the local convert → raw → page-count → col.add pipeline entirely.
• Writes the same wiki artifacts as the local long-doc path (shared _write_long_doc_artifacts) and compiles concepts via compile_long_doc.
• Registers a raw-less registry entry typed pageindex_cloud with a synthetic identity key pageindex-cloud:<doc_id> (sha256-keyed; re-import of the same doc-id is idempotent/skipped).
• openkb add runs under the exclusive KB lock, so the cloud-import path is serialized like every other ingest.
• openkb remove on an imported doc cleans up local artifacts only — the user's cloud corpus is never touched (the existing long_pdf gate already excludes the new type; proven by a regression test).
• Dependency pageindex switched to track git+https://github.com/VectifyAI/PageIndex.git@dev for the cloud API surface (lock pins a concrete commit). Follow-up: re-pin to an exact tag once a new PageIndex release is published.

Notes

• Builds on the collision-resistant doc_name groundwork (resolve_doc_name, registry path identity), now in main. Rebased onto main after fix: collision-resistant doc_name — same-stem documents no longer overwrite each other #96 merged; base is main.
• Known limitation: cloud OCR returns text only, so imported docs have no per-page images (images: []). Concept compilation reads the summary, so output quality is unaffected.

Test Plan

Automated (full suite green — 742 passed on the rebased branch):

• uv run --extra dev pytest -q
• Targeted: uv run --extra dev pytest tests/test_converter.py tests/test_indexer.py tests/test_add_command.py tests/test_remove.py -v

Manual (requires PAGEINDEX_API_KEY and a real cloud doc-id):

• export PAGEINDEX_API_KEY=... then openkb add --from-pageindex-cloud <DOC_ID> → wiki/summaries + wiki/sources written, concepts compiled, doc appears in openkb list.
• Re-run the same command → [SKIP] Already imported (idempotent).
• openkb add foo.pdf --from-pageindex-cloud X → errors "not both"; openkb add with neither → "Provide a PATH...".
• Unset PAGEINDEX_API_KEY then import → clear error, nothing written.
• openkb remove <doc> on the imported doc → local artifacts removed, cloud doc still present in PageIndex.