nextcloud-ingest

Nextcloud Ingest

Push Nextcloud files into the epistemic-graph knowledge graph in their richest modality. A single call stores the file's raw bytes as a content-addressed :Blob + :AssetOccurrence node and, when the file is a document or image, extracts its text (read_any / OCR) into a linked :Document for semantic search. This is the package's "maximum ingestion" seam — richer than the declarative folder-listing connector preset.

When to use

• Make a specific Nextcloud file durable in the KG (raw bytes + searchable text).
• Bring a PDF/office doc/image into the graph so it can be retrieved semantically.

When NOT to use

• Just download bytes or reorganize storage → nextcloud-files.
• Manage sharing grants → nextcloud-shares.

Prerequisites & environment

Connect via the mcp-client skill against the nextcloud-agent MCP server. A reachable epistemic-graph engine is required for anything to be stored; with no engine the tool no-ops cleanly (returns status: skipped) so the call is always safe.

Variable	Required	Notes
NEXTCLOUD_URL / NEXTCLOUD_USERNAME / NEXTCLOUD_PASSWORD	✅	WebDAV creds
TLS_PROFILE / TLS_PROFILE_REF	optional	AgentConfig transport profile; peer verification is mandatory

INGESTTOOL gates this tool category.

Tools & actions

This is a single-purpose typed tool (not action-routed).

Tool	Parameter	Purpose
nextcloud_ingest_file	path	Fetch + natively ingest one file into the KG

What it stores

• :Blob + :AssetOccurrence — the raw file bytes, content-addressed (deduped by digest), carrying remote_path, file_id, etag, mime, and size.
• :Document (when the file is pdf/office/txt/csv/html or an image) — the extracted plain text, source_uri nextcloud://<path>, linked back to the file via hasBlob.

Recipes

Ingest a report PDF:

nextcloud_ingest_file(path="Documents/report.pdf")

Returns e.g.:

{"status":"ingested","path":"Documents/report.pdf","asset_id":"media:…","digest":"…","size_bytes":48213,"media_type":"file","doc_id":"nextcloud:document:…"}