curador

Yes, you can batch classify manuals and PDFs by running a recursive directory scan that applies deterministic heuristics first, then uses Claude for ambiguous items. The pipeline generates a human review report and requires approval before routing files to memory or git repositories.

Batch processing handles sensitive files by blocking them from standard ingestion and routing them to a vault-pending location. The system enforces safety approvals and quarantines discardable items before any commit or memory update occurs.

Yes, LGPD consent controls require a recorded opt-in entry before scanning another developer's local folders. The system prevents directory triage and batch processing of other users' paths without this explicit consent.

The best way to classify ambiguous items during batch document ingestion is a two-step pipeline applying deterministic heuristics first, then invoking Claude only for unresolved items. This ensures safe ingestion outputs while minimizing AI processing overhead.

Yes, you can use Node.js scripts for recursive directory triage and file classification. The pipeline relies on Node and MJS modules to execute deterministic heuristics, track file states in JSONL files, and route outputs to git or MCP pathways.

Batch processing requires an approval gate to ensure policy-compliant knowledge ingestion. It blocks sensitive material and prevents automatic commits, requiring human review of the generated report before any move-to-repo or apply step happens.