dell/dell-data-orchestration-engine

Plan and operate governed image-classification workflows in Dell Data Orchestration Engine (Dataloop), including dataset readiness, marketplace model selection, subsets, training, deployment, prediction, and annotation review. Use when a user requests an image classifier or model-assisted image labeling in Dataloop. Do not use for dataset exploration, RAG pipelines, or local model training outside Dataloop.

Plan, build, validate, and query governed document RAG pipelines in Dell Data Orchestration Engine (Dataloop). Use for PDF ingestion, chunking, embeddings, searchable output datasets, RAG pipeline graphs, installation, or bounded semantic queries. Do not use for a generic local vector database request.

Configure, authenticate, validate, and troubleshoot the Dell Data Orchestration Engine (Dataloop) Python SDK MCP integration. Use for Dataloop API-key setup, bundled MCP runtime installation, MCP profile selection, connection validation, TLS settings, or tool-catalog and authorization problems. Do not use for DDAE or DDSE connection setup.

Assess and handle personally identifiable information in authorized Dataloop datasets using aggregate reports, reviewed findings, metadata writeback, and filters. Use for PII inventory, remediation planning, or policy-driven exclusion. Do not use for PII in local files or non-Dataloop sources. Never reveal detected raw PII values or certify a dataset as PII-free.

Discover, profile, and inspect authorized Dataloop projects, datasets, items, and annotations. Use when a user asks what data is available, wants dataset composition or quality evidence, needs a bounded item search, or wants to understand labels without changing data. Do not use for DDAE, Elastic, or unrelated local-file analysis.

Validate, install, run, inspect, and recover authorized Dell Data Orchestration Engine pipelines through Dataloop. Use when a user asks to validate a pipeline, start an approved execution, inspect execution status, stop (pause) or delete an installed pipeline, or terminate an exact stuck execution. Do not use to design a RAG graph, explore datasets, or manage local operating-system processes.