Skip to content

DataMeshOps

License Python Hackathon

DataMeshOps is an autonomous metadata-aware data remediation and ML governance engine built for the Build with DataHub: The Agent Hackathon. It features a high-performance shared core library (datameshops-core) powering two autonomous agents that deeply integrate with DataHub via the MCP Server.


Agents

🛠️ DataMeshOps-Healer — Grand Prize Target

Challenge: Agents That Do Real Work

Autonomously triages broken data pipelines, traces 3-hop upstream lineage via DataHub MCP, synthesizes non-destructive SQL fixes, and writes remediation knowledge back to the DataHub metadata graph.

→ Full Healer Docs

🔒 DataMeshOps-Guard — Category Prize Target

Challenge: Production ML Agents

Protects production ML models by tracing feature store lineage, detecting target leakage and upstream schema drift, quarantining compromised models, and raising blocking governance incidents via DataHub MCP.

→ Full Guard Docs


Quickstart

git clone https://github.com/rohitdas-ai/datameshops.git
cd datameshops
python3 -m venv venv && source venv/bin/activate
pip install -e .

# Run DataMeshOps-Healer (remediation agent)
python3 -m datameshops_healer.cli run --dataset nyc-taxi

# Run DataMeshOps-Guard (ML protection agent)
python3 -m datameshops_guard.cli audit --model customer_churn_v4

No API key needed — both agents run with deterministic mock fallback. Supply OPENAI_API_KEY or GEMINI_API_KEY for live LLM-powered execution.


Key DataHub Integration Points

MCP Tool Used By Purpose
search Both Find entities in DataHub catalog
get_lineage Both 3-hop upstream/downstream traversal
list_schema_fields Both Schema field inspection
get_dataset_queries Healer Historic SQL query fetch
add_tags Both Write #auto-remediated, #ml-quarantined
add_structured_properties Both Write audit and health properties
update_description Both Append Markdown post-mortems
raise_incident Guard Create blocking governance incidents
update_incident_status Healer Resolve open incidents