not_null_fct_transactions_merchant failed (23 rows)
        │
        ▼
Agent fetches failing rows → reads fct lineage → traces to int_ → traces to stg_ → checks raw source
        │
        ▼
Root cause: 23 rows in raw.bank_transactions have NULL narration.
Merchant extraction returns NULL when narration is NULL.
Fix: Add COALESCE(narration, '') in stg_bank__transactions.
Severity: HIGH

python agent.py \
  --test not_null_fct_transactions_merchant \
  --model fct_transactions \
  --column merchant \
  --verbose

python server.py   # starts on port 5051

curl -X POST http://localhost:5051/investigate \
  -H "Content-Type: application/json" \
  -d '{"test_name": "not_null_fct_transactions_merchant", "model": "fct_transactions", "column": "merchant"}'

claude mcp add -s user \
  -e GCP_PROJECT=your-project \
  -e BQ_LOCATION=asia-south1 \
  -e DBT_MANIFEST_PATH=/path/to/dbt_bank/target/manifest.json \
  -e DBT_RUN_RESULTS_PATH=/path/to/dbt_bank/target/run_results.json \
  -e GEMINI_API_KEY=your-key \
  dbt-investigator \
  -- /path/to/venv/bin/python /path/to/mcp_server.py

openclaw mcp set dbt-investigator '{
  "command": "/path/to/venv/bin/python",
  "args": ["/path/to/mcp_server.py"],
  "cwd": "/path/to/data-quality-agent",
  "env": {
    "GCP_PROJECT": "your-project",
    "GEMINI_API_KEY": "your-key",
    "DBT_MANIFEST_PATH": "/path/to/manifest.json",
    "DBT_RUN_RESULTS_PATH": "/path/to/run_results.json"
  }
}'
openclaw mcp probe   # → dbt-investigator: 3 tools ✔

git clone https://github.com/ARAVINDHRAJA123/data-quality-agent.git
cd data-quality-agent

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# Auth
gcloud auth application-default login

# Set environment
export GCP_PROJECT=your-project
export BQ_LOCATION=asia-south1
export DBT_MANIFEST_PATH=/path/to/dbt_bank/target/manifest.json
export DBT_RUN_RESULTS_PATH=/path/to/dbt_bank/target/run_results.json

# LLM (pick one)
export GEMINI_API_KEY=your-key      # free
export ANTHROPIC_API_KEY=your-key  # paid

cd /path/to/dbt_project && dbt compile
# manifest.json is now at target/manifest.json

data-quality-agent/
├── agent.py          ← agentic investigation loop (Claude + Gemini)
├── server.py         ← Flask webhook server
├── mcp_server.py     ← FastMCP server (any MCP client)
├── report.py         ← incident report formatter
├── tools/
│   ├── bq_tools.py   ← BigQuery: failing rows, queries, freshness
│   └── dbt_tools.py  ← manifest: lineage, SQL, test results
├── tests/
│   └── test_tools.py ← 11 unit tests (no BQ/LLM needed)
├── reports/          ← saved incident reports (markdown)
└── requirements.txt