Operational Intelligence

Operational intelligence here means deterministic, structured signals derived from dbt artifacts—graphs, execution timelines, readiness snapshots, and discovery—without requiring an LLM to interpret raw JSON.

Design principles

1. Artifact-driven — answers come from parsed manifest.json / run_results.json, not from chat.
2. Deterministic — same inputs produce the same structured outputs for CI and coding agents.
3. Shared substrate — a common analysis engine backs the CLI, MCP, and web interfaces for different workflows.
4. Coding-agent-friendly — JSON, field filtering, and MCP resident caches reduce token churn.

Examples (no LLM required)

CI gate after dbt run

dbt-tools status --dbt-target ./target --json

Readiness full means manifest and run results are both present—see Check run health.

Resolve “the orders model” without guessing unique_id

dbt-tools discover --dbt-target ./target "orders" --json

Ranked matches include reasons you can paste into tickets or coding-agent context—same contract as web Inventory and header search (Discovery parity).

Investigate slow nodes in the browser

npx @dbt-tools/web --dbt-target ./target

Use Timeline (?view=timeline) and Runs (?view=runs) after artifacts are confirmed—see Investigate slow runs.

Where to read more

• Discovery parity
• Deep links
• Local and remote artifacts
• Ecosystem at a glance