Front · How it works
Define an agent in YAML. Trigger from anywhere.
Each Programmable Agent is a YAML file in your repo. Commit it. Trigger it via GraphQL, the Python SDK, or the Paradime CLI. Read back the results — a PR, a fix, a summary — when the run completes.
01
Commit
Drop a YAML at
.dinoai/agents/<name>.yml and push.02
Trigger
Hit the GraphQL endpoint, use the Python SDK, or fire the CLI.
03
Read back
Pick up the final output — PR, summary, file — when status hits
COMPLETED.04
Compose
Delegate from one agent to another. Build orchestrators, sub-agents, callbacks.
# .dinoai/agents/test-maintainer.yml name: test-maintainer version: 1 role: > dbt™ test specialist focused on improving coverage and catching schema drift before it reaches production. goal: > Review changed models in the latest PR and propose missing not_null / unique / accepted_values tests. tools: mode: allowlist list: [read_file, run_sql_query, github_create_pull_request]
YAML
Agent definition format
versioned, code-reviewed
versioned, code-reviewed
3
Trigger surfaces
GraphQL · Python SDK · CLI
GraphQL · Python SDK · CLI
Isolated
Sandbox per run
scoped tools, audit-trailed
scoped tools, audit-trailed
Back · Seven agents to fork today
Every agent below ships as a YAML in our reference repo. Fork it, tune the role and goal to your stack, deploy. Each one is production-grade — these are the same agents we use internally.
AGENT 01
Pipeline Incident Commander
RoleOrchestrates triage when a pipeline fails. Spawns three sub-agents in parallel — log analysis, warehouse profiling, owner lookup.
OutputStructured incident report posted to
#incidents with root cause, owner, suggested fix.TriggerAirflow / Bolt
on_failure_callbackAGENT 02
Cost Optimizer — Snowflake
RoleScans Snowflake query history, finds the most expensive dbt™ models, proposes refactors.
OutputPull requests with before/after cost projections and rewritten model SQL.
TriggerNightly cron / weekly sweep
AGENT 03
Cost Optimizer — BigQuery
RoleSame idea as #02, tuned for BigQuery's slot and bytes-billed pricing model.
OutputPRs targeting BigQuery-specific patterns — partition pruning, slot allocation.
TriggerNightly cron / weekly sweep
AGENT 04
Doc Backfiller
RoleSweeps your dbt™ project for undocumented models and columns. Generates descriptions grounded in actual upstream SQL.
OutputSingle PR with backfilled
descriptions in .yml files.TriggerWeekly cron or post-merge hook
AGENT 05
Test Maintainer
RoleFinds gaps in dbt™ test coverage — missing
not_null, unique, accepted_values — and proposes additions.OutputFollow-up PR layered on top of the original.
TriggerOn PR open
AGENT 06
Linear Backlog Agent
RoleTriages incoming data tickets, enriches with warehouse context, routes to the right team.
OutputUpdated Linear issue with context, owner, complexity score.
TriggerLinear webhook on
data labelAGENT 07
End-to-End PR Reviewer
RoleReads diffs, runs targeted dbt™ tests against the changed surface, checks for materialization & incremental-strategy regressions.
OutputStructured review comment with verdict, callouts, and suggested follow-ups.
TriggerOn PR open / PR update
+ MORE ON THE WAY
Schema migrator, freshness sweeper, lineage explainer, semantic-layer maintainer.
StatusRoadmap items, shipping over the next 90 days.
Want one?Email product@paradime.io with the agent shape you need.
docs.paradime.io · Programmable Agents reference
Open the docs →
