Programmable Agents quick start

Get a Programmable Agent running in three steps.

There are two ways to create an agent:

In the app: build it on the Agents page and choose at deploy time whether it lives only in your workspace (live straight away, no repo file) or is committed to your repo as YAML through a pull request. See Creating Agents in the App.
As YAML in your repo: commit a file under .dinoai/agents/, which is what this quick start walks through.

Both kinds of agent are triggered the same way and appear in the same agent list.

Create an Agent YAML

Commit this file at .dinoai/agents/data-quality-checker.yml in your repo:

name: data-quality-checker
version: 1

role: >
  dbt™ Data Quality Specialist focused on test coverage and documentation
  completeness across analytics models.

goal: >
  Ensure all dbt™ models have appropriate tests and are fully documented
  before promotion to production.

backstory: >
  You have deep knowledge of dbt™ best practices and data quality patterns.
  You are methodical and prefer to verify before acting.

tools:
  mode: allowlist
  list:
    - read_file
    - search_files_and_directories
    - ripgrep_search
    - run_sql_query
    - list_all_snowflake_databases
    - list_all_columns_in_snowflake_table
    - post_slack_message

slack:
  channel: "#data-quality-alerts"

Trigger the Agent from Bolt

The simplest way to run an agent is on a Bolt schedule. In the Bolt schedule editor, click Add command, pick Run Paradime DinoAI Agent, and choose your agent plus a task:

Agent Name — pick data-quality-checker from the dropdown (populated from .dinoai/agents/*.yml on the schedule’s git branch).
Task — describe what to do, for example Audit all staging models for missing tests. This becomes the agent’s opening message.
Save Command, set the trigger (cron, on-merge, on-completion), and deploy.

If you manage schedules as code, add the equivalent CLI command directly:

schedules:
  - name: nightly data quality
    slug: nightly-data-quality-f6g7h8
    environment: production
    git_branch: main
    commands:
      - paradime dinoai --agent=data-quality-checker --message="Audit all staging models for missing tests and post a summary to the data-quality-alerts Slack channel"
    schedule: "0 2 * * *"
    timezone: UTC

See Triggering from Bolt for the full walkthrough.

Watch the Run

The agent’s messages stream into the Bolt command’s stdout, so you can follow progress live in the Bolt run logs. The Bolt run inherits the agent’s outcome: it exits 0 when the agent reaches COMPLETED and 1 on FAILED or EXPIRED. If a Slack channel is configured in the YAML, the same messages post to that thread.

Prefer to trigger from your own code?

You can also trigger agents programmatically over the Paradime API or Python SDK, instead of from Bolt:

from paradime import Paradime

paradime = Paradime(
    api_endpoint="API_ENDPOINT",
    api_secret="prdm_cmp_...",   # your account API key
    workspace_uid="WORKSPACE_UID",
)

run = paradime.dinoai_agents.trigger_run_and_wait(
    agent="data-quality-checker",
    message="Focus on stg_orders",
)

for msg in run.messages:
    print(f"[{msg.role}] {msg.content}")

See API Reference for the GraphQL and SDK details, including polling and multi-turn messaging.

Next Steps

YAML Configuration — Learn all available fields and options
API Reference — Full GraphQL mutation and query reference
Tools Reference — See all tools available to your agents
Examples — See tutorials on how to get started deploying agents

Slack Bot Agent