YAML Definition
One file, one flow
Workflows are described in YAML: variables, steps, output. Steps pass data via
{{...}} templates — query results feed directly into AI, and AI conclusions
drive the next step.
-
Step types:
query/ai/condition/foreach - Variables: string / number / connection, with defaults and required validation
- Template syntax: single values, row sets, full table JSON all referenceable
- Error strategies: abort / skip / fallback; configurable timeouts
Cross-Database Orchestration
One flow, multiple databases
Each step can bind to a different connection and database. For example: query orders from PostgreSQL, fetch logistics from MySQL, then let AI summarize into a report.
steps:
- type: query
id: get_orders
connection: "{{pg_conn}}"
sql: |
SELECT order_id, amount FROM test_orders
WHERE uid = '{{uid}}'
- type: condition
id: check_orders
if: "steps.get_orders.rows_count > 0"
then_steps:
- type: query
id: get_logistics
connection: "{{mysql_conn}}"
sql: |
SELECT carrier, tracking_no, status
FROM test_logistics
WHERE order_id IN ({{steps.get_orders.rows.*.order_id}})
output:
format: markdown
template: |
## Orders & logistics for user {{uid}}
{{steps.get_orders.result}}
{{steps.get_logistics.result}}
Four Ways to Run
Run from anywhere
- Connection window AI sidebar: select workflow, fill variables, view results
- Standalone Workflow window: tabbed management for multiple flows, full-screen focus
- External AI tools: list and run your flows from compatible external AI tools
- AI generation: let AI generate an executable workflow from your requirements
