Configuration
Configuration is split across the stack on purpose.
Product-Level Rule
Section titled “Product-Level Rule”Static process configuration and deployment settings live in config files and environment variables. Per-workflow and per-role behavior lives in platform-managed records and task contracts.
Main Places To Look
Section titled “Main Places To Look”- product stack environment and image config in
agirunner/.env - top-level stack topology in
agirunner/docker-compose.yml - platform environment and startup config in
agirunner-platformwhen you are working on platform internals directly - runtime process config in
agirunner-runtime/configs/runtime-config.example.yaml - model, tool, and execution policy in the dashboard and platform API
Configuration Layers
Section titled “Configuration Layers”Use this mental model:
- deploy-time configuration for process startup, ports, credentials, database URLs, and image references
- platform-managed configuration for roles, models, tools, workspaces, playbooks, and settings edited through the dashboard or API
- task-level contracts for the exact execution posture delivered to a runtime on claim
What Not To Do
Section titled “What Not To Do”Do not treat the runtime as the place where workflow meaning is configured. Workflow behavior belongs in the platform and in authored playbooks. The runtime should consume explicit contracts, not infer business rules from loose configuration.