Workflow engine

This documentation describes the workflow engine: its components, the lifecycle of a run, and how to build and operate it. Pages are ordered for linear reading. Each page uses only terms defined on earlier pages or in the glossary.

Page	Content
Overview	Components and design rationale, without implementation detail
1. Introduction	Purpose, design principles, platform services
2. Components	Architecture diagram and component reference
3. Run lifecycle	One run from start to archival
4. Workflow definitions	Structure, fields, references, registration checks
5. Task types	Interpreter, activity, and worker types
6. The API	Endpoints by group
7. Interpreter	Orchestration, execution loop, timeouts, statuses
8. System tasks	Activities and outbound calls
9. Workers	Task definitions, queues, worker loop, hosting
10. Human gates	Notification, decision, timeouts, write-tier rule
11. Events and schedules	Triggers, event handlers, status events, schedules
12. History and archive	Stores, tables, search, redaction, archival
13. Run control	Operations and test runs
14. Model and tool tasks	Model tasks, trust tiers, agents
15. Authentication and permissions	Token flow, callers, worker scope, tool scope, secrets
16. Monitoring	Tracing, metrics, operator views
17. Deployment	Resources and pipeline
18. Build order	Components and build steps
19. Design decisions	Decisions and reasons
Glossary	Terms