The 4 Layer System
The platform is intentionally split into API, Builder, Runtime, and Node System layers so product boundaries remain clear as the stack grows.
4 Layer System
A. Platform API
- Gateway
- Auth and session boundary
- Rate limit and contract enforcement
B. Builder
- Canvas UI
- Draft system
- Validation before runtime handoff
C. Runtime
- Execution engine
- Queue dispatch
- Retry, log, and output persistence
D. Node System
- Node definitions
- Execution mapping
- Future AI-capable node families
Data Flow
Flow JSON
Serializable representation of nodes and edges.
→
Validate
Shape checks, node compatibility, edge sanity.
→
Execution
Runtime steps run against stored flow definition.
→
Step Logs
Per-node observability and error handling.
→
Output
Stored result or downstream side-effect.
Boundary rule
The builder should never become the runtime. It prepares and validates a flow definition. The runtime owns actual execution state.
Database
D1 is the default SQLite-compatible persistence layer for early platform stages.
| Table | Purpose | Typical records |
|---|---|---|
users |
Identity and operator metadata. | Founder, developer, operator, or future workspace user. |
flows |
Stored workflow definitions and drafts. | Flow JSON, metadata, publish state. |
executions |
Runtime run records. | Status, timestamps, result snapshots, retry counters. |
logs |
Per-step and system event logs. | Validation output, node events, failures, debug traces. |
System Boundaries
flow.iai.one
Owns the workflow builder and execution behavior.
app.iai.one
Owns human-facing dashboards, review queues, and product-facing operational surfaces.
mail.iai.one
Owns delivery behavior for communication outputs.
home.iai.one
Owns the core platform vision and narrative boundary for the overall ecosystem.