Execution
The execution layer orchestrates LLM calls, data transformations, and validated outputs based on your workflow configuration.
Core Configuration
| Setting | Purpose | Options |
|---|---|---|
run_mode | Processing strategy | online (real-time), batch (cost-optimized) |
granularity | Processing scope | record (per-item), file (all items) |
guard | Conditional execution | Skip or filter based on data |
retry | Error handling | Automatic retry for transient failures |
versions | Parallel variations | Multiple iterations with different parameters |
Quick Reference
| Need | Configuration |
|---|---|
| Lower costs, can wait 24h | run_mode: batch |
| Immediate responses | run_mode: online |
| Per-item transformations | granularity: record |
| Aggregation/exports | granularity: file |
| Conditional execution | guard: { condition: "...", on_false: filter } |
| Handle transient failures | retry: { max_attempts: 3, on_exhausted: return_last } |
Schema Analysis
Analyze workflow schemas and field dependencies before making API calls:
agac schema -a my_workflow
Shows input/output schemas for each action and helps catch field reference errors.
See Also
- Run Modes — Batch vs online execution
- Granularity — Record vs file processing
- Guards — Conditional action execution
- Retry — Automatic error handling
- Version Actions — Parallel processing with iterations
- Workflow Dependencies — Cross-workflow orchestration
- Context Handling — Data flow between actions