Seed Data

Seed Data loads static reference data (syllabi, rubrics, lookups) into workflow context—available to all actions without repetition.

Configuration

defaults:
  context_scope:
    seed_path:
      exam_syllabus: $file:syllabus.json
      grading_rubric: $file:rubric.yaml

File Reference Syntax

Syntax	Description
$file:filename.json	Load JSON file from seed_data/ directory
$file:filename.yaml	Load YAML file from seed_data/ directory
$file:path/to/file.json	Load from subdirectory within seed_data/

Directory Structure

my_workflow/
├── agent_config/
│   └── my_workflow.yml
├── agent_io/
└── seed_data/           # Seed data files here
    ├── syllabus.json
    └── rubric.yaml

Accessing Seed Data

Use the seed prefix in field references:

prompt: |
  Exam: {{ seed.exam_syllabus.exam_name }}
  {% for skill in seed.exam_syllabus.skills_measured %}
  - {{ skill.skill_area }}
  {% endfor %}

Multiple Seed Files

defaults:
  context_scope:
    seed_path:
      syllabus: $file:exam_syllabus.json
      rubric: $file:grading_rubric.yaml

Access each with its assigned name: {{ seed.syllabus.exam_name }}, {{ seed.rubric.criteria }}.

Seed Data vs Source Data

Aspect	Seed Data	Source Data
Scope	Same for all records	Different per record
Loaded from	seed_data/ directory	agent_io/staging/ directory
Reference	{{ seed.name.field }}	{{ source.field }}

Best Practices

1. Keep seed data focused: Use specific files, not monolithic configs
2. Use descriptive names: grading_criteria not data1
3. Version seed data: Include version/effective_date in seed files

Workflow-Level vs Action-Level

Workflow-level (recommended):

defaults:
  context_scope:
    seed_path:
      syllabus: $file:syllabus.json

Per-action (merged with defaults):

actions:
  - name: specialized_action
    context_scope:
      seed_path:
        special_data: $file:special.json