Seed Data

AEGIS provides multiple seed scripts that populate the database and knowledge graph with sample data for development and demos. This page documents each seed script, what it creates, and how to run it.

Seed Scripts Overview

Running All Seeds

After starting infrastructure with docker compose up -d and ensuring all services are running:

# 1. Seed the knowledge graph (requires KG service running on port 8003)
curl -X POST http://localhost:8003/seed
 
# 2. Seed skills, checklists, rules, and demo data
cd services/orchestration-engine
poetry run python -m orchestration.seed_skills
poetry run python -m orchestration.seed_checklists
poetry run python -m orchestration.seed_rules
poetry run python -m orchestration.seed_demo_data

Knowledge Graph Seed

Endpoint: POST http://localhost:8003/seed

File: services/knowledge-graph-service/src/knowledge_graph/seed.py

Creates a realistic Permian Basin scenario in the Apache AGE oilgas graph. This is the most comprehensive seed script, building the full entity topology.

Entities Created

Relationships Created

The seed data connects entities with edges including:

• OPERATED_BY — wells, leases, facilities to operators
• LOCATED_IN — wells to leases and fields
• LOCATED_ON — wells to wellpads
• COMPLETED_IN — wells to formations
• GOVERNED_BY — wells to regulations
• OFFSET_TO — bidirectional between nearby wells
• FLARES_AT — authorizations to wells/leases
• PRODUCES_TO — wells to facilities
• FEEDS_INTO — facilities to pipeline routes
• CONNECTS_TO — infrastructure projects to leases

Demo Scenarios

The seed data supports three demo scenarios:

1. Compliance blast radius: Mitchell Ranch 1H -> Pad A -> Tank Battery -> Gathering Line
2. Infrastructure failure: Delaware Basin Compressor -> 3 wells, third-party operator
3. Planning surface: Howard R-32 expiring -> Howard Connector -> Spraberry system

Well Production Data

Every well includes full Form PR production data with 12 months of history:

{
    "production_oil_bbls": 5100.0,
    "production_gas_mcf": 10800.0,
    "production_casinghead_mcf": 420.0,
    "production_condensate_bbls": 85.0,
    "production_water_bbls": 2400.0,
    "gas_sold_mcf": 9200.0,
    "gas_used_on_lease_mcf": 380.0,
    "gas_flared_mcf": 1100.0,
    "gas_vented_mcf": 120.0,
    "producing_days": 30,
    "allocation_method": "test",
    "form_pr_period": "2026-02",
    "production_history_12m": [
        {"period": "2026-02", "oil_bbls": 5100, "gas_mcf": 10800, ...},
        {"period": "2026-01", "oil_bbls": 5300, "gas_mcf": 11200, ...},
        ...
    ]
}

Skill Registry Seed

Command: poetry run python -m orchestration.seed_skills (from services/orchestration-engine/)

File: services/orchestration-engine/src/orchestration/seed_skills.py

Populates the skills table with the three-tier skill definitions for Rule 37 and Rule 32 agents, plus checklist-item-level skills.

Skills Seeded

Rule 37 Skills:

Rule 32 Skills:

Checklist-Item Skills (mapped to specific checklist items):

The seed also creates skills that correspond 1:1 to checklist items:

• Rule 37: 11 skills (r37-exception-type, r37-field-rules, r37-offset-identification, etc.)
• Rule 32: 10 skills
• Form PR: 8 skills
• Flaring Monitor: 6 skills

Each checklist-item skill specifies which SSE events it emits (e.g., DATA_TABLE_UPDATE, FORM_FIELD_UPDATE, ARTIFACT_GENERATED, CHECKLIST_ITEM_UPDATE).

Three-Tier Architecture

Each skill has three injection tiers:

• Tier 1 manifest (~50 tokens): name, description, triggers, domain_tags
• Tier 2 definition (~200-800 tokens): full spec with steps, requirements, output format
• Tier 3 artifact refs: references to large content artifacts in skill_artifacts

The agent also seeds corresponding records in the agents table for each of the four agents (rule37, rule32, compliance-monitor, flaring-monitor) along with their configurations.

Checklist Template Seed

Command: poetry run python -m orchestration.seed_checklists (from services/orchestration-engine/)

File: services/orchestration-engine/src/orchestration/seed_checklists.py

Creates checklist templates for all four compliance domains.

Templates Seeded

Checklist Item Structure

Each item defines:

• index — Order within the checklist
• name — Human-readable item name
• item_type — data, document, form, artifacts, or validation
• completion_method — auto (agent-only) or hybrid (agent + human)
• agent_can — What the agent is capable of doing for this item
• user_must — What the human user is responsible for
• required_for_submission — Whether this item blocks filing submission

Example Rule 37 checklist items:

Rule Version Seed

Command: poetry run python -m orchestration.seed_rules (from services/orchestration-engine/)

File: services/orchestration-engine/src/orchestration/seed_rules.py

Populates the rule_versions table with statewide and field-specific rules.

Rules Seeded

Each rule version includes structured rule_data as JSONB with the actual regulatory parameters (distances, thresholds, deadlines, etc.).

Demo Data Seed

Command: poetry run python -m orchestration.seed_demo_data (from services/orchestration-engine/)

File: services/orchestration-engine/src/orchestration/seed_demo_data.py

Seeds realistic compliance data for a 15-minute demo walkthrough.

What Gets Seeded

• 25 wells with Permian Basin API numbers and compliance status distribution: 60% green (compliant), 25% amber (action needed), 15% red (overdue).
• Compliance status records in the compliance_status table for 4 domains per well (Rule 37, Rule 32, Form PR, Flaring).
• Pre-built Rule 37 checklist at ~40% completion for Mitchell Ranch 1H.
• Flare events with burn rate data for the flaring dashboard.
• Chart-ready historical data for compliance trends and deadline distributions.

Compliance Distribution

The demo data creates a realistic distribution of compliance statuses: