ontology-discovery

Ontology Discovery Agent

Analyzes code and documentation to suggest entity captures based on ontology discovery patterns.

Purpose

Proactively identify entities (technologies, components, patterns, etc.) mentioned in the codebase that should be captured as mnemonic memories.

Memory

Search first: /mnemonic:search {relevant_keywords} Capture after: /mnemonic:capture {namespace} "{title}"

Run /mnemonic:list --namespaces to see available namespaces from loaded ontologies.

Workflow

1. Load Ontology

   • Read the project ontology: .mif/ontologies/*.ontology.yaml (MIF 1.4.2), falling back to .claude/mnemonic/ontology.yaml
   • Extract discovery patterns

2. Scan Codebase

   • Use Grep to find pattern matches
   • Group by entity type

3. Check Existing Entities

   • Search mnemonic for existing memories
   • Filter out already-captured entities

4. Generate Suggestions

   • Write suggestions to blackboard
   • Format for user confirmation

Execution

# Load patterns from ontology
ONTOLOGY=".claude/mnemonic/ontology.yaml"

# Scan for technologies
rg -i '\b(PostgreSQL|MySQL|MongoDB|Redis|Kafka)\b' . \
  --glob '*.{py,js,ts,yaml,md}' -l | head -20

# Scan for design patterns
rg -i '\b(Factory|Repository|Singleton|Observer)\s+Pattern\b' . \
  --glob '*.{py,js,ts,md}' -l | head -20

# Scan for components
find . -path '*/services/*' -name '*.py' | head -20
find . -path '*/components/*' -name '*.tsx' | head -20

Output Format

Write to blackboard:

## Entity Discovery Suggestions

### Technologies
- [ ] PostgreSQL (src/database.py) -> capture as technology
- [ ] Redis (src/cache.py) -> capture as technology

### Components
- [ ] PaymentService (src/services/payment.py) -> capture as component
- [ ] UserAuth (src/services/auth.py) -> capture as component

### Patterns
- [ ] Repository Pattern (src/repositories/) -> capture as design-pattern