[Database] Define Pydantic v2 models and initial relational schema (incidents, models, audits)

[saviornt]

Description

Define strongly-typed Pydantic v2 models for validation/serialization and matching SQLAlchemy table definitions for the core entities: incidents, model registry, policy audits, and FAISS vector metadata.

Type

• Task

Focus Area (pick one)

• Shared Utils & Models

Priority

• High

Acceptance Criteria

• File: shared/src/database/schemas.py
  • Pydantic v2 models: IncidentBase, IncidentCreate, IncidentRead, IncidentInDB
  • Pydantic v2 models: ModelRegistryBase, ModelCreate, ModelInDB
  • Pydantic v2 models: PolicyAuditBase, PolicyAuditCreate, PolicyAuditInDB
  • Pydantic v2 model: FaissVectorMetadata
  • Use field_validator, ConfigDict(from_attributes=True), strict mode where appropriate
• File: shared/src/database/models.py (or inside sqlite.py)
  • SQLAlchemy 2.0 declarative base classes matching above
  • Tables created: incidents, model_registry, policy_audits, faiss_vectors_metadata
  • Appropriate indexes: timestamp ranges, device_mac, incident_type, vector_id
  • Modern mapped_column syntax with typing
• Google-style docstrings on every model and field
• Alembic initial migration stub created (even if not applied yet)
• mypy strict mode passes

Blocker / Dependencies

• [Database] Implement robust async SQLite session factory + connection pooling

Notes / Links

• Reference: tentative schema in docs/database_schema.md
• Related files: shared/src/database/session.py