LoreKeeper MCP

## Why The codebase lacks a true canonical data representation. API responses, OrcBrew entities, and cached data use inconsistent field naming (`key` vs `slug`, `challenge` vs `challenge_rating`, `description` vs `desc`), different type representations (nested objects vs strings), and missing validation for OrcBrew imports. This causes: 1. Transformation logic scattered across API clients, parsers, and repositories 2. OrcBrew data bypassing Pydantic validation entirely 3. Field name mismatches requiring ad-hoc handling at multiple layers 4. "Monster" naming inconsistent with Open5e v2's "Creature" terminology ## What Changes ### Phase 1: Core Model Standardization - Create `src/lorekeeper_mcp/models/` package for canonical models - Rename `Monster` model to `Creature` throughout codebase - Add shared slug/key normalization validator to all models - Standardize description field mapping (`description` → `desc`) - Repositories MUST accept and return only canonical Pydantic models (not dicts) ### Phase 2: OrcBrew Validation - Add OrcBrew-specific Pydantic models in `models/orcbrew/` - Validate OrcBrew entities through Pydantic before caching - Expand `entity_mapper.py` to capture all fields (not just indexed ones) ### Phase 3: Simplified Equipment Models - Simplify `DamageType` to store string with optional object for API compatibility - Add `damage_type_name` computed property on `Weapon` - Flatten `WeaponProperty` structure for simpler storage ### Phase 4: Field Normalization - Standardize challenge rating handling: `challenge`, `challenge_rating`, `challenge_rating_text` → `challenge_rating` (str) + `challenge_rating_decimal` (float) - Add sensible defaults for missing OrcBrew fields ## Impact - **Affected specs**: - `orcbrew-parser` (MODIFIED: validation requirements) - `entity-cache` (MODIFIED: document/model storage) - NEW: `canonical-models` spec - **Affected code**: - `src/lorekeeper_mcp/api_clients/models/` → Move to `src/lorekeeper_mcp/models/` - `src/lorekeeper_mcp/parsers/entity_mapper.py` → Use canonical models - `src/lorekeeper_mcp/api_clients/open5e_v1.py` → Use `Creature` not `Monster` - `src/lorekeeper_mcp/api_clients/open5e_v2.py` → Use `Creature` not `Monster` - `src/lorekeeper_mcp/repositories/` → Update to accept/return canonical models only (not dicts) - `src/lorekeeper_mcp/tools/` → Update imports - All tests referencing `Monster` class - **Breaking changes**: None for external API; internal model location changes