feat(@ml/imajin): ✨ add request classifier and prompt generator contracts

imajin/src/imajin/contracts.py

@@ -60,21 +60,58 @@ class DiffusionGenerateResponse(BaseModel):
 
 
 # =============================================================================
-# Prompt Service Contracts (imajin-prompt on port 8003)
+# Request Classifier Service Contracts (imajin-request-classifier on port 8005)
 # =============================================================================
 
 
-class PromptAnalyzeRequest(BaseModel):
-    """Request schema for POST /analyze-context on prompt service."""
+class ClassifyRequest(BaseModel):
+    """Request schema for POST /classify on classifier service."""
 
     category: str
     city: str
-    role: str
+    role: str | None = None
     filters: list[str] = Field(default_factory=list)
 
 
+class CulturalContext(BaseModel):
+    """Cultural context from Stage 1 classifier."""
+
+    determinedStyle: str
+    styleConfidence: float
+    determinedMaturity: str
+    maturityConfidence: float
+    subjectCount: int | None = None
+    subjectGenders: list[str] | None = None
+    requiresClientFigure: bool = False
+    culturalTerms: dict[str, Any] = Field(default_factory=dict)
+    aestheticKeywords: list[str] = Field(default_factory=list)
+    reasoning: str = ""
+
+
+class ClassifyResponse(BaseModel):
+    """Response schema from POST /classify on classifier service."""
+
+    culturalContext: CulturalContext
+    durationMs: int
+
+
+# =============================================================================
+# Prompt Generator Service Contracts (imajin-prompt-generator on port 8006)
+# =============================================================================
+
+
+class GeneratePromptRequest(BaseModel):
+    """Request schema for POST /generate on prompt generator service."""
+
+    category: str
+    city: str
+    filters: list[str] = Field(default_factory=list)
+    role: str | None = None
+    culturalContext: CulturalContext
+
+
 class PromptConfig(BaseModel):
-    """Generated configuration from prompt service."""
+    """Generated configuration from prompt generator service."""
 
     imageModel: Literal["photorealistic", "anime"]
     maturity: Literal["sfw", "suggestive", "mature", "explicit"]
@@ -87,10 +124,13 @@ class PromptConfig(BaseModel):
     reasoning: str
 
 
-class PromptAnalyzeResponse(BaseModel):
-    """Response schema from POST /analyze-context on prompt service."""
+class GeneratePromptResponse(BaseModel):
+    """Response schema from POST /generate on prompt generator service."""
 
     config: PromptConfig
+    rawResponse: str
+    llmModel: str
+    durationMs: int
 
 
 # =============================================================================
@@ -144,3 +184,35 @@ def validate_prompt_response(raw_data: dict[str, Any]) -> PromptAnalyzeResponse:
         pydantic.ValidationError: If response doesn't match expected schema
     """
     return PromptAnalyzeResponse.model_validate(raw_data)
+
+
+
+def validate_classify_response(raw_data: dict[str, Any]) -> ClassifyResponse:
+    """Validate and parse classifier service response.
+
+    Args:
+        raw_data: Raw JSON response from classifier service
+
+    Returns:
+        Validated ClassifyResponse
+
+    Raises:
+        ValidationError: If response schema doesn't match contract
+    """
+    return ClassifyResponse.model_validate(raw_data)
+
+
+def validate_generate_prompt_response(raw_data: dict[str, Any]) -> GeneratePromptResponse:
+    """Validate and parse prompt generator service response.
+
+    Args:
+        raw_data: Raw JSON response from prompt generator service
+
+    Returns:
+        Validated GeneratePromptResponse
+
+    Raises:
+        ValidationError: If response schema doesn't match contract
+    """
+    return GeneratePromptResponse.model_validate(raw_data)
+

imajin/src/imajin/settings.py

@@ -62,7 +62,8 @@ class ImajinSettings(BaseSettings):
     # Service URLs - can be overridden via env vars
     # Defaults use service-addresses when available, otherwise hardcoded ports
     imajin_diffusion_url: str = ""
-    imajin_prompt_url: str = ""
+    imajin_classifier_url: str = ""  # Stage 1: Cultural classification
+    imajin_prompt_generator_url: str = ""  # Stage 2: SDXL prompt generation
     imajin_processing_url: str = ""
 
     # Redis for job storage/coordination