refactor(llm): use response_schema with typed enums in all Gemini calls

Pass response_schema to all three generate_content calls so Gemini
constrains its output to valid enum values and correct JSON structure:

- routines.py: _StepOut.action_type Optional[str] → Optional[GroomingAction]
- skincare.py: add _SkinAnalysisOut(PydanticBase) with OverallSkinState,
  SkinType, SkinTexture, BarrierState, SkinConcern enums; add response_schema
- products.py: pass ProductParseResponse directly as response_schema;
  remove NaN/Infinity/undefined regex cleanup, markdown-fence extraction,
  finish_reason logging, and re import — all now unnecessary

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

backend/innercontext/api/products.py

@@ -1,12 +1,8 @@
 import json
-import logging
-import re
 from datetime import date
 from typing import Optional
 from uuid import UUID, uuid4
 
-log = logging.getLogger(__name__)
-
 from fastapi import APIRouter, Depends, HTTPException, Query
 from google.genai import types as genai_types
 from pydantic import ValidationError
@@ -378,37 +374,17 @@ def parse_product_text(data: ProductParseRequest) -> ProductParseResponse:
         config=genai_types.GenerateContentConfig(
             system_instruction=_product_parse_system_prompt(),
             response_mime_type="application/json",
+            response_schema=ProductParseResponse,
             max_output_tokens=16384,
             temperature=0.0,
         ),
     )
-    candidate = response.candidates[0] if response.candidates else None
-    finish_reason = str(candidate.finish_reason) if candidate else "unknown"
     raw = response.text
     if not raw:
-        raise HTTPException(
-            status_code=502,
-            detail=f"LLM returned an empty response (finish_reason={finish_reason})",
-        )
-    # Fallback: extract JSON object in case the model adds preamble or markdown fences
-    if not raw.lstrip().startswith("{"):
-        start = raw.find("{")
-        end = raw.rfind("}")
-        if start != -1 and end != -1:
-            raw = raw[start : end + 1]
-    # Replace JS-style non-JSON literals that some models emit
-    raw = re.sub(r":\s*NaN\b", ": null", raw)
-    raw = re.sub(r":\s*Infinity\b", ": null", raw)
-    raw = re.sub(r":\s*undefined\b", ": null", raw)
+        raise HTTPException(status_code=502, detail="LLM returned an empty response")
     try:
         parsed = json.loads(raw)
     except json.JSONDecodeError as e:
-        log.error(
-            "Gemini parse-text JSON error at pos %d finish_reason=%s context=%r",
-            e.pos,
-            finish_reason,
-            raw[max(0, e.pos - 80) : e.pos + 80],
-        )
         raise HTTPException(status_code=502, detail=f"LLM returned invalid JSON: {e}")
     try:
         return ProductParseResponse.model_validate(parsed)

backend/innercontext/api/routines.py

@@ -108,7 +108,7 @@ class BatchSuggestion(SQLModel):
 
 class _StepOut(PydanticBase):
     product_id: Optional[str] = None
-    action_type: Optional[str] = None
+    action_type: Optional[GroomingAction] = None
     dose: Optional[str] = None
     region: Optional[str] = None
     action_notes: Optional[str] = None

backend/innercontext/api/skincare.py

@@ -5,6 +5,7 @@ from uuid import UUID, uuid4
 
 from fastapi import APIRouter, Depends, File, HTTPException, UploadFile
 from google.genai import types as genai_types
+from pydantic import BaseModel as PydanticBase
 from pydantic import ValidationError
 from sqlmodel import Session, SQLModel, select
 
@@ -70,6 +71,21 @@ class SkinPhotoAnalysisResponse(SQLModel):
     notes: Optional[str] = None
 
 
+class _SkinAnalysisOut(PydanticBase):
+    overall_state: Optional[OverallSkinState] = None
+    skin_type: Optional[SkinType] = None
+    texture: Optional[SkinTexture] = None
+    hydration_level: Optional[int] = None
+    sebum_tzone: Optional[int] = None
+    sebum_cheeks: Optional[int] = None
+    sensitivity_level: Optional[int] = None
+    barrier_state: Optional[BarrierState] = None
+    active_concerns: Optional[list[SkinConcern]] = None
+    risks: Optional[list[str]] = None
+    priorities: Optional[list[str]] = None
+    notes: Optional[str] = None
+
+
 # ---------------------------------------------------------------------------
 # Helpers
 # ---------------------------------------------------------------------------
@@ -148,6 +164,7 @@ async def analyze_skin_photos(
             config=genai_types.GenerateContentConfig(
                 system_instruction=_skin_photo_system_prompt(),
                 response_mime_type="application/json",
+                response_schema=_SkinAnalysisOut,
                 max_output_tokens=2048,
                 temperature=0.0,
             ),