feat: update to June 2026 models (Claude Sonnet 4.6, Gemini 3.5 Flash), add glossary button, and implement cost factor quota & vision fallback

services/auth_service.py

@@ -501,19 +501,21 @@ def check_usage_limits(user: User) -> Dict[str, Any]:
     }
 
 
-def record_usage(user_id: str, pages_count: int, use_credits: bool = False) -> bool:
-    """Record document translation usage"""
+def record_usage(
+    user_id: str, pages_count: int, use_credits: bool = False, cost_factor: int = 1
+) -> bool:
+    """Record document translation usage with optional cost factor depending on AI model"""
     user = get_user_by_id(user_id)
     if not user:
         return False
 
     updates = {
-        "docs_translated_this_month": user.docs_translated_this_month + 1,
-        "pages_translated_this_month": user.pages_translated_this_month + pages_count,
+        "docs_translated_this_month": user.docs_translated_this_month + cost_factor,
+        "pages_translated_this_month": user.pages_translated_this_month + (pages_count * cost_factor),
     }
 
     if use_credits:
-        updates["extra_credits"] = max(0, user.extra_credits - pages_count)
+        updates["extra_credits"] = max(0, user.extra_credits - (pages_count * cost_factor))
 
     result = update_user(user_id, updates)
     return result is not None

services/translation_service.py

@@ -903,11 +903,12 @@ RULES:
                 else "image/png"
             )
 
-            # Determine a vision model. If the current model doesn't support vision,
-            # use a fast vision fallback model like google/gemini-2.0-flash-001
             vision_model = self.model
-            if "deepseek" in vision_model:
-                vision_model = "google/gemini-2.0-flash-001"
+            vision_supported_keywords = ["gpt-4", "claude", "gemini", "fable", "pixtral", "llama-3.2", "grok-2"]
+            has_vision = any(kw in vision_model.lower() for kw in vision_supported_keywords)
+            
+            if not has_vision:
+                vision_model = "google/gemini-3.5-flash"
 
             session = self._get_session()
             payload = {
@@ -1083,7 +1084,10 @@ ADDITIONAL CONTEXT AND INSTRUCTIONS:
         try:
             import openai
 
-            client = openai.OpenAI(api_key=self.api_key)
+            client_kwargs = {"api_key": self.api_key}
+            if self.base_url:
+                client_kwargs["base_url"] = self.base_url
+            client = openai.OpenAI(**client_kwargs)
 
             # Read and encode image
             with open(image_path, "rb") as img_file:
@@ -1097,8 +1101,20 @@ ADDITIONAL CONTEXT AND INSTRUCTIONS:
                 else "image/png"
             )
 
+            # Determine a vision model. If the current model doesn't support vision,
+            # use a fast vision fallback model
+            vision_model = self.model
+            vision_supported_keywords = ["gpt-4", "claude", "gemini", "fable", "pixtral", "llama-3.2", "grok-2"]
+            has_vision = any(kw in vision_model.lower() for kw in vision_supported_keywords)
+            
+            if not has_vision:
+                if self.base_url and "openrouter.ai" in self.base_url:
+                    vision_model = "google/gemini-3.5-flash"
+                else:
+                    vision_model = "gpt-4o-mini"
+
             response = client.chat.completions.create(
-                model=self.model,  # gpt-4o and gpt-4o-mini support vision
+                model=vision_model,
                 messages=[
                     {
                         "role": "user",
@@ -1121,7 +1137,7 @@ ADDITIONAL CONTEXT AND INSTRUCTIONS:
 
             return response.choices[0].message.content.strip()
         except Exception as e:
-            logger.warning("openai_vision_error", error_type=type(e).__name__)
+            logger.warning("openai_vision_error", error_type=type(e).__name__, error=str(e))
             return ""