Feat: Implement token tracking, soft delete, and Admin UI improvements

app/api/endpoints/gemini.py

@@ -1,5 +1,8 @@
 from fastapi import APIRouter, Depends, Request
-from app.api.deps import get_api_key
+from app.api.deps import get_api_key, get_current_module
+from app.models.module import Module
+from sqlalchemy.orm import Session
+from app.core.database import get_db
 from app.core.limiter import limiter
 from app.core.config import settings
 from pydantic import BaseModel
@@ -26,16 +29,29 @@ def get_gemini_client():
 async def gemini_chat(
     request: Request,
     chat_data: LLMRequest,
-    api_key: str = Depends(get_api_key)
+    api_key: str = Depends(get_api_key),
+    module: Module = Depends(get_current_module),
+    db: Session = Depends(get_db)
 ):
     client = get_gemini_client()
     
     try:
         if not client:
+             # Mock response
+             response_text = f"MOCK: Gemini response to '{chat_data.prompt}'"
+             if module:
+                 # Estimate tokens for mock
+                 prompt_tokens = len(chat_data.prompt) // 4
+                 completion_tokens = len(response_text) // 4
+                 module.ingress_tokens += prompt_tokens
+                 module.egress_tokens += completion_tokens
+                 module.total_tokens += (prompt_tokens + completion_tokens)
+                 db.commit()
+
              return {
                 "status": "mock",
                 "model": "gemini",
-                "response": f"MOCK: Gemini response to '{chat_data.prompt}'"
+                "response": response_text
             }
             
         # Using the async generation method provided by the new google-genai library
@@ -45,6 +61,20 @@ async def gemini_chat(
             contents=chat_data.prompt
         )
         
+        # Track usage if valid module
+        if module:
+            # Estimate tokens since metadata might vary
+            # 1 char ~= 0.25 tokens (rough estimate if exact count not returned)
+            # Gemini response usually has usage_metadata
+            usage = response.usage_metadata
+            prompt_tokens = usage.prompt_token_count if usage else len(chat_data.prompt) // 4
+            completion_tokens = usage.candidates_token_count if usage else len(response.text) // 4
+            
+            module.ingress_tokens += prompt_tokens
+            module.egress_tokens += completion_tokens
+            module.total_tokens += (prompt_tokens + completion_tokens)
+            db.commit()
+        
         return {
             "status": "success",
             "model": "gemini",