From 7e75d15a27765f8e6263fee07d03e5a7fe060225 Mon Sep 17 00:00:00 2001
From: adriano <adriano.andrade@hittelco.com>
Date: Wed, 24 Sep 2025 17:48:51 -0300
Subject: [PATCH] feat: updated diarization report using llm like gemini flash
 2.5

---
 backend/app/services/report_service.py | 50 +++++++++++++++++++++++---
 1 file changed, 46 insertions(+), 4 deletions(-)

diff --git a/backend/app/services/report_service.py b/backend/app/services/report_service.py
index 10715e4..e2d9d84 100644
--- a/backend/app/services/report_service.py
+++ b/backend/app/services/report_service.py
@@ -129,6 +129,34 @@ class TranscriptionReportService:
         # Executa agregação principal
         self.mongo_results = list(collection.aggregate(pipeline))
         self.unique_ids = [doc["_id"] for doc in self.mongo_results] 
+        
+        #  TEST
+        # print("======> self.mongo_results: ", self.mongo_results)
+        
+        # if rowMongo := next((m for m in self.mongo_results), None): 
+        #     print("============>", rowMongo["totalCost"])
+            
+        #     stt = rowMongo.get('usageByType', {}).get('stt', {}) 
+                    
+        #     if not stt:            
+        #         stt = rowMongo.get('usageByType', {}).get('input-audio', {})   
+                
+        #         # minutes
+        #         usageMinute = round(stt.get('usage', 0) / 1920, 2) 
+                
+        #         # seconds
+        #         token_by_second = 1920 / 60
+        #         usageSeconds = round(stt.get('usage', 0) / token_by_second)
+                 
+                                    
+        #         print("======> stt_model: ", stt.get('product', 'unknown'))
+        #         print("======> stt_provider: ", stt.get('provider', 'unknown'))  
+        #         print("======> stt_cost: ", rowMongo["totalCost"])   
+        #         print("======> stt_usage_minute: ", f"{usageMinute:.2f}")  
+        #         print("======> stt_usageSeconds: ", f"{usageSeconds}") 
+
+        # exit(1)
+        # END TEST
 
         # Pipeline para contagem total
         count_pipeline = [
@@ -192,10 +220,24 @@ class TranscriptionReportService:
                     row["llm_provider"] = token_output.get('provider','unknown')    
                     
                     stt = rowMongo.get('usageByType', {}).get('stt',{})
-                    row["stt_model"] = stt.get('product', 'unknown')
-                    row["stt_provider"] = stt.get('provider', 'unknown')
-                    row["stt_cost"] = stt.get('usageCost', 0)
-                    row["stt_usage"] = stt.get('usage', 0)
+                    
+                    if not stt:                        
+                        stt = rowMongo.get('usageByType', {}).get('input-audio',{})
+                        
+                        # seconds 
+                        token_by_second = 1920 / 60 # Cobrança de token por minuto do modelo gemini flash 2.5 input de audio
+                        usageSeconds = round(stt.get('usage', 0) / token_by_second)
+                                              
+                        row["stt_model"] = stt.get('product', 'unknown')
+                        row["stt_provider"] = stt.get('provider', 'unknown')  
+                        row["stt_cost"] = rowMongo["totalCost"]
+                        row["stt_usage"] = usageSeconds
+                        
+                    else:    
+                        row["stt_model"] = stt.get('product', 'unknown')
+                        row["stt_provider"] = stt.get('provider', 'unknown')
+                        row["stt_cost"] = stt.get('usageCost', 0)
+                        row["stt_usage"] = stt.get('usage', 0)
 
                     row["total_min"] = f"{(int(row['total_billsec']) / 60):.2f}"