[NOTICKET] Fix: model cost display in langfuse

externals/storages/azure_blob.py

@@ -10,28 +10,6 @@ from utils.logger import get_logger
 logger = get_logger("azure-blob")
 
 
-# async def get_blob_service_client() -> BlobServiceClient:
-#     try:
-#         # credential = AzureSasCredential(AzureBlobConstants.SAS_KEY)
-#         credential = AzureBlobConstants.SAS_KEY
-#         logger.info(f"✅ Initialized Azure AI Cred: Using **Azure SAS Key Credential**")
-        
-#         _blob_service_client = BlobServiceClient(
-#             account_url=AzureBlobConstants.ENDPOINT,
-#             credential=credential,
-#         )
-#         return _blob_service_client
-#     except Exception as E:
-#         logger.error(f'❌ Getting blob service client error, {E}')
-#         logger.error(f'⚠️ using DefaultAzureCredential')
-#         credential=DefaultAzureCredential()
-#         _blob_service_client = BlobServiceClient(
-#             account_url=AzureBlobConstants.ENDPOINT,
-#             credential=credential,
-#         )
-#         return _blob_service_client
-
-
 async def get_container_client() -> ContainerClient:
     try:
         container_client = ContainerClient(
@@ -39,6 +17,7 @@ async def get_container_client() -> ContainerClient:
             container_name=AzureBlobConstants.CONTAINER_NAME,
             credential=AzureSasCredential(AzureBlobConstants.SAS_KEY)
         )
+        logger.info(f'✅ Getting container client success')
         return container_client
     except Exception as E:
         logger.error(f'❌ Getting container client error, {E}')
@@ -50,19 +29,6 @@ async def get_container_client() -> ContainerClient:
         )
         return container_client
 
-# async def get_container_client() -> ContainerClient:
-#     service = await get_blob_service_client()
-#     container = service.get_container_client(
-#         AzureBlobConstants.CONTAINER_NAME
-#     )
-
-#     try:
-#         await container.create_container()
-#         logger.info("✅ Blob container created")
-#     except Exception:
-#         pass  # already exists
-
-#     return container
 
 
 # ---------- HELPERS ----------
@@ -71,49 +37,7 @@ def build_blob_path(filename: str) -> str:
 
 
 # ---------- UPLOAD ----------
-# async def upload_pdf(
-#     file: UploadFile,
-# ) -> Dict[str, Any]:
-
-#     if file.content_type != "application/pdf":
-#         raise HTTPException(
-#             status_code=status.HTTP_400_BAD_REQUEST,
-#             detail="Only PDF files are allowed",
-#         )
-
-#     container = await get_container_client()
-
-#     blob_name = f"{uuid.uuid4()}.pdf"
-#     blob_path = build_blob_path(blob_name)
-
-#     blob_client = container.get_blob_client(blob_path)
-
-#     size = 0
-#     chunks = []
-
-#     async for chunk in file.stream():
-#         size += len(chunk)
-#         if size > AzureBlobConstants.MAX_FILE_SIZE_MB * 1024 * 1024:
-#             raise HTTPException(
-#                 status_code=status.HTTP_413_REQUEST_ENTITY_TOO_LARGE,
-#                 detail="File too large",
-#             )
-#         chunks.append(chunk)
-
-#     await blob_client.upload_blob(
-#         b"".join(chunks),
-#         overwrite=False,
-#         content_type="application/pdf",
-#     )
-
-#     logger.info(f"✅ Uploaded PDF: {blob_path}")
-
-#     return {
-#         "blob_name": blob_name,
-#         "path": blob_path,
-#         "url": blob_client.url,
-#         "size_bytes": size,
-#     }
+
 def helper_path_maker(
     tenant_id: str,
     user_id: str,
@@ -123,8 +47,6 @@ def helper_path_maker(
     return f"{in_folder}/{tenant_id}/{user_id}/{filename}"
 
 
-
-
 async def upload_pdf(file: UploadFile, tenant_id: str, user_id: str) -> dict:
     container_client = await get_container_client()
     blob_name = helper_path_maker(tenant_id, user_id, file.filename)
@@ -139,6 +61,7 @@ async def upload_pdf(file: UploadFile, tenant_id: str, user_id: str) -> dict:
         content_settings=ContentSettings(
             content_type=file.content_type or "application/pdf"
         ),
+        max_concurrency=4,
     )
 
     return {

services/agentic/profile_scoring.py

@@ -408,8 +408,8 @@ class AgenticScoringService:
             llm=llm,
             input_llm=input_llm,
             metadata_observability=MetadataObservability(
-                fullname="system",
-                task_id="system",
+                fullname=self.user.full_name,
+                task_id=self.user.user_id,
                 agent=self._ai_matching.__name__,
             )
         )

services/base/BaseGenerator.py

@@ -64,73 +64,23 @@ class BaseAIGenerator:
         """private helper for retries"""
         return chain.invoke(input_llm, config=config)
     
-    # @trace_runtime
-    # async def agenerate(self):
-    #     try:
-    #         chain = self.prompt | self.llm
-    #         config = {"callbacks": [langfuse_handler]}
-
-    #         with langfuse.start_as_current_observation(
-    #             as_type='generation',
-    #             name=self.name,
-    #             input=self.input_llm,
-    #         ) as trace:
-    #             trace.update_trace(user_id=self.metadata_observability.fullname, 
-    #                             session_id=self.metadata_observability.task_id, 
-    #                             metadata=self.metadata_observability.model_dump()
-    #                             )
-    #             output = await self._asafe_invoke(chain=chain,
-    #                                                 input_llm=self.input_llm, 
-    #                                                 config=config)
-    #             trace.update_trace(output=output)
-    #             return output
-    #     except Exception as E:
-    #         logger.error(f"❌ BaseGenerator, agenerate error, {E}")
-    #         return None
-    
-    # @trace_runtime
-    # async def generate(self):
-    #     try:
-    #         chain = self.prompt | self.llm
-    #         config = {"callbacks": [langfuse_handler]}
-
-    #         with langfuse.start_as_current_observation(
-    #             as_type='generation',
-    #             name=self.name,
-    #             input=self.input_llm,
-    #         ) as trace:
-    #             trace.update_trace(user_id=self.metadata_observability.fullname, 
-    #                             session_id=self.metadata_observability.task_id, 
-    #                             metadata=self.metadata_observability.model_dump()
-    #                             )
-    #             output = self._safe_invoke(chain=chain,
-    #                                         input_llm=self.input_llm, 
-    #                                         config=config)
-    #             trace.update_trace(output=output)
-    #             return output
-    #     except Exception as E:
-    #         logger.error(f"❌ BaseGenerator, generate error, {E}")
-    #         return None
-
     @trace_runtime
     async def agenerate(self):
         trace = None
         try:
-            chain = self.prompt | self.llm
-            config = {"callbacks": [langfuse_handler]}
-
             # ✅ Create trace (no context manager, no end())
             trace = langfuse.trace(
                 name=self.name,
                 input=self.input_llm,
-            )
-
-            trace.update(
                 user_id=self.metadata_observability.fullname,
                 session_id=self.metadata_observability.task_id,
                 metadata=self.metadata_observability.model_dump(),
             )
 
+            handler = trace.get_langchain_handler() 
+            config = {"callbacks": [handler]}
+            chain = self.prompt | self.llm
+
             output = await self._asafe_invoke(
                 chain=chain,
                 input_llm=self.input_llm,
@@ -156,20 +106,19 @@ class BaseAIGenerator:
     async def generate(self):
         trace = None
         try:
-            chain = self.prompt | self.llm
-            config = {"callbacks": [langfuse_handler]}
 
             trace = langfuse.trace(
                 name=self.name,
                 input=self.input_llm,
-            )
-
-            trace.update(
                 user_id=self.metadata_observability.fullname,
                 session_id=self.metadata_observability.task_id,
                 metadata=self.metadata_observability.model_dump(),
             )
 
+            handler = trace.get_langchain_handler() 
+            config = {"callbacks": [handler]}
+            chain = self.prompt | self.llm
+
             output = self._safe_invoke(
                 chain=chain,
                 input_llm=self.input_llm,

services/knowledge/extract_profile.py

@@ -149,8 +149,8 @@ class KnowledgeExtractService:
                 input_llm=input_llm,
                 llm=llm,
                 metadata_observability=MetadataObservability(
-                    fullname="local test",
-                    task_id=raw_profile.get("profile_id"),
+                    fullname=self.user.full_name,
+                    task_id=self.user.user_id,
                     agent=self.extract.__name__,
                 )
             )