Update app.py

app.py

@@ -25,6 +25,7 @@ from fastapi import FastAPI, File, UploadFile
 import time
 import requests
 from datetime import datetime
+from io import BytesIO
 import os
 os.environ["GROQ_API_KEY"] = "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"
 os.environ["TAVILY_API_KEY"] = "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"
@@ -77,33 +78,53 @@ tavily_search = TavilySearchResults(max_results=4)
 
 @app.post("/uploadpdfs")
 async def upload_file(file: UploadFile = File(...)):
-    # global vector_index
+    # Read file contents into memory
+    contents = await file.read()
+    file_stream = BytesIO(contents)
 
-    # Save uploaded file to a temp directory
-    temp_dir = "temp_uploads"
-    os.makedirs(temp_dir, exist_ok=True)
-    file_id = str(uuid.uuid4())
-    file_path = os.path.join(temp_dir, f"{file_id}_{file.filename}")
-
-    with open(file_path, "wb") as f:
-        shutil.copyfileobj(file.file, f)
-
-    # Load and parse document
-    documents = SimpleDirectoryReader(input_files=[file_path]).load_data()
+    # Load and parse document directly from memory
+    documents = SimpleDirectoryReader(input_file_objs=[file_stream], input_file_names=[file.filename]).load_data()
     parser = SentenceSplitter(chunk_size=300, chunk_overlap=50)
     nodes = parser.get_nodes_from_documents(documents)
 
     # Create or update vector index
     embed_model = HuggingFaceEmbedding(model_name="WhereIsAI/UAE-Large-V1")
-    # if vector_index is None:
     if not hasattr(app.state, "vector_index") or app.state.vector_index is None:
         app.state.vector_index = VectorStoreIndex(nodes, embed_model=embed_model)
-        message = "New vector index created and file stored."
+        message = "New vector index created and file processed in-memory."
     else:
         app.state.vector_index.insert_nodes(nodes)
-        message = "File stored and vector index updated."
+        message = "File processed in-memory and vector index updated."
 
     return {"message": message, "filename": file.filename}
+# async def upload_file(file: UploadFile = File(...)):
+#     # global vector_index
+
+#     # Save uploaded file to a temp directory
+#     temp_dir = "temp_uploads"
+#     os.makedirs(temp_dir, exist_ok=True)
+#     file_id = str(uuid.uuid4())
+#     file_path = os.path.join(temp_dir, f"{file_id}_{file.filename}")
+
+#     with open(file_path, "wb") as f:
+#         shutil.copyfileobj(file.file, f)
+
+#     # Load and parse document
+#     documents = SimpleDirectoryReader(input_files=[file_path]).load_data()
+#     parser = SentenceSplitter(chunk_size=300, chunk_overlap=50)
+#     nodes = parser.get_nodes_from_documents(documents)
+
+#     # Create or update vector index
+#     embed_model = HuggingFaceEmbedding(model_name="WhereIsAI/UAE-Large-V1")
+#     # if vector_index is None:
+#     if not hasattr(app.state, "vector_index") or app.state.vector_index is None:
+#         app.state.vector_index = VectorStoreIndex(nodes, embed_model=embed_model)
+#         message = "New vector index created and file stored."
+#     else:
+#         app.state.vector_index.insert_nodes(nodes)
+#         message = "File stored and vector index updated."
+
+#     return {"message": message, "filename": file.filename}
 
 
 class QueryRequest(BaseModel):