Update rag_server.py

rag_server.py

@@ -13,21 +13,22 @@ from transformers import AutoModel
 import streamlit as st
 
 # --- Konfiguration ---
-HF_API_TOKEN = os.environ.get("HF_API_TOKEN")  # Lesen Sie den Token aus der Umgebungsvariable
+os.environ["HF_HOME"] = "/app/cache"  # Specify cache path
+HF_API_TOKEN = os.environ.get("HF_API_TOKEN")  # Read token from environment variable
 MODEL_NAME = "dannyk97/mistral-screenplay-model"
 
 # --- Hilfsfunktionen ---
 
 def query_huggingface_inference_endpoints(prompt):
     """
-    Stellt eine Anfrage an die Hugging Face Inference API.
+    Sends a request to the Hugging Face Inference API.
     """
     try:
         client = InferenceClient(token=HF_API_TOKEN)
         result = client.text_generation(prompt, model=MODEL_NAME)
         return result
     except Exception as e:
-        return f"Fehler bei der Anfrage an Hugging Face API: {e}"
+        return f"Error in query_huggingface_inference_endpoints: {e}"
 
 # Function to download PDF from Google Drive
 def download_pdf_from_drive(drive_link):
@@ -50,17 +51,22 @@ def extract_text_from_pdf(pdf_stream):
 # Function to split text into chunks
 def chunk_text(text, chunk_size=500, chunk_overlap=50):
     text_splitter = RecursiveCharacterTextSplitter(
-        chunk_size=chunk_size,
-        chunk_overlap=chunk_overlap,
-        length_function=len
+        chunk_size=chunk_size, chunk_overlap=chunk_overlap
     )
     return text_splitter.split_text(text)
 
 # Function to create embeddings and store in FAISS
 def create_embeddings_and_store(chunks):
-    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
-    vector_db = FAISS.from_texts(chunks, embedding=embeddings)
-    return vector_db
+    try:
+        embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+        vector_db = FAISS.from_texts(chunks, embedding=embeddings)
+        return vector_db
+    except Exception as e:
+        print(f"Error creating embeddings: {e}")
+        print("Using dummy embeddings to proceed (functionality will be limited).")
+        # Fallback to a simpler embedding model (but this might not work well)
+        vector_db = FAISS.from_texts(["fallback text"], HuggingFaceEmbeddings(model_name="all-mpnet-base-v2")) #Ggf mit "" ersetzen, falls die Implementierung nicht passt.
+        return vector_db
 
 # Function to query the vector database and interact with Hugging Face Inference API
 def query_vector_db(query, vector_db):
@@ -71,7 +77,7 @@ def query_vector_db(query, vector_db):
     # Interact with the Text Generation API
     prompt = f"Nutze diesen Kontext um die Frage zu beantworten: {context}\nFrage: {query}"
     try:
-        output = query_huggingface_inference_endpoints(prompt) # Hier wurde das Modell nicht angegeben
+        output = query_huggingface_inference_endpoints(prompt) #Keine Modelangabe mehr
         return output
     except Exception as e:
         return f"FEHLER: {str(e)}"