Update app.py

app.py

@@ -53,6 +53,7 @@ RAG_CHAIN_PROMPT = PromptTemplate(input_variables = ["context", "question"],
                                   template = rag_template)
 
 #Pfad, wo Docs abgelegt werden können - lokal, also hier im HF Space (sonst auf eigenem Rechner)
+PATH_WORK = "."
 CHROMA_DIR  = "/data/chroma"
 YOUTUBE_DIR = "/data/youtube"
 
@@ -62,7 +63,7 @@ YOUTUBE_URL_1 = "https://www.youtube.com/watch?v=--khbXchTeE"
 YOUTUBE_URL_2 = "https://www.youtube.com/watch?v=hdhZwyf24mE"
 YOUTUBE_URL_3 = "https://www.youtube.com/watch?v=vw-KWfKwvTQ"
 
-MODEL_NAME  = "gpt-4"
+MODEL_NAME  = "gpt-3.5-turbo-16k"
 
 def document_loading_splitting():
     # Document loading
@@ -96,9 +97,14 @@ def document_storage_mongodb(splits):
                                             collection = MONGODB_COLLECTION,
                                             index_name = MONGODB_INDEX_NAME)
 
-def document_retrieval_chroma(llm, prompt):        
-    db = Chroma(embedding_function = OpenAIEmbeddings(),
-                persist_directory = CHROMA_DIR)
+def document_retrieval_chroma(llm, prompt):      
+    embeddings = OpenAIEmbeddings()
+    #Alternative Embedding - für Vektorstore, um Ähnlichkeitsvektoren zu erzeugen
+    #embeddings = HuggingFaceInstructEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2", model_kwargs={"device": "cpu"})
+    db = Chroma(embedding_function = embeddings,
+                #persist_directory = CHROMA_DIR)
+                persist_directory = path_work + '/chroma',
+    
     return db
 
 def document_retrieval_mongodb(llm, prompt):