Update app.py

app.py

@@ -19,6 +19,7 @@ from langchain.embeddings import HuggingFaceInstructEmbeddings, HuggingFaceEmbed
 from langchain.prompts import PromptTemplate
 from langchain.embeddings.openai import OpenAIEmbeddings
 from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.retrievers.self_query.base import SelfQueryRetriever
 from langchain.vectorstores import Chroma
 from chromadb.errors import InvalidDimensionException
 
@@ -198,7 +199,10 @@ def document_retrieval_chroma():
     #ChromaDb um die embedings zu speichern
     db = Chroma(embedding_function = embeddings, persist_directory = PATH_WORK + CHROMA_DIR)
     print ("Chroma DB bereit ...................")
-    return db
+    llm =  OpenAI(temperature=0.5)
+    retriever =  SelfQueryRetriever.from_llm(llm,vectorstore,document_content_description,metadata_field_info,enable_limit=True,verbose=True,)
+
+    return db, retriever
 
 
 
@@ -213,10 +217,10 @@ def llm_chain(prompt):
 
 #langchain nutzen, um prompt an llm zu leiten, aber vorher in der VektorDB suchen, um passende splits zum Prompt hinzuzufügen
 #prompt mit RAG!!!
-def rag_chain(prompt, db):
+def rag_chain(prompt, db, retriever):
     rag_template = "Nutze die folgenden Kontext Teile am Ende, um die Frage zu beantworten . " + template + "Frage: " + prompt + "Kontext Teile: "
-    retriever = db.as_retriever()
-    retrieved_chunks = retriever.search(prompt, k=3)
+    retrieved_chunks = retriever.get_relevant_documents(prompt)
+    print
     neu_prompt = rag_template 
     for i, chunk in enumerate(retrieved_chunks):
         neu_prompt += f"{i+1}. {chunk}\n"
@@ -276,9 +280,9 @@ def generate(text, history, rag_option, model_option,  temperature=0.5, max_new_
             if not splittet:
                 splits = document_loading_splitting()
                 document_storage_chroma(splits)
-            db = document_retrieval_chroma()
+            db, retriever = document_retrieval_chroma()
             #mit RAG:
-            neu_text_mit_chunks = rag_chain(text, db)
+            neu_text_mit_chunks = rag_chain(text, db, retriever)
             #für Chat LLM:
             #prompt = generate_prompt_with_history_openai(neu_text_mit_chunks, history)
             #als reiner prompt: