Update app.py

app.py

@@ -3,10 +3,10 @@ import warnings
 import gradio as gr
 from dotenv import load_dotenv
 
-from langchain.chains import ConversationalRetrievalChain
 from langchain_community.vectorstores import FAISS
 from langchain_community.embeddings import AzureOpenAIEmbeddings
-from langchain_community.chat_models import AzureChatOpenAI
+from langchain.embeddings.openai import OpenAIEmbeddings
+from openai import AzureOpenAI
 
 # Patch Gradio bug (schema parsing issue)
 import gradio_client.utils
@@ -17,49 +17,56 @@ load_dotenv()
 AZURE_OPENAI_API_KEY = os.getenv("AZURE_OPENAI_API_KEY")
 AZURE_OPENAI_ENDPOINT = os.getenv("AZURE_OPENAI_ENDPOINT")
 AZURE_OPENAI_LLM_DEPLOYMENT = os.getenv("AZURE_OPENAI_LLM_DEPLOYMENT")
+embeddings = OpenAIEmbeddings()
 
-if not all([AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_LLM_DEPLOYMENT]):
+if not all([AZURE_OPENAI_API_KEY, AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_LLM_DEPLOYMENT, OPENAI_EMBEDDING]):
     raise ValueError("Azure OpenAI environment variables are missing.")
 
 # Suppress warnings
 warnings.filterwarnings("ignore")
 
-# Initialize embedding model (must be deployed separately)
-embeddings = AzureOpenAIEmbeddings(
-    azure_deployment="text-embedding-ada-002",  # make sure this deployment exists
-    azure_endpoint=AZURE_OPENAI_ENDPOINT,
-    openai_api_key=AZURE_OPENAI_API_KEY,
-    openai_api_version="2024-08-01-preview",
-    chunk_size=1000
-)
+# Initialize embedding model
+# embeddings = AzureOpenAIEmbeddings(
+#     azure_deployment=OPENAI_EMBEDDING,
+#     azure_endpoint=AZURE_OPENAI_ENDPOINT,
+#     openai_api_key=AZURE_OPENAI_API_KEY,
+#     openai_api_version="2024-08-01-preview",
+#     chunk_size=1000
+# )
 
 # Load FAISS vector store
 vectorstore = FAISS.load_local(
     "faiss_index_sysml", embeddings, allow_dangerous_deserialization=True
 )
 
-# Initialize LLM
-llm = AzureChatOpenAI(
-    deployment_name=AZURE_OPENAI_LLM_DEPLOYMENT,
+# Initialize Azure OpenAI client directly
+client = AzureOpenAI(
+    api_key=AZURE_OPENAI_API_KEY,
     azure_endpoint=AZURE_OPENAI_ENDPOINT,
-    openai_api_key=AZURE_OPENAI_API_KEY,
-    openai_api_version="2024-08-01-preview",
-    temperature=0.5
-)
-
-# Build conversational chain with history
-qa = ConversationalRetrievalChain.from_llm(
-    llm=llm,
-    retriever=vectorstore.as_retriever(),
-    return_source_documents=False
+    api_version="2024-08-01-preview"
 )
 
 history = []
 
-# Chatbot logic
+# Chatbot logic using AzureOpenAI directly
 def sysml_chatbot(message, history):
-    result = qa({"question": message, "chat_history": history})
-    answer = result["answer"]
+    # Perform retrieval
+    retriever = vectorstore.as_retriever()
+    docs = retriever.get_relevant_documents(message)
+    context = "\n\n".join(doc.page_content for doc in docs[:4])
+
+    # Compose prompt with retrieved context
+    system_prompt = "You are a helpful assistant knowledgeable in SysML. Use the context below to answer the user's question.\n\nContext:\n" + context
+
+    response = client.chat.completions.create(
+        model=AZURE_OPENAI_LLM_DEPLOYMENT,
+        messages=[
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": message}
+        ]
+    )
+
+    answer = response.choices[0].message.content
     history.append((message, answer))
     return "", history