Update app.py

app.py

@@ -20,25 +20,22 @@ HF_MODEL1 = 'HuggingFaceH4/zephyr-7b-beta'
 vector_path = 'faiss_index'
 hf_token = os.environ["HUGGINGFACEHUB_API_TOKEN"]
 
-# Initialize your embedding model
-embedding_model = HuggingFaceEmbeddings(model_name=EMB_MODEL1)
-
-# Load FAISS from relative path
-if os.path.exists("faiss_index"):
-    vectordb = FAISS.load_local(vector_path, embedding_model, allow_dangerous_deserialization=True)
-else:
-    raise FileNotFoundError("FAISS index not found in Space. Please upload it to faiss_index/")
-
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature
+def respond(message, history #,
+    #system_message,
+    #max_tokens,
+    #temperature,
     #top_p
     ):
-
+    
+    # Initialize your embedding model
+    embedding_model = HuggingFaceEmbeddings(model_name=EMB_MODEL1)
+    
+    # Load FAISS from relative path
+    if os.path.exists("faiss_index"):
+        vectordb = FAISS.load_local(vector_path, embedding_model, allow_dangerous_deserialization=True)
+    else:
+        raise FileNotFoundError("FAISS index not found in Space. Please upload it to faiss_index/")
+    
     # define retriever object
     retriever = vectordb.as_retriever(search_type="similarity", search_kwargs={"k": 5})
 
@@ -46,32 +43,34 @@ def respond(
     llm = HuggingFaceHub(
         repo_id=MISTRAL_MODEL1, 
         huggingfacehub_api_token=hf_token,
-        model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens}
+        model_kwargs={"temperature": 0.5, "max_new_tokens": 512}
         )
+    
     # create a RAG pipeline
     qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)
     #generate results
     result = qa_chain.invoke(message)
+    responce = result['result']
     
-    yield result['result']
+    yield responce
 
 
 demo = gr.ChatInterface(
     respond,
     type="messages", 
-    autofocus=False,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=128, maximum=1024, value=512, step=128, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    autofocus=False #,
+    #additional_inputs=[
+    #    gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+    #    gr.Slider(minimum=128, maximum=1024, value=512, step=128, label="Max new tokens"),
+    #    gr.Slider(minimum=0.1, maximum=1.0, value=0.7, step=0.1, label="Temperature"),
+    #    gr.Slider(
+    #        minimum=0.1,
+    #        maximum=1.0,
+    #        value=0.95,
+    #        step=0.05,
+    #        label="Top-p (nucleus sampling)",
+    #    ),
+    #],
 )