Update app.py

app.py

@@ -9,35 +9,89 @@ import pickle
 import time
 setattr(httpcore, 'SyncHTTPTransport', 'AsyncHTTPProxy')
 
+"""
+For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
+"""
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 model = SentenceTransformer('intfloat/multilingual-e5-large-instruct')
 
 def get_detailed_instruct(task_description: str, query: str) -> str:
-        return f'Instruct: {task_description}\nQuery: {query}'
+    return f'Instruct: {task_description}\nQuery: {query}'
 
-def respond(message,
-    max_tokens = 2048,
-    temperature = 0.7,
-    top_p = 0.95,
-    ):
-    #system role
-    messages = [{"role": "system", "content": "You are a sunni moslem bot that always give answer based on quran, hadith, and the companions of prophet Muhammad!"}]
+
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+):
+    messages = [{"role": "system", "content": system_message}]
 
     #make a moslem bot
     messages.append({"role": "user", "content": "I want you to answer strictly based on quran and hadith"})
     messages.append({"role": "assistant", "content": "I'd be happy to help! Please go ahead and provide the sentence you'd like me to analyze. Please specify whether you're referencing a particular verse or hadith (Prophetic tradition) from the Quran or Hadith, or if you're asking me to analyze a general statement."})
+
+    #adding fatwa references
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    selected_references = torch.load('selected_references.sav', map_location=torch.device(device))
+    encoded_questions = torch.load('encoded_questions.sav', map_location=torch.device(device))
+    
+    task = 'Given a web search query, retrieve relevant passages that answer the query'
+    queries = [
+        get_detailed_instruct(task, message)
+    ]
+    print("start\n")
+    print(time.time())
+
+    query_embeddings = model.encode(queries, convert_to_tensor=True, normalize_embeddings=True)
+    scores = (query_embeddings @ encoded_questions.T) * 100
+    selected_references['similarity'] = scores.tolist()[0]
+    sorted_references = selected_references.sort_values(by='similarity', ascending=False)
+    sorted_references = sorted_references.iloc[:1]
+    sorted_references = sorted_references.sort_values(by='similarity', ascending=True)
+    print(sorted_references.shape[0])
+    print(sorted_references['similarity'].tolist())
+    print("sorted references\n")
+    print(time.time())
+
+    from googletrans import Translator
+    translator = Translator()
+    
+    for index, row in sorted_references.iterrows():
+        if(type(row["user"]) is str and type(row['assistant']) is str):
+            try:
+                translator = Translator()
+                print(index)
+                print(f'{row["user"]}')
+                translated = translator.translate(f'{row["user"]}', src='ar', dest='en')
+                print(translated)
+                user = translated.text
+                print(user)
+                assistant = translator.translate(row['assistant']).text
+                messages.append({"role": "user", "content":user })
+                messages.append({"role": "assistant", "content": assistant})
+            except Exception as error:
+                print("1. An error occurred:", error)
+                print("adding fatwa references exception occurred")
+    
+    print("append references\n")
+    print(time.time())
     
     #adding more references
     df = pd.read_csv("moslem-bot-reference.csv", sep='|')
     for index, row in df.iterrows():
         messages.append({"role": "user", "content": row['user']})
         messages.append({"role": "assistant", "content": row['assistant']})
-    
-    print("added more references\n")
-    print(time.time())
+        
+    for val in history:
+        if val[0]:
+            messages.append({"role": "user", "content": val[0]})
+        if val[1]:
+            messages.append({"role": "assistant", "content": val[1]})
 
     #latest user question
-    from googletrans import Translator
     translator = Translator()
     en_message = ""
     message_language = "en"
@@ -56,13 +110,9 @@ def respond(message,
         messages.append({"role": "user", "content": message})
         print("An error occurred:", error)
         print("en_message exception occurred")
-    
-    
-    print(messages)
-    print("added last question\n")
-    print(time.time())
 
     response = ""
+
     for message in client.chat_completion(
         messages,
         max_tokens=max_tokens,
@@ -73,46 +123,22 @@ def respond(message,
         token = message.choices[0].delta.content
 
         response += token
-        yield response
-    """
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        try:
-            print("cek1")
-            if(message):
-                print("cek2")
-                if len(message.choices)>0:
-                    print("cek3")
-                    token = message.choices[0].delta.content
-                    response += token
-                    if(len(response)>0):
-                        print("cek4")
-                        translated = translator.translate(response, src='en', dest=message_language)
-                        if not (translated is None):
-                            print("cek5")
-                            translated_response = translated.text
-                            yield translated_response
-                        else:
-                            yield response
-                    else:
-                        yield response
-                else:
-                    yield response
-            else:
-                yield response
-        except Exception as error:
-            print("An error occurred:", error)
+
+        translated_response = translator.translate(response, src='en', dest=message_language).text
+        if(len(translated_response)>0):
+            yield translated_response
+        else:
             yield response
-    """
-demo = gr.Interface(
-    fn=respond,
+
+
+"""
+For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
+"""
+demo = gr.ChatInterface(
+    respond,
     additional_inputs=[
-        gr.Slider(minimum=1, maximum=2048, value=2048, step=1, label="Max new tokens"),
+        gr.Textbox(value="You are a sunni moslem bot that always give answer based on quran, hadith, and the companions of prophet Muhammad!", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
             minimum=0.1,
@@ -121,16 +147,9 @@ demo = gr.Interface(
             step=0.05,
             label="Top-p (nucleus sampling)",
         ),
-    ],  
-    inputs="textbox", 
-    outputs="textbox",  
-    cache_examples="lazy",
-    examples=[
-                ["Why is men created?"],
-                ["Please tell me about superstition!"],
-                ["How moses defeat pharaoh?"],
-            ],
-    title="Moslem Bot")
-    
+    ],
+)
+
+
 if __name__ == "__main__":
-    demo.launch()   
+    demo.launch()