Update app.py

app.py

@@ -67,86 +67,61 @@ index.add(text_embeddings)
 #index = faiss.read_index("./resourse/embeddings_ngap.faiss")
 
 print("Finish the model init process")
+def format_chat_prompt(message, chat_history):
+    prompt = ""
+    for turn in chat_history:
+        user_message, bot_message = turn
+        prompt = f"{prompt}\nUser: {user_message}\nAssistant: {bot_message}"
+    prompt = f"{prompt}\nUser: {message}\nAssistant:"
+    return prompt
 
-model.config["promptTemplate"] = "[INST] {0} [/INST]"
-model.config["systemPrompt"] = "Tu es un assitant et tu dois répondre en français"
-model._is_chat_session_activated = False
+def respond(message, chat_history):
 
-max_new_tokens = 2048
 
-def generater(message, history):
-    prompt = "<s>"
-    for user_message, assistant_message in history:
-        prompt += model.config["promptTemplate"].format(user_message)
-        prompt += assistant_message +  "</s>"
-    prompt += model.config["promptTemplate"].format(message)
 
-    question_embeddings = np.array([get_text_embedding(message)])
+        prompt = message
 
-    #D, I = index.search(question_embeddings, k=3) # distance, index
-    #retrieved_chunk = [chunks[i] for i in I.tolist()[0]]
+        context.append({'role':'user', 'content':f"{prompt}"})
 
-    #prompt += "Informations de contexte ci-dessous:"
-    #prompt += [i for i in retrieved_chunk]
-    
-    outputs = []    
-    for token in model.generate(prompt=prompt, temp=0.5, top_k = 40, top_p = 1, max_tokens = max_new_tokens, streaming=True):
-        outputs.append(token)
-        yield "".join(outputs)
-
-def vote(data: gr.LikeData):
-    if data.liked:
-        return
-    else:
-        return
-
-chatbot = gr.Chatbot(avatar_images=('./resourse/user-icon.png', './resourse/chatbot-icon.png'),bubble_full_width = False)
-"""
-additional_inputs=[
-    gr.Slider(
-        label="temperature",
-        value=0.5,
-        minimum=0.0,
-        maximum=2.0,
-        step=0.05,
-        interactive=True,
-        info="Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.",
-    ),
-    gr.Slider(
-        label="top_p",
-        value=1.0,
-        minimum=0.0,
-        maximum=1.0,
-        step=0.01,
-        interactive=True,
-        info="0.1 means only the tokens comprising the top 10% probability mass are considered. Suggest set to 1 and use temperature. 1 means 100% and will disable it",
-    ),
-    gr.Slider(
-        label="top_k",
-        value=40,
-        minimum=0,
-        maximum=1000,
-        step=1,
-        interactive=True,
-        info="limits candidate tokens to a fixed number after sorting by probability. Setting it higher than the vocabulary size deactivates this limit.",
-    )
-]
-"""
-additional_inputs=[
-gr.UploadButton(file_types=[".pdf",".csv",".doc"])
-]
-iface = gr.ChatInterface(
-    fn = generater,
-    title=title,
-    description = description,
-    chatbot=chatbot,
-    additional_inputs=additional_inputs,
+        tokenized_chat = tokenizer.apply_chat_template(context, tokenize=True, add_generation_prompt=True, return_tensors="pt")
+
+        outputs = model.generate(tokenized_chat, max_new_tokens=1000, temperature = 0.0)
+
+        bot_message = tokenizer.decode(outputs[0]).split("<|assistant|>")[-1].replace("</s>","")
+
+        context.append({'role':'assistant', 'content':f"{bot_message}"})
+
+        chat_history.append((message, bot_message))
+        return "", chat_history
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Assistant virtuel Ameli")
+    gr.Markdown("Mes réponses sont générées par IA. Elles peuvent être fausses ou imprécises.")
+    with gr.Row():
+        with gr.Column(scale=1):
+          #msg = gr.Audio(sources=["microphone"])
+
+          audio_file = gr.Audio(sources=["microphone"])
+          b1 = gr.Button("Posez votre question à l'oral")
+
+          text = gr.Textbox(lines =5)
+
+
+
+          b1.click(speech_to_text, inputs=audio_file, outputs=text)
+
+
+          #msg = gr.Textbox(label="Posez votre question")
+          btn = gr.Button("Soumettre la question")
+
+
+        with gr.Column(scale=2, min_width=50):
+          chatbot = gr.Chatbot(height=700) #just to fit the notebook
+          clear = gr.ClearButton(components=[text, chatbot], value="Clear console")
 
-)
 
-with gr.Blocks(css="./resourse/style/custom.css") as demo:
-    chatbot.like(vote, None, None)
-    iface.render()
+    btn.click(respond, inputs=[text, chatbot], outputs=[text, chatbot])
+    text.submit(respond, inputs=[text, chatbot], outputs=[text, chatbot]) #Press enter to submit
 
 if __name__ == "__main__":
     demo.queue(max_size=3).launch()