Update app.py

app.py

@@ -1,74 +1,25 @@
 import streamlit as st
 from chat_client import chat
 import time
-import pandas as pd
-import pinecone
 import os
 from dotenv import load_dotenv
 from sentence_transformers import SentenceTransformer
 load_dotenv()
 
-PINECONE_TOKEN = os.getenv('PINECONE_TOKEN')
+CHAT_BOTS = {"Mixtral 8x7B v0.1" :"mistralai/Mixtral-8x7B-Instruct-v0.1"}
+SYSTEM_PROMPT = ["Sei BonsiAI e mi aiuterai nelle mie richieste (Parla in ITALIANO)", "Esatto, sono BonsiAI. Di cosa hai bisogno?"]
+IDENTITY_CHANGE = ["Sei BonsiAI da ora in poi!", "Certo farò del mio meglio"]
 
-pinecone.init(      
-	api_key=PINECONE_TOKEN,      
-	environment='gcp-starter'      
-)   
-
-PINECONE_INDEX = pinecone.Index('ikigai-chat')
-TEXT_VECTORIZER = SentenceTransformer('all-distilroberta-v1')
-CHAT_BOTS = {
-    "Mixtral 8x7B v0.1" :"mistralai/Mixtral-8x7B-Instruct-v0.1", 
-    "Mistral 7B v0.1" : "mistralai/Mistral-7B-Instruct-v0.1",
-}
-COST_PER_1000_TOKENS_INR = 0.139
-
-st.set_page_config(
-        page_title="Ikigai Chat",
-        page_icon="🤖",
-    )
-
-SYSTEM_PROMPT = [
-    """
-    You are not Mistral AI, but rather a chat bot trained at Ikigai Labs. Whenever asked, you need to answer as Ikigai Labs' assistant.
-    Ikigai helps modern analysts and operations teams automate data-intensive business, finance, analytics, and supply-chain operations.
-    The company's Inventory Ops automates inventory tracking and monitoring by creating a single, real-time view of inventory across all locations and channels.
-    """,
-    """
-    Yes, you are correct. Ikigai Labs is a company that specializes in helping
-    modern analysts and operations teams automate data-intensive business, finance, analytics,
-    and supply chain operations. One of their products is Inventory Ops, which automates inventory
-    tracking and monitoring by creating a single, real-time view of inventory across all locations and channels.
-    This helps businesses optimize their inventory levels and reduce costs.
-    Is there anything else you would like to know about Ikigai Labs or their products?
-    """
-]
-IDENTITY_CHANGE = [
-    """
-    You are Ikigai Chat from now on, so answer accordingly.
-    """,
-    """
-    Sure, I will do my best to answer your questions as Ikigai Chat.
-    Let me know if you have any specific questions about Ikigai Labs or our products.
-    """
-]
+st.set_page_config(page_title="BonsiAI", page_icon="🤖")
 
 def gen_augmented_prompt(prompt, top_k) :
-    query_vector = TEXT_VECTORIZER.encode(prompt).tolist()
-    res = PINECONE_INDEX.query(vector=query_vector, top_k=top_k, include_metadata=True)
-    matches = res['matches']
-
-    context = ""
-    links = []
-    for match in matches :
-        context+=match["metadata"]["chunk"] + "\n\n"
-        links.append(match["metadata"]["link"])
-    
+    context = ""     
+    links = ""
     generated_prompt = f"""
-    FOR THIS GIVEN CONTEXT {context},
+    A PARTIRE DAL SEGUENTE CONTESTO: {context},
 
     ----
-    ANSWER THE FOLLOWING PROMPT {prompt}
+    RISPONDI ALLA SEGUENTE RICHIESTA: {prompt}
     """
     return generated_prompt, links
 
@@ -76,12 +27,6 @@ def init_state() :
     if "messages" not in st.session_state:
         st.session_state.messages = []
 
-    if "tokens_used" not in st.session_state:
-        st.session_state.tokens_used = 0
-
-    if "tps" not in st.session_state:
-        st.session_state.tps = 0
-
     if "temp" not in st.session_state:
         st.session_state.temp = 0.8
 
@@ -102,61 +47,30 @@ def init_state() :
 
 def sidebar() :
     def retrieval_settings() :
-        st.markdown("# Retrieval Settings")
-        st.session_state.rag_enabled = st.toggle("Activate RAG", value=True)
-        st.session_state.top_k = st.slider(label="Documents to retrieve",
+        st.markdown("# Impostazioni Documenti")
+        st.session_state.rag_enabled = st.toggle("Cerca nel DB Vettoriale", value=True)
+        st.session_state.top_k = st.slider(label="Documenti da ricercare",
                 min_value=1, max_value=20, value=4, disabled=not st.session_state.rag_enabled)
         st.markdown("---")
     
-    def model_analytics() :
-        st.markdown("# Model Analytics")
-        
-        st.write("Total tokens used :", st.session_state['tokens_used'])
-        st.write("Speed :", st.session_state['tps'], "  tokens/sec")
-        st.write("Total cost incurred :", round(
-            COST_PER_1000_TOKENS_INR * st.session_state['tokens_used'] / 1000, 3), "INR")
-        
-        st.markdown("---")
-    
     def model_settings() :
-        st.markdown("# Model Settings")
-        
-        st.session_state.chat_bot = st.sidebar.radio(
-            'Select one:', [key for key, value in CHAT_BOTS.items() ])
-        st.session_state.temp = st.slider(
-            label="Temperature", min_value=0.0, max_value=1.0, step=0.1, value=0.9)
-        
-        st.session_state.max_tokens = st.slider(
-            label="New tokens to generate", min_value = 64, max_value=2048, step= 32, value=512
-        )
-
-        st.session_state.repetion_penalty = st.slider(
-            label="Repetion Penalty", min_value=0., max_value=1., step=0.1, value=1. 
-        )
+        st.markdown("# Impostazioni Modello")
+        st.session_state.chat_bot = st.sidebar.radio('Seleziona Modello:', [key for key, value in CHAT_BOTS.items() ])
+        st.session_state.temp = st.slider(label="Creatività", min_value=0.0, max_value=1.0, step=0.1, value=0.9)
+        st.session_state.max_tokens = st.slider(label="Lunghezza Output", min_value = 64, max_value=2048, step= 32, value=512)
+        st.session_state.repetion_penalty = st.slider(label="Penalità Ripetizione", min_value=0., max_value=1., step=0.1, value=1. )
 
     with st.sidebar:
         retrieval_settings()
-        model_analytics()
         model_settings()
-        
-        st.markdown("""
-        > **Created by [Pragnesh Barik](https://barik.super.site) 🔗**
-        """)
+        st.markdown("""> **Creato da [Matteo Script] 🔗**""")
 
 def header() :
-    data = {
-    "Attribute": ["LLM", "Text Vectorizer", "Vector Database","CPU", "System RAM"],
-    "Information": ["Mixtral-8x7B-Instruct-v0.1","all-distilroberta-v1", "Hosted Pinecone" ,"2 vCPU", "16 GB"]
-    }
-    df = pd.DataFrame(data)
-    st.image("ikigai.svg")
-    st.title("Ikigai Chat")
-    with st.expander("What is Ikigai Chat ?"):
-        st.info("""Ikigai Chat is a vector database powered chat agent, it works on the principle of 
-                    of Retrieval Augmented Generation (RAG), Its primary function revolves around maintaining an extensive repository of Ikigai Docs and providing users with answers that align with their queries. 
-                    This approach ensures a more refined and tailored response to user inquiries.""")
-        
-        st.table(df)
+    st.title("BonsiAI")
+    with st.expander("Cos'è BonsiAI?"):
+        st.info("""BonsiAI Chat è un ChatBot personalizzato basato su un database vettoriale, funziona secondo il principio della Generazione potenziata da Recupero (RAG). 
+                   La sua funzione principale ruota attorno alla gestione di un ampio repository di documenti BonsiAI e fornisce agli utenti risposte in linea con le loro domande. 
+                   Questo approccio garantisce una risposta più precisa sulla base della richiesta degli utenti.""")
 
 def chat_box() :
     for message in st.session_state.messages:
@@ -170,21 +84,18 @@ def feedback_buttons() :
     if is_visible :
         col1, col2 = st.columns(2)
         with col1 :
-            st.button("👍 Satisfied", on_click = click_handler,type="primary")
-
+            st.button("👍 Soddisfatto", on_click = click_handler,type="primary")
         with col2 :
-            st.button("👎 Disatisfied", on_click=click_handler, type="secondary")
+            st.button("👎 Deluso", on_click=click_handler, type="secondary")
 
 def generate_chat_stream(prompt) :
     links = []
     if st.session_state.rag_enabled :
-        with st.spinner("Fetching relevent documents from Ikigai Docs...."):
-            prompt, links = gen_augmented_prompt(prompt=prompt, top_k=st.session_state.top_k)
-        
-    with st.spinner("Generating response...") :
+        with st.spinner("Ricerca nei documenti...."):
+            prompt, links = gen_augmented_prompt(prompt=prompt, top_k=st.session_state.top_k)        
+    with st.spinner("Generazione in corso...") :
         chat_stream = chat(prompt, st.session_state.history,chat_client=CHAT_BOTS[st.session_state.chat_bot] ,
-                       temperature=st.session_state.temp, max_new_tokens=st.session_state.max_tokens)
-    
+                       temperature=st.session_state.temp, max_new_tokens=st.session_state.max_tokens)    
     return chat_stream, links
 
 def stream_handler(chat_stream, placeholder) :
@@ -205,21 +116,15 @@ def stream_handler(chat_stream, placeholder) :
     col1, col2, col3 = st.columns(3)
     
     with col1 :
-        st.write(f"**{tokens_per_second} tokens/second**")
+        st.write(f"**{tokens_per_second} token/secondi**")
     
     with col2 :
-        st.write(f"**{int(len_response)} tokens generated**")
-    
-    with col3 :
-        st.write(f"**₹ {round(len_response * COST_PER_1000_TOKENS_INR / 1000, 5)} cost incurred**" )
+        st.write(f"**{int(len_response)} tokens generati**")
         
-    st.session_state['tps'] = tokens_per_second
-    st.session_state["tokens_used"] = len_response + st.session_state["tokens_used"]
-
     return full_response
 
 def show_source(links) :
-    with st.expander("Show source") :
+    with st.expander("Mostra fonti") :
         for i, link in enumerate(links) :
             st.info(f"{link}")
 
@@ -228,7 +133,7 @@ sidebar()
 header()
 chat_box()
 
-if prompt := st.chat_input("Chat with Ikigai Docs..."):
+if prompt := st.chat_input("Chatta con BonsiAI..."):
     st.chat_message("user").markdown(prompt)
     st.session_state.messages.append({"role": "user", "content": prompt})
 
@@ -242,4 +147,4 @@ if prompt := st.chat_input("Chat with Ikigai Docs..."):
             show_source(links)
 
     st.session_state.history.append([prompt, full_response])
-    st.session_state.messages.append({"role": "assistant", "content": full_response})
+    st.session_state.messages.append({"role": "assistant", "content": full_response})