deploy app

.env_example

@@ -1,2 +1,3 @@
 REDIS_KEY = ''
-OPENAI_API_KEY = ''
+OPENAI_API_KEY = ''
+HUGGINGFACEHUB_API_TOKEN = ''

main.py → app.py

@@ -1,29 +1,106 @@
 import streamlit as st
-from chatbot import llm_chain, chain
 from sentence_transformers import SentenceTransformer
 from redis.commands.search.query import Query
-from database import redis_conn
+import redis
+from langchain.prompts import PromptTemplate
+from langchain import HuggingFaceHub
+from langchain.chains import LLMChain
+from langchain.memory import ConversationBufferMemory
+from langchain.chat_models import ChatOpenAI
+from langchain.callbacks.base import BaseCallbackHandler
+import os
+from dotenv import load_dotenv
 import numpy as np
 
+load_dotenv()
+redis_key = os.getenv('REDIS_KEY')
+HUGGINGFACEHUB_API_TOKEN = os.getenv('HUGGINGFACEHUB_API_TOKEN')
+repo_id = 'tiiuae/falcon-7b-instruct'
 
+class StreamHandler(BaseCallbackHandler):
+    def __init__(self, container, initial_text="", display_method='markdown'):
+        self.container = container
+        self.text = initial_text
+        self.display_method = display_method
+
+    def on_llm_new_token(self, token: str, **kwargs) -> None:
+        self.text += token + " "
+        display_function = getattr(self.container, self.display_method, None)
+        if display_function is not None:
+            display_function(self.text)
+        else:
+            raise ValueError(f"Invalid display_method: {self.display_method}")
+        
 
 st.title('My Amazon shopping buddy 🏷️')
 st.caption('🤖 Powered by Falcon Open Source AI model')
-st.session_state['disabled']= False
+
+#connect to redis database
+@st.cache_resource()
+def redis_connect():
+  redis_conn = redis.Redis(
+    host='redis-12882.c259.us-central1-2.gce.cloud.redislabs.com',
+    port=12882,
+    password=redis_key)
+  return redis_conn
+
+redis_conn = redis_connect()
+
+#the encoding keywords chain
+@st.cache_resource()
+def encode_keywords_chain():
+    falcon_llm_1 = HuggingFaceHub(repo_id = repo_id, model_kwargs={'temperature':0.1,'max_new_tokens':500},huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN)
+    prompt = PromptTemplate(
+        input_variables=["product_description"],
+        template="Create comma seperated product keywords to perform a query on a amazon dataset for this user input: {product_description}",
+    )
+    chain = LLMChain(llm=falcon_llm_1, prompt=prompt)
+    return chain
+chain = encode_keywords_chain()
+#the present products chain
+
+@st.cache_resource()
+def present_products_chain():
+    template = """You are a salesman. Be kind, detailed and nice.  take the given context and Present the given queried search result in a nice way as answer to the user_msg. dont ask questions back or freestyle and invent followup conversation! 
+    {chat_history}
+    user:{user_msg}
+    Chatbot:"""
+    prompt = PromptTemplate(
+        input_variables=["chat_history", "user_msg"],
+        template=template
+    )
+    memory = ConversationBufferMemory(memory_key="chat_history")
+    llm_chain = LLMChain(
+        llm = ChatOpenAI(openai_api_key=os.getenv('OPENAI_API_KEY'),temperature=0.8,model='gpt-3.5-turbo'),
+        prompt=prompt,
+        verbose=False,
+        memory=memory,
+    )
+    return llm_chain
+ 
+
+
+llm_chain = present_products_chain()
+
+@st.cache_resource()
+def embedding_model():
+    embedding_model = SentenceTransformer('sentence-transformers/all-distilroberta-v1')
+    return embedding_model
+
+embedding_model = embedding_model()
 
 if "messages" not in st.session_state:
     st.session_state["messages"] = [{"role": "assistant", "content": "Hey im your online shopping buddy, how can i help you today?"}]
 for msg in st.session_state["messages"]:
     st.chat_message(msg["role"]).write(msg["content"])
 
-prompt = st.chat_input(key="user_input",disabled=st.session_state.disabled )
-embedding_model = SentenceTransformer('sentence-transformers/all-distilroberta-v1')
+prompt = st.chat_input(key="user_input" )
+
 if prompt:
     st.session_state["messages"].append({"role": "user", "content": prompt})
     st.chat_message('user').write(prompt)
     st.session_state.disabled = True
     keywords = chain.run(prompt)
-
     #vectorize the query
     query_vector = embedding_model.encode(keywords)
     query_vector = np.array(query_vector).astype(np.float32).tobytes()

chatbot.py

@@ -1,47 +0,0 @@
-from langchain.prompts import PromptTemplate
-from langchain import HuggingFaceHub
-from langchain.chains import LLMChain
-from langchain.memory import ConversationBufferMemory
-from redis.commands.search.query import Query
-import time
-import os
-from dotenv import load_dotenv
-import numpy as np
-load_dotenv()
-HUGGINGFACEHUB_API_TOKEN = os.getenv('HUGGINGFACEHUB_API_TOKEN')
-repo_id = 'tiiuae/falcon-7b-instruct'
-
-falcon_llm_1 = HuggingFaceHub(repo_id = repo_id, model_kwargs={'temperature':0.1,'max_new_tokens':500},huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN)
-
-prompt = PromptTemplate(
-    input_variables=["product_description"],
-    template="Create comma seperated product keywords to perform a query on a amazon dataset for this user input: {product_description}",
-)
-
-chain = LLMChain(llm=falcon_llm_1, prompt=prompt)
-
-# code The response
-repo_id_2 = 'tiiuae/falcon-7b' 
-template = """You are a salesman. Be kind, detailed and nice.  take the given context and Present the given queried search result in a nice way as answer to the user_msg. dont ask questions back or freestyle and invent followup conversation! just 
-
-{chat_history}
-{user_msg}
-Chatbot:"""
-
-prompt = PromptTemplate(
-    input_variables=["chat_history", "user_msg"],
-    template=template
-)
-memory = ConversationBufferMemory(memory_key="chat_history")
-
-llm_chain = LLMChain(
-    llm = HuggingFaceHub(repo_id = repo_id_2, model_kwargs={'temperature':0.8,'max_new_tokens':500}),
-    prompt=prompt,
-    verbose=False,
-    memory=memory,
-)
-
-
-
-
-

database.py

@@ -11,5 +11,3 @@ redis_conn = redis.Redis(
   host='redis-12882.c259.us-central1-2.gce.cloud.redislabs.com',
   port=12882,
   password=redis_key)
-
-print('connected to redis')

preprocess.py

@@ -8,6 +8,7 @@ from database import redis_conn
 from utilities import create_flat_index, load_vectors
 
 
+
 #set maximum length for text fields
 MAX_TEXT_LENGTH = 512
 

readme.md

@@ -1,3 +1,13 @@
+---
+title: EcomShoppingBuddy
+emoji: 🌍
+colorFrom: pink
+colorTo: red
+sdk: streamlit
+sdk_version: 1.27.2
+app_file: app.py
+pinned: false
+---
 **Description:**  
 An ***e-commerce chatBot*** which goes through the Amazon dataset products and suggests the most suitable goods according to the user needs.  
 By utilizing the power of product embeddings and large language models exploiting Langchain and Redis technologies alongside the open source sentence-transformer embedding model and Falcon LLM, this chatbot acts as a real salesperson, can understand the client's request and efficiently search for relevant product recommendations based on the user description and present them in an engaging and informative manner.  

requirements.txt

@@ -2,6 +2,7 @@ langchain == 0.0.242
 openai == 0.27.8
 redis == 5.0.1
 pandas == 2.0.3
+numpy == 1.24.3
 sentence-transformers == 2.2.2
 tiktoken == 0.5.1
 streamlit == 1.27.2