Initial Model

app.py

@@ -0,0 +1,61 @@
+import gradio as gr
+import pickle
+from chain import get_chain
+
+
+with open("db.pkl", "rb") as f:
+    global vectorstore
+    vectorstore = pickle.load(f)
+
+chain = get_chain()
+
+def chat(inp, history):
+    history = history or []
+    output = chain({'input_documents': vectorstore.similarity_search(inp, k=4), 'query': inp}, return_only_outputs=True)
+    response = output['output_text']
+    history.append((inp, response))
+    return history, history
+
+
+block = gr.Blocks(css=".gradio-container {background-color: lightgray}")
+
+with block:
+    with gr.Row():
+        with gr.Column(scale=1):
+            # gr.Markdown("""<img src=https://huggingface.co/spaces/.jpeg></left>""")
+            gr.Markdown("""<img alt="Image" title="icon" src="./images/octo.jpeg" />""")
+        with gr.Column(scale=15):
+            gr.Markdown("<h1><left>Your Beer Sommelier</left></h1>")
+            gr.Markdown("""<h3><left> Your friendly beer sommelier is here to help you choose the right beer for you!</left></h3>""")
+
+    chatbot = gr.Chatbot()
+
+    with gr.Row():
+        message = gr.Textbox(
+            label="What's your question?",
+            placeholder="Do you have any recommendations for IPA with tropical fruite flavors?",
+            lines=1,
+        )
+    with gr.Row():
+        submit = gr.Button(value="Send", variant="secondary", size="lg")
+
+    gr.Examples(
+        examples=[
+            "What does Bock taste like? Do you recommend it?",
+            "I like Lagunitas IPA. Do you have anything similar?",
+            "Do you have any recommendations for Pilsner?",
+            "What kind of German beers do you have?"
+        ],
+        inputs=message,
+    )
+
+    state = gr.State()
+
+    message.submit(chat, inputs=[message, state], outputs=[chatbot, state])
+    message.submit(lambda x: gr.update(value=''), [], [message])
+
+    submit.click(chat, inputs=[message, state], outputs=[chatbot, state])
+    submit.click(lambda x: gr.update(value=''), [], [message])
+
+
+block.launch(debug=True, share=True)

chain.py

@@ -0,0 +1,50 @@
+"""Generate QA Chain to answer the question given matches (db)"""
+import os
+import openai
+from langchain.chains.question_answering import load_qa_chain
+from langchain.chat_models import ChatOpenAI
+from langchain.llms import OpenAI
+from langchain.memory import ConversationBufferMemory, ConversationBufferWindowMemory
+from langchain.prompts import PromptTemplate
+from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
+
+openai.api_key  = os.environ['OPENAI_API_KEY']
+
+def get_chain():
+    template = """
+    You are a ChatBot at a restaurant with expertise in beers and good at making recommendations based on the user's preferences.
+    You are a given a context delimited by "=========", which is extracted part of the restaurant's beer list. 
+    Customer's question is given below delimited by "```````".
+
+    Based on the context, question and chat history, please respond in a friendly, conversational manner based on the context, describing features of beers. 
+    When asked to make recommendations, make three to four. Do not mention ABV or IBU unless asked about it. Remember you are trying to promote your beers.
+
+    If asked about something that is not related to beers in any absolute way, say you are only good at recommending beers in a witty way and redirect the conversation
+    by asking about customer's preferences such as beer style, flavor, etc..
+
+    If you don't find the answer to the user's question with the context provided to you below,
+    answer that you don't have the requested beer and ask about customer's preferences such as beer style, flavor, etc..
+    Finish by proposing your help for anything else.
+
+    CONTEXT:
+    =========
+    {context}
+    =========
+
+    QUESTION: 
+    ````````
+    {query}
+    ````````
+    CHAT HISTORY: 
+    {chat_history}
+
+    ANSWER:
+    """
+
+    prompt = PromptTemplate(input_variables=["chat_history", "query", "context"], template=template)
+    memory = ConversationBufferMemory(memory_key="chat_history", input_key="query")
+    chat = ChatOpenAI(temperature=0, streaming=True)
+
+    chain = load_qa_chain(llm=chat, chain_type="stuff", memory=memory, prompt=prompt)
+
+    return chain

ingest.py

@@ -0,0 +1,74 @@
+"""Scrape selected data from the 'untapped' website, put them in a list of dictionaries, 
+embed and store them into a vectorstore."""
+
+import requests
+import os
+import openai
+from bs4 import BeautifulSoup
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.embeddings import CohereEmbeddings, HuggingFaceHubEmbeddings
+from langchain.vectorstores import FAISS, Annoy, DocArrayInMemorySearch
+from langchain.document_loaders import CSVLoader
+import csv
+import pickle
+
+openai.api_key  = os.environ['OPENAI_API_KEY']
+
+
+def ingest_data():
+    headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/39.0.2171.95 Safari/537.36'}
+    url='https://untappd.com/v/gourmet-haus-staudt/15392'
+    response = requests.get(url, headers=headers)
+    soup = BeautifulSoup(response.text, 'html.parser')
+
+    links = soup.find_all('a', attrs={'data-track':'menu', 'data-href': ":"})
+
+    beer_list = []
+
+    for link in links:
+        page_url = 'https://untappd.com' + link.get('href')
+        
+        if page_url:
+            page_response = requests.get(page_url, headers=headers)
+            page_soup = BeautifulSoup(page_response.text, 'html.parser')
+            
+            # Extract the "relevant" page content
+            box_info = page_soup.find('div', class_="box b_info")
+
+            if box_info:
+                name = box_info.find('h1').get_text()
+                brewery = box_info.find(class_="brewery").get_text().strip()
+                style = box_info.find(class_="style").get_text()
+                description = box_info.find(class_="beer-descrption-read-less")
+                description.find('a').extract()
+                description = description.get_text().strip()
+                abv = box_info.find(class_="abv").get_text().strip()
+                ibu = box_info.find(class_="ibu").get_text().strip()
+                ratings = box_info.find(class_="num").get_text().replace("(", "").replace(")", "")
+                n_ratings = box_info.find(class_="raters").get_text().strip()
+            
+            beer_dict = {"name": name, "brewery": brewery, "style": style, "description": description, "alcohol_by_volume": abv, "ibu_bitterness_unit": ibu, "ratings": ratings, "n_ratings": n_ratings}
+            
+            beer_list.append(beer_dict)
+
+    keys = beer_list[0].keys()
+
+    with open('beers.csv', 'w', newline='') as output_file:
+        dict_writer = csv.DictWriter(output_file, keys)
+        dict_writer.writeheader()
+        dict_writer.writerows(beer_list)
+
+    file = 'beers.csv'
+    loader = CSVLoader(file_path=file)
+    data = loader.load()
+
+    embeddings = OpenAIEmbeddings()
+    db = FAISS.from_documents(data, embeddings)
+
+    # Save vectorstore
+
+    with open("db.pkl", "wb") as f:
+        pickle.dump(db, f)
+
+if __name__ == "__main__":
+    ingest_data()

requirement.txt

@@ -0,0 +1,10 @@
+langchain
+openai
+python-dotenv
+faiss-cpu
+annoy
+tiktoken
+cohere
+chromadb
+beautifulsoup4
+gradio