app.py

import streamlit as st
from langchain.llms import OpenAI
from langchain.prompts import PromptTemplate
from langchain import FewShotPromptTemplate
from langchain.prompts.example_selector import LengthBasedExampleSelector
from dotenv import load_dotenv


def getLLMResponse(query, age_option,tasktype_option):
    examples = []
    llm = OpenAI(temperature=.9, model="gpt-3.5-turbo-instruct")


    example_template = """
    Question: {query}
    Response: {answer}
    """

    example_prompt = PromptTemplate(
        input_variables=["query", "answer"],
        template=example_template
    )


    prefix = """You are a {template_ageoption}, and you are going to {template_tasktype_option} , 
    you give one answer for each query. it is strictly limited to 1 answer only, and the answer MUST be LESS THAN 200 words. 
    For a tweet, you SHOULD NOT give more than 280 characters. If it is not to write for a tweet, DO NOT give a tweet suggestion in your answer.
    """

    suffix = """
    Question: {template_userInput}
    Response: """

    example_selector = LengthBasedExampleSelector(
        examples=examples,
        example_prompt=example_prompt,
        max_length = numberOfWords
    )


    new_prompt_template = FewShotPromptTemplate(
        example_selector=example_selector,  # use example_selector instead of examples
        example_prompt=example_prompt,
        prefix=prefix,
        suffix=suffix,
        input_variables=["template_userInput","template_ageoption","template_tasktype_option"],
        example_separator="\n"
    )

  
    print(new_prompt_template.format(template_userInput=query,template_ageoption=age_option,template_tasktype_option=tasktype_option))
    response=llm(new_prompt_template.format(template_userInput=query,template_ageoption=age_option,template_tasktype_option=tasktype_option))
    print(response)

    return response

#UI Starts here
# Load environment variables
load_dotenv()  # Make sure your .env file path is correct

# Streamlit App Configuration
st.set_page_config(page_title="PitchPal: Your Friendly Copy Assistant",
                   page_icon="💻",
                   layout="wide",
                   initial_sidebar_state="collapsed")

# Custom CSS for styling
st.markdown(
    """
    <style>
    .big-font {
        font-size:20px !important;
    }
    .title-font {
        font-size:30px !important;
        font-weight: bold;
    }
    .streamlit-container {
        margin-top: 2rem;
    }
    </style>
    """, unsafe_allow_html=True)

# Header Section
st.markdown("<h1 style='text-align: center; color: #1144aa'>PitchPal: Your Efficient Sales Copy Assistant</h1>", unsafe_allow_html=True)
st.markdown("<h3 style='text-align: center; color: #333'>Craft compelling sales copy with ease</h3>", unsafe_allow_html=True)
st.markdown("<p style='text-align: right; font-size:14px;'>By <a href='https://entzyeung.github.io/portfolio/index.html'>Lorentz Yeung</a></p>", unsafe_allow_html=True)

# User Input Section with Improved Layout
col1, col2 = st.columns(2)

with col1:
    form_input = st.text_area('Enter the product or service:', 'PlayStation 6', height=150)

with col2:
    tasktype_option = st.selectbox(
        'Marketing copy type:',
        ('Twitter post', 'Sales copy', 'Product description'),
        index=1)
    age_option = st.selectbox(
        'Audience age group:',
        ('Below 18', '18-45', '46-65', '> 65'),
        index=1)

# Submit Button for Generating Sales Copy
submit = st.button("Generate Sales Copy")

if submit:
    response = getLLMResponse(form_input, age_option, tasktype_option)
    st.markdown("## Generated Sales Copy")
    st.write(response)  # Display the LLM response

# Note: Ensure that all functions and logic related to LLM response generation are correctly implemented
# and replace placeholder texts and functions with your actual application code.