app.py

import streamlit as st
import pandas as pd
import requests
import json

# Define your API key and endpoint
api_key = 'AIzaSyAQ4dXlOkF8rPC21f6omTS4p6v-uJ2vVIg'
url = "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp:generateContent"
headers = {'Content-Type': 'application/json'}

# Cache the sentiment analysis function to improve performance
@st.cache_data
def analyze_sentiment(text):
    """
    Analyze the sentiment of the given text using the Gemini API.
    """
    system_prompt = """
    You are a Sentiment Analysis Tool (SEA). Analyze the following comments and classify the sentiment of each as positive, neutral, or negative.
    Return the results in the following format:
    Comment: <comment>
    Sentiment: <sentiment>
    ---
    Additionally, provide actionable insights into customer satisfaction trends in the following format:
    ### Suggestions for Improvement:
    - <suggestion 1>
    - <suggestion 2>
    """
    data = {
        "contents": [{
            "parts": [{"text": f"{system_prompt}\n\n{text}"}]
        }]
    }
    response = requests.post(url, headers=headers, json=data, params={'key': api_key})
    if response.status_code == 200:
        return response.json()
    else:
        st.error(f"Request failed with status code {response.status_code}: {response.text}")
        return None

def read_file_content(file, file_type):
    """
    Read the entire content of the file based on its type.
    """
    if file_type == 'csv':
        df = pd.read_csv(file)
        text = ' '.join(df.apply(lambda x: ' '.join(x.dropna().astype(str)), axis=1))
    elif file_type == 'xlsx':
        df = pd.read_excel(file)
        text = ' '.join(df.apply(lambda x: ' '.join(x.dropna().astype(str)), axis=1))
    elif file_type == 'json':
        df = pd.read_json(file)
        text = ' '.join(df.apply(lambda x: ' '.join(x.dropna().astype(str)), axis=1))
    elif file_type == 'txt' or file_type == 'md':
        text = file.read().decode('utf-8')
    else:
        st.error("Unsupported file type.")
        return None
    return text

def process_large_text(text, chunk_size=5000):
    """
    Split large text into smaller chunks for processing.
    """
    chunks = [text[i:i + chunk_size] for i in range(0, len(text), chunk_size)]
    return chunks

def display_sentiment_results(response_text, file_name):
    """
    Display sentiment analysis results for a single file.
    """
    # Parse comments and sentiments
    results = []
    lines = response_text.split('\n')
    comment = None
    sentiment = None
    for line in lines:
        if line.startswith("Comment:"):
            comment = line.replace("Comment:", "").strip()
        elif line.startswith("Sentiment:"):
            sentiment = line.replace("Sentiment:", "").strip()
            if comment and sentiment:
                results.append((comment, sentiment))
                comment = None
                sentiment = None

    # Display results
    st.write(f"### Sentiment Analysis Results for **{file_name}**")
    df_results = pd.DataFrame(results, columns=['Comment/Prompt', 'Sentiment'])
    st.dataframe(df_results)

    # Sentiment distribution
    sentiment_counts = df_results['Sentiment'].value_counts().reset_index()
    sentiment_counts.columns = ['Sentiment', 'Count']
    with st.expander(f"View Sentiment Distribution for {file_name}"):
        st.bar_chart(sentiment_counts.set_index('Sentiment'))

    # Suggestions
    suggestions = []
    current_section = None
    for line in lines:
        if line.startswith("### Suggestions for Improvement:"):
            current_section = "Suggestions for Improvement"
        elif current_section and line.startswith("- "):
            suggestions.append(line.replace("- ", "").strip())
    if suggestions:
        st.write("### Suggestions for Improvement")
        for suggestion in suggestions:
            st.write(f"- {suggestion}")
    else:
        st.warning("No suggestions available.")

    # CSV download
    output_file = f"sentiment_analysis_results_{file_name}.csv"
    df_results.to_csv(output_file, index=False)
    st.download_button(
        label=f"Download Results for {file_name} as CSV",
        data=open(output_file, 'rb').read(),
        file_name=output_file,
        mime='text/csv',
    )

# Streamlit layout
st.set_page_config(page_title="Sentiment Analysis Tool", layout="wide")
st.title("Sentiment Analysis Tool (SEA) 💬")
st.write("Analyze customer feedback with sentiment classification and actionable insights.")

# Sidebar for instructions
with st.sidebar:
    st.header("Instructions 📄")
    st.write("""
    1. Upload one or more files containing customer feedback in the main area.
    2. Analyze real-time feedback using the text input box.
    3. Download sentiment analysis results as CSV files.
    """)
    st.write("---")
    st.header("About")
    st.write("This app uses the Gemini API for sentiment analysis and provides actionable insights.")

# Main layout with tabs
tab1, tab2 = st.tabs(["📁 File Analysis", "✍️ Real-Time Feedback"])

with tab1:
    st.write("### Upload one or more files for batch sentiment analysis:")
    uploaded_files = st.file_uploader("Choose files", type=["csv", "xlsx", "json", "txt", "md"], accept_multiple_files=True)
    if uploaded_files:
        for uploaded_file in uploaded_files:
            file_type = uploaded_file.name.split('.')[-1]
            with st.spinner(f"Processing {uploaded_file.name}..."):
                # Read the entire file content
                text = read_file_content(uploaded_file, file_type)
                if text:
                    # Process large text in chunks if necessary
                    chunks = process_large_text(text)
                    combined_results = ""
                    for chunk in chunks:
                        sentiment_result = analyze_sentiment(chunk)
                        if sentiment_result:
                            response_text = sentiment_result.get('candidates', [{}])[0].get('content', {}).get('parts', [{}])[0].get('text', '').strip()
                            combined_results += response_text + "\n"
                    if combined_results:
                        display_sentiment_results(combined_results, uploaded_file.name)
                    else:
                        st.error(f"Sentiment analysis failed for {uploaded_file.name}.")

with tab2:
    st.write("### Enter your feedback for real-time analysis:")
    feedback_input = st.text_area("Enter your feedback:", placeholder="Type your feedback here...")
    if st.button("Analyze Sentiment"):
        if feedback_input.strip() == "":
            st.warning("Please enter some feedback to analyze.")
        else:
            with st.spinner("Analyzing sentiment..."):
                sentiment_result = analyze_sentiment(feedback_input)
                if sentiment_result:
                    sentiment = sentiment_result.get('candidates', [{}])[0].get('content', {}).get('parts', [{}])[0].get('text', '').strip().lower()
                    if "positive" in sentiment:
                        st.success(f"Sentiment: **Positive** 😊")
                    elif "neutral" in sentiment:
                        st.info(f"Sentiment: **Neutral** 😐")
                    elif "negative" in sentiment:
                        st.error(f"Sentiment: **Negative** 😠")
                    else:
                        st.warning(f"Sentiment: **Unknown** 🤔")
                else:
                    st.error("Sentiment analysis failed.")