app.py

import gradio as gr
import requests
from bs4 import BeautifulSoup
from nltk import download, sent_tokenize
import google.generativeai as genai
import os
import re
import tempfile
import asyncio
import time

# Download NLTK data
download('punkt')
download('punkt_tab')

# Configure Gemini API using environment variable
api_key = os.environ.get("GEMINI_API_KEY")
if not api_key:
    raise ValueError("GEMINI_API_KEY not found in environment variables. Please set it in your environment.")

genai.configure(api_key=api_key)

# Use gemini-1.5-flash for faster and more accessible text analysis
try:
    model = genai.GenerativeModel('gemini-1.5-flash')
except Exception as e:
    # Fallback: List available models if the specified model is not found
    print(f"Error initializing model: {str(e)}")
    print("Available models:")
    for m in genai.list_models():
        print(m.name)
    raise ValueError("Failed to initialize gemini-1.5-flash. Check available models above and update the model name.")

# Prompt for Gemini to analyze text with specified output format
PROMPT = """
You are an AI content reviewer. Analyze the provided text for the following:
1. Grammar Issues: Identify and suggest corrections for grammatical errors.
2. Legal Policy Violations: Flag content that may violate common legal policies (e.g., copyright infringement, defamation, incitement to violence).
3. Crude/Abusive Language: Detect crude, offensive, or abusive language.
4. Sensitive Topics: Identify content related to sensitive topics such as racism, gender bias, or other forms of discrimination.

Return the results in the following markdown format:
# Blog Review Report

## Grammar Corrections
1. [Heading of issue]
   - CONTENT: [Exact line or part of text with the issue]
   - SUGGESTION: [Suggested correction]
   - ISSUE: [Description of the issue]

2. [Heading of next issue]
   - CONTENT: [Exact line or part of text with the issue]
   - SUGGESTION: [Suggested correction]
   - ISSUE: [Description of the issue]

[Continue numbering for additional issues or state "None detected"]

## Legal Policy Violations
- CONTENT: [Exact line or part of text with the issue]
  SUGGESTION: [Suggested action or correction]
  ISSUE: [Description of the legal violation]
[Or state "None detected"]

## Crude/Abusive Language
- [List instances of crude or abusive language or "None detected"]

## Sensitive Topics
- [List instances of sensitive topics or "None detected"]

For each issue, provide the exact text, a suggested correction or action, and a concise explanation. Be precise and ensure the output strictly follows the specified format.
"""

async def fetch_url_content(url):
    try:
        response = requests.get(url, timeout=10)
        response.raise_for_status()
        soup = BeautifulSoup(response.text, 'html.parser')
        # Extract text from common content tags
        content = ' '.join([p.get_text(strip=True) for p in soup.find_all(['p', 'article', 'div'])])
        return content if content else "No readable content found on the page."
    except Exception as e:
        return f"Error fetching URL: {str(e)}"

async def review_blog(text_input, url_input, progress=gr.Progress()):
    # Determine input type based on which field is populated
    if text_input and not url_input:
        input_type = "Text"
        input_text = text_input
    elif url_input and not text_input:
        input_type = "URL"
        input_text = url_input
    else:
        return "Review Blog", "Error: Please provide input in either the Text or URL tab, but not both.", gr.update(visible=False)

    # Handle empty input
    if not input_text:
        return "Review Blog", "Error: No input provided.", gr.update(visible=False)

    # Handle URL input
    if input_type == "URL":
        progress(0, desc="Fetching content...")
        input_text = await fetch_url_content(input_text)
        if input_text.startswith("Error"):
            return "Review Blog", input_text, gr.update(visible=False)

    # Tokenize input for analysis
    sentences = sent_tokenize(input_text)
    analysis_text = "\n".join(sentences)

    # Simulate progress for API call
    progress(0, desc="Generating report...")
    start_time = time.time()
    for i in range(1, 10):
        await asyncio.sleep(1)  # Simulate progress every second
        progress(i / 10, desc=f"Generating report... ({int(i * 10)}%)")
        if time.time() - start_time > 30:  # Timeout after 30 seconds
            return "Review Blog", "Error: API request timed out after 30 seconds.", gr.update(visible=False)

    # Query Gemini with the prompt
    try:
        response = await asyncio.to_thread(model.generate_content, PROMPT + "\n\nText to analyze:\n" + analysis_text)
        report = response.text.strip()
        # Ensure the response is markdown by removing any code fences
        report = re.sub(r'^markdown\n|$', '', report, flags=re.MULTILINE)
    except Exception as e:
        report = f"Error analyzing content with Gemini: {str(e)}"
        # Fallback: List available models for debugging
        print("Available models:")
        for m in genai.list_models():
            print(m.name)
        return "Review Blog", report, gr.update(visible=False)

    # Create a temporary file to store the report
    try:
        with tempfile.NamedTemporaryFile(mode='w', suffix='.md', delete=False, encoding='utf-8') as temp_file:
            temp_file.write(report)
            temp_file_path = temp_file.name
        progress(1.0, desc="Report generated!")
        return "Review Blog", report, gr.update(visible=True, value=temp_file_path)
    except Exception as e:
        return "Review Blog", f"Error creating temporary file: {str(e)}", gr.update(visible=False)

# Custom CSS for hover effect, loading state, and Inter font
custom_css = """
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');
.gradio-container {
    font-family: 'Inter', sans-serif !important;
}
.review-btn {
    transition: all 0.3s ease;
    font-weight: 500;
    background-color: #2c3e50;
    color: white;
    border-radius: 8px;
    padding: 10px 20px;
}
.review-btn:hover {
    background-color: #4CAF50;
    color: white;
    transform: scale(1.05);
}
.review-btn:disabled {
    opacity: 0.7;
    cursor: not-allowed;
}
.review-btn:disabled::after {
    content: ' ⏳';
}
.tab-nav button {
    font-family: 'Inter', sans-serif;
    font-weight: 500;
}
input, textarea {
    font-family: 'Inter', sans-serif;
}
.gr-progress {
    background-color: #e0e0e0;
    border-radius: 8px;
    overflow: hidden;
}
.gr-progress > div {
    background-color: #4CAF50;
    height: 20px;
    transition: width 0.3s ease;
}
"""

# Gradio UI with Tabs
with gr.Blocks(theme=gr.themes.Monochrome(), css=custom_css) as demo:
    gr.Markdown("# 📝 AI Blog Reviewer")
    gr.Markdown("Enter blog text or a URL to review for grammar, legal issues, crude language, and sensitive topics. The report is generated in markdown format.")

    with gr.Tabs():
        with gr.TabItem("Text"):
            text_input = gr.Textbox(lines=8, label="Blog Content", placeholder="Paste your blog text here...")
        with gr.TabItem("URL"):
            url_input = gr.Textbox(lines=1, label="Blog URL", placeholder="Enter the blog URL here...")

    status_button = gr.Button(value="Review Blog", elem_classes=["review-btn"])

    gr.Markdown("### 📄 Review Report")
    report_output = gr.Markdown()
    download_btn = gr.File(label="Download Report", visible=False)

    # Bind the review button to process inputs
    status_button.click(
        fn=review_blog,
        inputs=[text_input, url_input],
        outputs=[status_button, report_output, download_btn]
    )

demo.launch()