content_engine.py

import os
import re
import requests
import markdown
from dotenv import load_dotenv
from langchain_google_genai import ChatGoogleGenerativeAI

load_dotenv()

# ---------------------------------------------------------------
# Gemini LLM
# ---------------------------------------------------------------
llm = ChatGoogleGenerativeAI(
    model="gemini-2.5-flash",
    google_api_key=os.getenv("GOOGLE_API_KEY"),
    temperature=0.3,
)


# ---------------------------------------------------------------
# Extract video ID from any YouTube URL format
# ---------------------------------------------------------------
def extract_video_id(url: str) -> str:
    patterns = [
        r"(?:v=)([A-Za-z0-9_-]{11})",
        r"youtu\.be/([A-Za-z0-9_-]{11})",
        r"shorts/([A-Za-z0-9_-]{11})",
        r"embed/([A-Za-z0-9_-]{11})",
    ]
    for pattern in patterns:
        match = re.search(pattern, url)
        if match:
            return match.group(1)
    raise ValueError(
        f"Could not extract a valid YouTube video ID from: {url}\n"
        "Please use a standard YouTube link like https://www.youtube.com/watch?v=..."
    )


# ---------------------------------------------------------------
# Fetch transcript via Supadata API
# Sign up free at https://supadata.ai — 500 transcripts/month
# Add SUPADATA_API_KEY to HF Space secrets
# ---------------------------------------------------------------
def fetch_transcript(video_id: str) -> str:
    api_key = os.getenv("SUPADATA_API_KEY")

    if not api_key:
        raise EnvironmentError(
            "SUPADATA_API_KEY is not set.\n"
            "• Sign up free at https://supadata.ai\n"
            "• Add SUPADATA_API_KEY to HF Space → Settings → Variables and Secrets"
        )

    try:
        resp = requests.get(
            "https://api.supadata.ai/v1/youtube/transcript",
            headers={"x-api-key": api_key},
            params={"videoId": video_id, "lang": "en"},
            timeout=30,
        )
    except requests.exceptions.ConnectionError:
        raise ConnectionError("Could not reach Supadata API. Check your internet connection.")
    except requests.exceptions.Timeout:
        raise TimeoutError("Supadata API timed out. Try again.")

    if resp.status_code == 401:
        raise EnvironmentError(
            "Invalid SUPADATA_API_KEY. Check the key in your HF Space secrets."
        )
    if resp.status_code == 404:
        raise ValueError(
            "No transcript found for this video.\n"
            "The video may have no subtitles, be private, or age-restricted."
        )
    if resp.status_code == 429:
        raise RuntimeError(
            "Supadata free tier limit reached (500/month).\n"
            "Upgrade at https://supadata.ai or try again next month."
        )
    if not resp.ok:
        raise RuntimeError(f"Supadata API error {resp.status_code}: {resp.text}")

    data = resp.json()

    # Supadata returns: { "content": [ {"text": "...", "offset": 0, "duration": 2} ] }
    content = data.get("content", [])
    if not content:
        raise ValueError("Supadata returned an empty transcript.")

    # Join all text chunks into a single string
    transcript = " ".join(
        chunk.get("text", "") for chunk in content if chunk.get("text")
    ).strip()

    if not transcript:
        raise ValueError("Transcript was empty after processing.")

    return transcript


# ---------------------------------------------------------------
# Main transcript entry point
# ---------------------------------------------------------------
def extract_transcript(url: str) -> str:
    video_id = extract_video_id(url)
    return fetch_transcript(video_id)


# ---------------------------------------------------------------
# Summarize transcript into a Markdown article
# ---------------------------------------------------------------
def summarize(text: str) -> str:
    safe_text = text[:8000]
    prompt = f"""Convert this YouTube transcript into a HIGH-QUALITY Markdown blog article.

STRICT RULES:
- Use ## for main sections
- Use ### for sub-sections
- Use bullet points (-) for lists
- Each bullet on its own line
- Clean, readable prose — no filler words
- No inline messy formatting
- Do NOT include a title (it will be added separately)

Transcript:
{safe_text}
"""
    return llm.invoke(prompt).content


# ---------------------------------------------------------------
# Generate a short blog title
# ---------------------------------------------------------------
def generate_title(article: str) -> str:
    prompt = f"""Generate ONLY ONE short, catchy blog post title for this article.

Rules:
- One line only
- No quotes, no explanation, no numbering
- Suitable as an HTML <title> tag

Article (first 2000 chars):
{article[:2000]}
"""
    raw = llm.invoke(prompt).content.strip()
    return raw.split("\n")[0].strip().strip('"').strip("'")


# ---------------------------------------------------------------
# Full pipeline: URL → article text
# ---------------------------------------------------------------
def generate_article_from_url(url: str) -> str:
    transcript = extract_transcript(url)
    return summarize(transcript)


# ---------------------------------------------------------------
# Convert Markdown article → styled HTML file
# ---------------------------------------------------------------
def generate_web_files(article: str) -> None:
    try:
        title = generate_title(article)
    except Exception:
        title = "YouTube Generated Article"

    word_count = len(article.split())
    reading_time = max(1, word_count // 200)
    html_body = markdown.markdown(article, extensions=["extra"])

    html_content = f"""<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>{title}</title>
    <style>
        body {{
            font-family: Georgia, 'Times New Roman', serif;
            max-width: 750px;
            margin: 0 auto;
            padding: 24px 20px;
            line-height: 1.8;
            color: #1a1a1a;
            background: #fff;
        }}
        h1 {{ font-size: 2rem; margin-bottom: 4px; line-height: 1.3; }}
        h2 {{ font-size: 1.4rem; margin-top: 2rem; border-bottom: 1px solid #e0e0e0; padding-bottom: 4px; }}
        h3 {{ font-size: 1.15rem; margin-top: 1.5rem; }}
        p {{ margin: 0.9rem 0; }}
        ul, ol {{ padding-left: 1.4rem; margin: 0.7rem 0; }}
        li {{ margin: 0.3rem 0; }}
        .meta {{ color: #888; font-size: 0.9rem; font-family: sans-serif; margin-bottom: 2rem; }}
    </style>
</head>
<body>
    <h1>{title}</h1>
    <p class="meta">⏱ {reading_time} min read &nbsp;|&nbsp; Generated from YouTube</p>
    {html_body}
</body>
</html>
"""
    with open("generated_article.html", "w", encoding="utf-8") as f:
        f.write(html_content)