import os import requests import gradio as gr import re # === API SETUP === GROQ_API_KEY = "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" # Bạn nên dùng biến môi trường GROQ_API_URL = "https://api.groq.com/openai/v1/chat/completions" HEADERS = { "Authorization": f"Bearer {GROQ_API_KEY}", "Content-Type": "application/json" } # === BLOCKED TOPICS === BLOCKED_KEYWORDS = [ "sex", "drugs", "violence", "suicide", "death", "kill", "murder", "gun", "abuse", "politics", "terror", "crime", "war", "religion", "racism", "nude", "adult", "nsfw" ] def is_safe_topic(topic: str) -> bool: topic_lower = topic.lower() return not any(bad_word in topic_lower for bad_word in BLOCKED_KEYWORDS) # === PODCAST GENERATION === def generate_educational_podcast(topic: str, language: str, model: str = "llama3-70b-8192") -> str: prompt = f""" You are a scriptwriter for an educational podcast for students in grades 8 to 12. Create a short and engaging podcast script in **{language}** on the topic: "{topic}". The two hosts are named Ali and Talha. Ali starts the conversation. They should have a back-and-forth conversation about the topic. The tone should be friendly, easy to understand, and informative. Avoid music or sound effects. Keep the total length under 800 words. Only write their dialogue and a few narration lines if needed. Do NOT use Host 1/2 or generic names. """ data = { "model": model, "messages": [ {"role": "system", "content": "You are a creative and age-appropriate educational podcast writer."}, {"role": "user", "content": prompt.strip()} ], "temperature": 0.75, "max_tokens": 1024 } response = requests.post(GROQ_API_URL, headers=HEADERS, json=data) if response.status_code == 200: return response.json()["choices"][0]["message"]["content"].strip() else: return f"Error {response.status_code}: {response.text}" # === SCRIPT FORMATTING === def format_script(script: str) -> str: script = re.sub(r'\*\*(.*?)\*\*', r'\1', script) script = re.sub(r'(?i)^.*(music|sound effect).*$','', script, flags=re.MULTILINE) script = script.replace("Host 1:", "Ali:").replace("Host 2:", "Talha:") script = re.sub(r'\b[Aa]lex:', 'Ali:', script) script = re.sub(r'\b[Mm]aya:', 'Talha:', script) lines = script.strip().split("\n") formatted = "" for line in lines: line = line.strip() if not line: continue if line.startswith("Ali:"): content = line.replace("Ali:", "").strip() formatted += f"
🎙️ Ali: {content}
\n" elif line.startswith("Talha:"): content = line.replace("Talha:", "").strip() formatted += f"
🎧 Talha: {content}
\n" return formatted # === GRADIO UI === custom_css = """ .host1, .host2 { border-radius: 16px; padding: 14px; margin: 16px 0; box-shadow: 0 4px 10px rgba(255,255,255,0.2); font-size: 16px; line-height: 1.6; color: #ffffff; } .host1 { background: linear-gradient(135deg, #4b6cb7, #182848); } .host2 { background: linear-gradient(135deg, #2c3e50, #4ca1af); } """ def generate(topic, language): if not is_safe_topic(topic): return "⚠️ Restricted content. Please enter an appropriate educational topic." raw_script = generate_educational_podcast(topic, language) return format_script(raw_script) with gr.Blocks(css=custom_css) as demo: gr.Markdown("## 🎙️ Educational Podcast Generator for Students") gr.Markdown("Enter a topic and choose your language to generate a friendly podcast between Ali and Talha.") topic_input = gr.Textbox(label="Enter educational topic") lang_input = gr.Dropdown( choices=["English", "Vietnamese", "Spanish", "French", "Japanese"], value="English", label="Select output language" ) generate_btn = gr.Button("🎧 Generate Podcast") output_box = gr.HTML() generate_btn.click(fn=generate, inputs=[topic_input, lang_input], outputs=output_box) demo.launch()