Create app.py

app.py

@@ -0,0 +1,215 @@
+import gradio as gr
+import torch
+from transformers import pipeline
+import os
+
+print("Loading Whisper transcription model...")
+transcriber = pipeline(
+    "automatic-speech-recognition",
+    model="openai/whisper-base",
+    chunk_length_s=30,
+    stride_length_s=5,
+    return_timestamps=False,
+    device=0 if torch.cuda.is_available() else -1,
+)
+
+print("Loading summarization model...")
+summarizer = pipeline(
+    "summarization",
+    model="sshleifer/distilbart-cnn-12-6",
+    device=0 if torch.cuda.is_available() else -1,
+)
+
+print("Models ready.")
+
+
+def transcribe_audio(audio_path):
+    result = transcriber(audio_path)
+    return result["text"].strip()
+
+
+def chunk_text(text, max_tokens=900):
+    words = text.split()
+    chunks, current = [], []
+    for word in words:
+        current.append(word)
+        if len(current) >= max_tokens:
+            chunks.append(" ".join(current))
+            current = []
+    if current:
+        chunks.append(" ".join(current))
+    return chunks
+
+
+def summarize_transcript(transcript):
+    if not transcript.strip():
+        return "No transcript available to summarize."
+
+    word_count = len(transcript.split())
+
+    if word_count <= 900:
+        result = summarizer(
+            transcript,
+            max_length=200,
+            min_length=60,
+            do_sample=False,
+        )
+        return result[0]["summary_text"]
+
+    chunks = chunk_text(transcript, max_tokens=900)
+    chunk_summaries = []
+    for chunk in chunks:
+        r = summarizer(chunk, max_length=150, min_length=40, do_sample=False)
+        chunk_summaries.append(r[0]["summary_text"])
+
+    combined = " ".join(chunk_summaries)
+    if len(combined.split()) > 900:
+        combined = " ".join(combined.split()[:900])
+
+    final = summarizer(combined, max_length=250, min_length=80, do_sample=False)
+    return final[0]["summary_text"]
+
+
+def extract_action_items(transcript):
+    action_keywords = [
+        "will ", "should ", "need to ", "must ", "action:",
+        "todo:", "follow up", "follow-up", "assign", "deadline",
+        "by next", "responsible", "let's ", "we'll ", "i'll ", "you'll ",
+    ]
+    sentences = [
+        s.strip()
+        for s in transcript.replace("\n", " ").split(".")
+        if len(s.strip()) > 15
+    ]
+    actions = []
+    for sentence in sentences:
+        lower = sentence.lower()
+        if any(kw in lower for kw in action_keywords):
+            actions.append(f"• {sentence.strip()}.")
+
+    if not actions:
+        return "No specific action items detected."
+    return "\n".join(actions[:10])
+
+
+def extract_key_topics(summary):
+    stop_words = {
+        "the", "a", "an", "is", "are", "was", "were", "be", "been",
+        "being", "have", "has", "had", "do", "does", "did", "will",
+        "would", "could", "should", "may", "might", "shall", "can",
+        "and", "but", "or", "nor", "so", "yet", "both", "either",
+        "neither", "not", "only", "own", "same", "than", "too", "very",
+        "just", "because", "as", "until", "while", "of", "in", "on",
+        "at", "by", "for", "with", "about", "into", "through", "during",
+        "before", "after", "to", "from", "up", "down", "out", "this",
+        "that", "these", "those", "it", "its", "they", "their", "there",
+        "we", "our", "you", "your", "he", "she", "his", "her", "also",
+        "if", "any", "then", "what", "which", "who", "how", "all", "each",
+    }
+    words = summary.lower().split()
+    freq = {}
+    for w in words:
+        w_clean = w.strip(".,!?;:()'\"")
+        if w_clean and w_clean not in stop_words and len(w_clean) > 3:
+            freq[w_clean] = freq.get(w_clean, 0) + 1
+
+    top = sorted(freq, key=freq.get, reverse=True)[:8]
+    if not top:
+        return "Topics could not be extracted."
+    return "  •  ".join(t.title() for t in top)
+
+
+def analyze_meeting(audio_file):
+    if audio_file is None:
+        return ("Please upload an audio file.", "", "", "", "")
+
+    try:
+        transcript = transcribe_audio(audio_file)
+        if not transcript:
+            return ("Transcription produced no text. Try a clearer audio file.", "", "", "", "")
+
+        summary = summarize_transcript(transcript)
+        actions = extract_action_items(transcript)
+        topics = extract_key_topics(summary)
+
+        word_count = len(transcript.split())
+        stats = f"📊 {word_count} words transcribed  |  ~{word_count // 130 + 1} min read"
+
+        return transcript, summary, actions, topics, stats
+
+    except Exception as e:
+        return (f"Error during processing: {str(e)}", "", "", "", "")
+
+
+with gr.Blocks(
+    title="Meeting Audio Analyzer",
+    theme=gr.themes.Soft(),
+    css="""
+        #title { text-align: center; margin-bottom: 0.5rem; }
+        #subtitle { text-align: center; color: #666; margin-bottom: 1.5rem; font-size: 0.95rem; }
+        footer { display: none !important; }
+    """,
+) as demo:
+
+    gr.Markdown("# Meeting Audio Analyzer", elem_id="title")
+    gr.Markdown(
+        "Upload a meeting recording — get a full transcript, summary, action items, and key topics.",
+        elem_id="subtitle",
+    )
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            audio_input = gr.Audio(
+                label="Upload Meeting Audio",
+                type="filepath",
+                sources=["upload"],
+            )
+            analyze_btn = gr.Button("Analyze Meeting", variant="primary", size="lg")
+            stats_out = gr.Markdown(value="", label="")
+
+        with gr.Column(scale=2):
+            with gr.Tabs():
+                with gr.TabItem("Summary"):
+                    summary_out = gr.Textbox(
+                        label="Meeting Summary",
+                        lines=8,
+                        interactive=False,
+                        placeholder="Summary will appear here after analysis...",
+                    )
+                with gr.TabItem("Action Items"):
+                    actions_out = gr.Textbox(
+                        label="Action Items",
+                        lines=8,
+                        interactive=False,
+                        placeholder="Action items will appear here...",
+                    )
+                with gr.TabItem("Key Topics"):
+                    topics_out = gr.Textbox(
+                        label="Key Topics",
+                        lines=3,
+                        interactive=False,
+                        placeholder="Key topics will appear here...",
+                    )
+                with gr.TabItem("Full Transcript"):
+                    transcript_out = gr.Textbox(
+                        label="Full Transcript",
+                        lines=15,
+                        interactive=False,
+                        placeholder="Full transcript will appear here...",
+                    )
+
+    analyze_btn.click(
+        fn=analyze_meeting,
+        inputs=[audio_input],
+        outputs=[transcript_out, summary_out, actions_out, topics_out, stats_out],
+        show_progress=True,
+    )
+
+    gr.Markdown(
+        "Models: [Whisper Base](https://huggingface.co/openai/whisper-base) · "
+        "[DistilBART CNN](https://huggingface.co/sshleifer/distilbart-cnn-12-6) — "
+        "runs fully locally, no API keys needed."
+    )
+
+if __name__ == "__main__":
+    demo.launch()