add graph 2

UI/pocess_keyframes.py

@@ -1,11 +1,8 @@
+# app.py
+
+import os, shutil, time, json, sys, warnings
 import gradio as gr
 from pathlib import Path
-import warnings
-import os
-import shutil
-import json
-import time
-import sys
 
 sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), "..")))
 warnings.filterwarnings("ignore")
@@ -13,19 +10,11 @@ warnings.filterwarnings("ignore")
 from KeyFrameSelection.FeatureExtraction import process_video, save_records
 from KeyFrameSelection.Similarties import hash_filter, clip_filter
 from FrameProcessor.utils.io_utils import get_frames_from_folder, save_description_to_csv
-from FrameProcessor.processor.multi_frame import process_frames
+from FrameProcessor.graph.workflow import frame_processor
 from config.paths import output_csv_file, output_json_file
-# from llm.model import model
-from langchain_google_genai.chat_models import ChatGoogleGenerativeAI
-from dotenv import load_dotenv
-import os
 
-# Log printer
-def print_state(text: str):
-    yield text
 
-# Keyframe Extraction + Filtering
-def extract_and_filter_keyframes(video_path):
+def summarize_video(video_path):
     keyframe_dir = "outputs/keyframes"
     csv_path = "outputs/keyframes.csv"
 
@@ -33,13 +22,18 @@ def extract_and_filter_keyframes(video_path):
         shutil.rmtree("outputs")
     os.makedirs("outputs/final_output", exist_ok=True)
 
-    yield from print_state("✅ Starting video summarization...")
+    start = time.time()
+
+    # Step 1: Extract raw keyframes
     records, fps = process_video(video_path, interval_sec=10)
-    yield from print_state("🎞️ Keyframe extraction done.")
 
-    min_frames, max_iterations = 10, 20
+    # Step 2: Filter
+    min_frames = 10
+    max_iterations = 20
     iteration = 0
-    hash_threshold, ssim_threshold, clip_threshold = 5, 0.95, 0.90
+    hash_threshold = 5
+    ssim_threshold = 0.95
+    clip_threshold = 0.90
     filtered = records
 
     while len(filtered) >= min_frames and iteration < max_iterations:
@@ -51,108 +45,71 @@ def extract_and_filter_keyframes(video_path):
         iteration += 1
 
     save_records(filtered, keyframe_dir, csv_path, fps)
-    yield from print_state(f"✅ Frame filtering done: {len(filtered)} frames saved.")
-    yield from print_state("📂 Proceeding to multi-frame processing...")
-
     frame_paths = get_frames_from_folder(keyframe_dir)
-    yield frame_paths
-
-# Process frames using multi_frame.py
-def process_frames_batch(frame_paths):
-    if not frame_paths:
-        yield from print_state("❌ No frames to process.")
-        return
-
-    yield from print_state(f"🧠 Processing {len(frame_paths)} frames with model...")
-
-    results = process_frames(frame_paths)
-
-    for i, result in enumerate(results):
-        yield from print_state(f"🖼️ Frame: {os.path.basename(result['frame'])}")
-        yield from print_state(f"📌 Importance: {result.get('importance')}")
-        yield from print_state(f"📝 Reason: {result.get('reason')[:100]}")
-        yield from print_state("-" * 30)
 
-    yield results
-
-# Save results to JSON & CSV
-def save_outputs(results):
-    important_frames = [r for r in results if r["importance"] == "important"]
-    for result in important_frames:
-        save_description_to_csv(result, output_csv_file)
-
-    with open(output_json_file, 'w', encoding='utf-8') as f:
+    # Step 3: Graph processing on each frame
+    results = []
+    for frame_path in frame_paths:
+        state = {
+            "frame_path": frame_path,
+            "frame_data": {},
+            "frame_features": {},
+            "importance": "not_important",
+            "reason": "",
+            "description": {},
+            "next_step": "describe_frame"
+        }
+
+        try:
+            result = frame_processor.invoke(state)
+            results.append(result)
+
+            if result["importance"] == "important":
+                save_description_to_csv(result)
+
+        except Exception as e:
+            results.append({
+                "frame_path": frame_path,
+                "importance": "error",
+                "reason": str(e)
+            })
+
+    important = [r for r in results if r["importance"] == "important"]
+
+    with open(output_json_file, "w") as f:
         json.dump(results, f, indent=2, ensure_ascii=False)
 
-    yield from print_state("✅ Final saving done.")
-    yield from print_state(f"📝 Important frames: {len(important_frames)}")
-    yield from print_state(f"📁 CSV: {output_csv_file}")
-    yield from print_state(f"📁 JSON: {output_json_file}")
-    yield from print_state("🎉 All done!")
-
-# Main pipeline
-def run_full_pipeline(video_path):
-    start = time.time()
+    end = time.time()
+    return f"✅ Processed {len(important)} important frames out of {len(results)} in {end - start:.2f}s."
 
 
-    load_dotenv()
-
-    model = ChatGoogleGenerativeAI(
-                                    model="gemini-1.5-flash",
-                                    google_api_key=os.getenv("GOOGLE_API_KEY"),
-                                    convert_system_message_to_human=True
-                                    )
-
-    if not video_path:
-        yield from print_state("❌ No video uploaded.")
-        return
-
-    # Step 1: Extract
-    frame_paths_gen = extract_and_filter_keyframes(video_path)
-    frame_paths = None
-    for out in frame_paths_gen:
-        if isinstance(out, list):
-            frame_paths = out
-        else:
-            yield out
-
-    # Step 2: Process
-    results_gen = process_frames_batch(frame_paths)
-    results = None
-    for out in results_gen:
-        if isinstance(out, list):
-            results = out
-        else:
-            yield out
-
-    # Step 3: Save
-    for out in save_outputs(results):
-        yield out
+def process_uploaded_video(video_file):
+    if not video_file:
+        raise gr.Error("Please upload a video first.")
+    return summarize_video(video_file)
 
-    end = time.time()
-    yield from print_state(f"⏱️ Total Time: {end - start:.2f} sec")
 
-# Gradio UI
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("""
-    <div style='text-align: center; color: #e91e63; line-height: 1.8; margin-bottom: 30px;'>
-        <h1 style='margin-bottom: 20px;'>🎞️ Video Summarization UI</h1>
-        <p style='font-size: 18px;'>Upload your lecture or tutorial video</p>
-        <p style='font-size: 18px;'>Then click <b>Summarize</b> to extract key content</p>
-    </div>
-    """)
+    gr.Markdown(
+        """
+        <div style='text-align: center; color: #e91e63; line-height: 1.8; margin-bottom: 30px;'>
+            <h1>🎞️ Video Summarization UI hello</h1>
+            <p>Upload your lecture or tutorial video</p>
+            <p>Click <b>Summarize</b> to extract important frames and their content</p>
+        </div>
+        """
+    )
 
     with gr.Row():
         with gr.Column(scale=1, min_width=400):
             video_upload = gr.File(label="🎥 Upload Video", file_types=["video"], type="filepath")
-            summarize_btn = gr.Button("✨ Summarize", variant="primary", size="lg")
-            log_output = gr.Textbox(label="📋 Process Log", lines=25, interactive=False)
+            summarize_btn = gr.Button("✨ Summarize", variant="primary")
+            result_box = gr.Textbox(label="📄 Summary Result")
 
     summarize_btn.click(
-        fn=run_full_pipeline,
+        fn=process_uploaded_video,
         inputs=[video_upload],
-        outputs=[log_output],
-        show_progress=True
+        outputs=[result_box]
     )
 
 if __name__ == "__main__":