Update app.py

app.py

@@ -82,9 +82,11 @@ def extract_key_frames(video_file: str, key_frames_response: str) -> list:
     # Strip Markdown code block if present
     cleaned_response = key_frames_response.strip()
     if cleaned_response.startswith("```json") and cleaned_response.endswith("```"):
-        cleaned_response = cleaned_response[7:-3].strip()  # Remove ```json and ```
+        cleaned_response = cleaned_response[7:-3].strip()
     elif cleaned_response.startswith("```") and cleaned_response.endswith("```"):
-        cleaned_response = cleaned_response[3:-3].strip()  # Remove generic ```
+        cleaned_response = cleaned_response[3:-3].strip()
+    
+    print(f"Cleaned key frames response: {cleaned_response}")  # Debug output
     
     try:
         # Try parsing as JSON
@@ -92,15 +94,21 @@ def extract_key_frames(video_file: str, key_frames_response: str) -> list:
         if not isinstance(key_frames, list):
             raise ValueError("Response is not a list.")
     except json.JSONDecodeError as e:
-        print(f"JSON parsing failed: {str(e)}. Raw response: {key_frames_response}")
-        # Fallback: Attempt to parse plain text with timecodes (e.g., "00:00:03 - Scene")
+        print(f"JSON parsing failed: {str(e)}. Falling back to text parsing.")
+        # Fallback: Parse plain text with timecodes (e.g., "00:00:03 - Scene" or "00:00:03: Scene")
         key_frames = []
-        lines = key_frames_response.strip().split("\n")
+        lines = cleaned_response.strip().split("\n")
         for line in lines:
+            line = line.strip()
+            if not line:
+                continue
             if " - " in line:
                 timestamp, title = line.split(" - ", 1)
                 key_frames.append({"timecode": timestamp.strip(), "title": title.strip()})
-            elif ":" in line and len(line.split(":")) == 3:  # Rough check for HH:MM:SS
+            elif ": " in line and len(line.split(":")[0]) == 2:  # Check for HH:MM:SS format
+                timestamp, title = line.split(": ", 1)
+                key_frames.append({"timecode": timestamp.strip(), "title": title.strip()})
+            elif len(line.split(":")) == 3:  # Rough check for standalone HH:MM:SS
                 key_frames.append({"timecode": line.strip(), "title": "Untitled"})
     
     for frame in key_frames:
@@ -155,11 +163,26 @@ def analyze_video(video_file: str, user_query: str) -> tuple[str, list]:
         )
         summary = summary_response.text
 
-        # Step 2: Extract key frames
+        # Step 2: Extract key frames with few-shot examples
         key_frames_prompt = (
             "Identify key frames in this video and return them as a JSON array. "
             "Each object must have 'timecode' (in HH:MM:SS format) and 'title' describing the scene. "
-            "Ensure the response is valid JSON."
+            "Ensure the response is valid JSON. Here are examples of the expected format:\n"
+            "Example 1: For a video of a car chase:\n"
+            "```json\n"
+            "[\n"
+            "  {\"timecode\": \"00:00:00\", \"title\": \"Car chase begins on highway\"},\n"
+            "  {\"timecode\": \"00:00:10\", \"title\": \"Police car joins pursuit\"}\n"
+            "]\n"
+            "```\n"
+            "Example 2: For a nature video:\n"
+            "```json\n"
+            "[\n"
+            "  {\"timecode\": \"00:00:05\", \"title\": \"Bird flies across screen\"},\n"
+            "  {\"timecode\": \"00:00:15\", \"title\": \"Deer appears in forest\"}\n"
+            "]\n"
+            "```\n"
+            "Now, provide the key frames for this video in the same JSON format."
         )
         if user_query:
             key_frames_prompt += f" Focus on: {user_query}"
@@ -170,7 +193,7 @@ def analyze_video(video_file: str, user_query: str) -> tuple[str, list]:
         )
         key_frames = extract_key_frames(video_file, key_frames_response.text)
 
-        # Generate Markdown report (without Video URI)
+        # Generate Markdown report
         markdown_report = (
             "## Video Analysis Report\n\n"
             f"**Summary:**\n{summary}\n"
@@ -180,7 +203,7 @@ def analyze_video(video_file: str, user_query: str) -> tuple[str, list]:
             for i, (_, caption) in enumerate(key_frames, 1):
                 markdown_report += f"- Frame {i}: {caption}\n"
         else:
-            markdown_report += "\n*No key frames extracted. The model may not have returned valid JSON.*\n"
+            markdown_report += "\n*No key frames extracted. Check the console for the raw response.*\n"
 
         return markdown_report, key_frames