up

main.py

@@ -1,6 +1,6 @@
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
-from youtube_transcript_api import YouTubeTranscriptApi, TranscriptsDisabled, NoTranscriptFound, VideoUnavailable
+import requests
 import re
 from pydantic import BaseModel
 
@@ -15,7 +15,6 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# Pydantic model to ensure we receive proper JSON data
 class TranscriptRequest(BaseModel):
     url: str
 
@@ -26,6 +25,17 @@ def extract_video_id(url: str) -> str:
         return match.group(1)
     return None
 
+def parse_caption_xml(xml_string: str) -> str:
+    # Extract text content from the XML safely using regex
+    text_matches = re.findall(r'<text[^>]*>([^<]*)</text>', xml_string)
+    if not text_matches:
+        return ""
+    
+    full_text = " ".join(text_matches)
+    # Clean up HTML entities
+    full_text = full_text.replace("&amp;#39;", "'").replace("&amp;quot;", '"').replace("&amp;amp;", "&").replace("&#39;", "'").replace("&lt;", "<").replace("&gt;", ">")
+    return full_text
+
 @app.post("/api/transcript")
 async def get_transcript(req: TranscriptRequest):
     url = req.url
@@ -35,52 +45,87 @@ async def get_transcript(req: TranscriptRequest):
         raise HTTPException(status_code=400, detail="Invalid YouTube URL")
 
     try:
-        print(f"Attempting to fetch transcript for video_id: {video_id}")
+        print(f"Fetching InnerTube data for video_id: {video_id}")
+        
+        # YouTube's internal API endpoint
+        innertube_url = "https://www.youtube.com/youtubei/v1/player?prettyPrint=false"
+        
+        payload = {
+            "context": {
+                "client": {
+                    "clientName": "WEB",
+                    "clientVersion": "2.20241210.00.00",
+                    "hl": "en",
+                    "gl": "US"
+                }
+            },
+            "videoId": video_id
+        }
+        
+        headers = {
+            "Content-Type": "application/json",
+            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36",
+            "Origin": "https://www.youtube.com",
+            "Referer": "https://www.youtube.com/"
+        }
         
-        transcript_data = None
+        # 1. Fetch the video player data
+        response = requests.post(innertube_url, json=payload, headers=headers, timeout=5)
         
-        # 1. Try fetching English transcript first
-        try:
-            print("Trying English transcript...")
-            transcript_data = YouTubeTranscriptApi.get_transcript(video_id, languages=['en'])
-            print("Found English transcript.")
-        except NoTranscriptFound:
-            print("No English transcript found. Trying any available language...")
-            # 2. If no English, fetch the first available language automatically
-            try:
-                transcript_data = YouTubeTranscriptApi.get_transcript(video_id)
-                print("Found fallback language transcript.")
-            except Exception as e:
-                print(f"Fallback failed: {e}")
-                pass
+        if response.status_code != 200:
+            raise HTTPException(status_code=500, detail=f"YouTube API returned status {response.status_code}")
+            
+        data = response.json()
         
-        if not transcript_data:
+        # Check if video is playable
+        playability = data.get("playabilityStatus", {})
+        if playability.get("status") == "ERROR":
+            raise HTTPException(status_code=404, detail=playability.get("reason", "Video unavailable"))
+            
+        # 2. Extract caption tracks
+        caption_tracks = data.get("captions", {}).get("playerCaptionsTracklistRenderer", {}).get("captionTracks", [])
+        
+        if not caption_tracks:
             raise HTTPException(status_code=404, detail="No captions found for this video.")
-
-        # Combine all text lines safely
-        full_text = " ".join([line.get('text', '') for line in transcript_data])
+            
+        # Prefer English, fallback to first available
+        selected_track = None
+        for track in caption_tracks:
+            if track.get("languageCode") == "en":
+                selected_track = track
+                break
+                
+        if not selected_track:
+            selected_track = caption_tracks[0]
+            
+        caption_url = selected_track.get("baseUrl")
+        if not caption_url:
+            raise HTTPException(status_code=404, detail="Caption URL not found.")
+            
+        print(f"Fetching captions from URL...")
+        
+        # 3. Fetch the actual caption XML
+        caption_response = requests.get(caption_url, headers={"User-Agent": headers["User-Agent"]}, timeout=5)
         
-        # Clean up HTML entities
-        full_text = full_text.replace("&amp;#39;", "'").replace("&amp;quot;", '"').replace("&amp;amp;", "&").replace("&#39;", "'")
+        if caption_response.status_code != 200:
+            raise HTTPException(status_code=500, detail=f"Caption fetch failed with status {caption_response.status_code}")
+            
+        # 4. Parse the XML into plain text
+        full_text = parse_caption_xml(caption_response.text)
         
+        if not full_text.strip():
+            raise HTTPException(status_code=404, detail="Caption XML was empty.")
+            
         print(f"Successfully fetched transcript. Length: {len(full_text)} chars")
         return {"text": full_text}
 
     except HTTPException:
-        raise # Re-raise our custom HTTP exceptions
-    except TranscriptsDisabled:
-        raise HTTPException(status_code=403, detail="Captions are disabled for this video.")
-    except VideoUnavailable:
-        raise HTTPException(status_code=404, detail="Video is unavailable.")
+        raise 
+    except requests.exceptions.Timeout:
+        raise HTTPException(status_code=504, detail="Request to YouTube timed out.")
     except Exception as e:
-        # Catch-all: Log the exact error to Hugging Face logs so we can see what went wrong!
         error_name = type(e).__name__
         print(f"UNHANDLED ERROR for {video_id}: {error_name} - {str(e)}")
-        
-        # Check if it's a rate limit error dynamically
-        if "TooManyRequests" in error_name or "429" in str(e):
-             raise HTTPException(status_code=429, detail="YouTube is rate-limiting requests. Try again in a minute.")
-        
         raise HTTPException(status_code=500, detail=f"Failed to fetch transcript: {error_name} - {str(e)}")
 
 # KEEP-ALIVE ENDPOINT