Update main.py

main.py

@@ -6,16 +6,23 @@ No Gradio, no routing conflicts.
 
 import os
 import json
-import base64
 import tempfile
 import logging
-import time
 import requests
 import numpy as np
+import cloudinary
+import cloudinary.uploader
 from fastapi import FastAPI, Request
 from fastapi.responses import StreamingResponse, JSONResponse
 from fastapi.middleware.cors import CORSMiddleware
 
+# Cloudinary config — set these in your HF Space secrets
+cloudinary.config(
+    cloud_name = os.environ.get("CLOUD_NAME"),
+    api_key    = os.environ.get("API_KEY"),
+    api_secret = os.environ.get("API_SECRET"),
+)
+
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
@@ -86,8 +93,20 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
         yield {"status": "processing", "step": 5, "message": "Step 5/5 — Summarizing..."}
         summary = translator.summarize(transcript)
 
-        with open(clean1, "rb") as f:
-            enhanced_b64 = base64.b64encode(f.read()).decode("utf-8")
+        # Upload enhanced audio to Cloudinary — returns a URL instead of base64.
+        # This keeps the done SSE event tiny (~200 bytes) instead of ~700KB,
+        # which was causing the JSON to be split across 85+ TCP chunks.
+        try:
+            upload_result = cloudinary.uploader.upload(
+                clean1,
+                resource_type = "video",  # Cloudinary uses "video" for audio
+                folder        = "clearwave_enhanced",
+            )
+            enhanced_url = upload_result["secure_url"]
+            logger.info(f"Enhanced audio uploaded: {enhanced_url}")
+        except Exception as e:
+            logger.error(f"Cloudinary upload failed: {e}")
+            enhanced_url = None
 
         yield {
             "status":        "done",
@@ -96,7 +115,7 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
             "transcript":    transcript,
             "translation":   translation,
             "summary":       summary,
-            "enhancedAudio": enhanced_b64,
+            "enhancedAudio": enhanced_url,
             "stats": {
                 "language":             detected_lang.upper(),
                 "noise_method":         stats.get("noise_method", "noisereduce"),
@@ -185,4 +204,4 @@ async def process_url(request: Request):
         generate(),
         media_type="text/event-stream",
         headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
-    )
+    )