Fix Railway deployment: encoding bug, healthcheck, deps pinning, Dockerfile optimization

- worker/index.js: Remove Content-Encoding forwarding (CF Workers auto-decompress)
- main.py: Add /health endpoint, make Playwright pre-warm non-fatal
- railway.toml: Add [deploy] with healthcheck, restart policy
- requirements.txt: Pin all dependency versions
- Dockerfile: Use playwright install --with-deps, optimize layer caching

Dockerfile

@@ -1,37 +1,22 @@
 FROM python:3.11-slim
 
-# Install system dependencies for Playwright Chromium
+# Install ffmpeg (not a Playwright dep, needed for Instagram audio extraction)
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    wget \
     ffmpeg \
-    libnss3 \
-    libnspr4 \
-    libatk1.0-0 \
-    libatk-bridge2.0-0 \
-    libcups2 \
-    libdrm2 \
-    libxkbcommon0 \
-    libxcomposite1 \
-    libxdamage1 \
-    libxfixes3 \
-    libxrandr2 \
-    libgbm1 \
-    libpango-1.0-0 \
-    libcairo2 \
-    libasound2t64 \
-    libatspi2.0-0 \
-    libwayland-client0 \
     && rm -rf /var/lib/apt/lists/*
 
 WORKDIR /app
 
+# Layer 1: Python dependencies (cached unless requirements.txt changes)
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Install Playwright Chromium browser
-RUN playwright install chromium
+# Layer 2: Playwright Chromium + all system deps (auto-installs libnss3, libatk, etc.)
+# Using --with-deps eliminates manual apt-get for ~15 Chromium libraries
+RUN playwright install --with-deps chromium
 
+# Layer 3: Application code (changes most frequently)
 COPY . .
 
-# Railway provides PORT env var
-CMD uvicorn main:app --host 0.0.0.0 --port ${PORT:-8000}
+# Railway provides PORT env var; default to 8000 for local dev
+CMD ["sh", "-c", "uvicorn main:app --host 0.0.0.0 --port ${PORT:-8000}"]

main.py

@@ -413,8 +413,11 @@ def _pw_init_browser():
     return _ig_browser
 
 
-# Pre-warm browser at import time
-_pw_executor.submit(_pw_init_browser)
+# Pre-warm browser at import time (fire-and-forget; failure is non-fatal)
+try:
+    _pw_executor.submit(_pw_init_browser)
+except Exception:
+    logger.warning("[instagram] Failed to submit browser pre-warm task")
 
 
 def _pw_extract_embed(shortcode):
@@ -884,3 +887,9 @@ async def submit_feedback(request: FeedbackRequest):
 @app.get("/")
 async def root():
     return FileResponse("static/index.html")
+
+
+@app.get("/health")
+async def health_check():
+    """Lightweight health check for Railway. No external dependencies."""
+    return {"status": "ok"}

railway.toml

@@ -1,3 +1,9 @@
 [build]
 builder = "DOCKERFILE"
 dockerfilePath = "Dockerfile"
+
+[deploy]
+healthcheckPath = "/health"
+healthcheckTimeout = 120
+restartPolicyType = "ON_FAILURE"
+restartPolicyMaxRetries = 3

requirements.txt

@@ -1,6 +1,6 @@
-fastapi
-uvicorn[standard]
-youtube-transcript-api
-playwright
-groq
-requests
+fastapi==0.135.1
+uvicorn[standard]==0.41.0
+youtube-transcript-api==1.2.4
+playwright==1.58.0
+groq==1.1.1
+requests==2.32.5

worker/index.js

@@ -25,8 +25,8 @@ export default {
       headers.delete('cf-connecting-ip');
       headers.delete('cf-ray');
       headers.delete('cf-ipcountry');
-      // Force uncompressed response so downstream clients can parse it
-      // (Worker strips Content-Encoding header, so compressed bytes become unparsable)
+      // Force uncompressed response — prevents encoding mismatch
+      // (CF Workers auto-decompress bodies, so forwarding Content-Encoding would lie about the actual encoding)
       headers.set('Accept-Encoding', 'identity');
       // Override User-Agent with a browser UA
       headers.set('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');
@@ -54,15 +54,15 @@ export default {
       const response = await fetch(targetUrl, fetchOptions);
       const body = await response.arrayBuffer();
 
-      // Forward essential response headers including Content-Encoding
+      // Build clean response headers
+      // CRITICAL: Do NOT forward Content-Encoding — CF Workers auto-decompress
+      // the body, so the raw bytes are always uncompressed regardless of what
+      // YouTube's Content-Encoding header says. Forwarding it would cause
+      // the downstream client to try decompressing already-decompressed data.
       const responseHeaders = {
         'Content-Type': response.headers.get('Content-Type') || 'text/html',
         'Access-Control-Allow-Origin': '*',
       };
-      const contentEncoding = response.headers.get('Content-Encoding');
-      if (contentEncoding) {
-        responseHeaders['Content-Encoding'] = contentEncoding;
-      }
 
       return new Response(body, {
         status: response.status,