feat: add noVNC headed mode for interactive browser sessions

- Dockerfile with Xvfb + x11vnc + noVNC for browser-in-a-URL
- search.py auto-detects DISPLAY env to run headed vs headless
- pause_for_human() function for CAPTCHA/login handoff
- Facebook Marketplace search (headed mode only, needs login)
- Updated README with both modes

search.py

@@ -1,10 +1,20 @@
 """
 Car search crawler -- finds used hybrids under $10K near Woodinville WA.
 Uses Playwright (headless Chromium) to handle JS-heavy car sites.
+
+Two modes:
+  headless (default):  uv run python search.py
+  headed via noVNC:    docker run ... (see README)
+
+When running in the Docker container, the browser is visible at
+http://localhost:6080/vnc.html -- use this to handle CAPTCHAs,
+log into Facebook Marketplace, or visually inspect results.
 """
 
 import asyncio
 import json
+import os
+import sys
 from datetime import datetime, timezone
 from pathlib import Path
 
@@ -13,6 +23,26 @@ from playwright.async_api import async_playwright
 RESULTS_DIR = Path("results")
 RESULTS_DIR.mkdir(exist_ok=True)
 
+# If DISPLAY is set (noVNC container), run headed
+HEADED = bool(os.environ.get("DISPLAY"))
+
+
+async def pause_for_human(page, reason: str):
+    """Pause and prompt for human interaction via noVNC."""
+    if not HEADED:
+        print(f"  SKIP (headless): {reason}")
+        return False
+
+    print(f"\n{'=' * 60}")
+    print(f"  HUMAN NEEDED: {reason}")
+    print(f"  Open http://localhost:6080/vnc.html to interact")
+    print(f"  Press Enter here when done...")
+    print(f"{'=' * 60}")
+
+    # Wait for user to press Enter in the terminal
+    await asyncio.get_event_loop().run_in_executor(None, input)
+    return True
+
 
 async def search_craigslist(page) -> list[dict]:
     """Craigslist Seattle -- simple HTML, most reliable to scrape."""
@@ -24,7 +54,7 @@ async def search_craigslist(page) -> list[dict]:
         "&search_distance=50"
         "&sort=date"
     )
-    print(f"  Searching Craigslist: {url}")
+    print(f"  Searching Craigslist...")
     await page.goto(url, wait_until="domcontentloaded", timeout=30000)
     await page.wait_for_timeout(2000)
 
@@ -96,6 +126,44 @@ async def search_cargurus(page) -> list[dict]:
     return listings
 
 
+async def search_facebook(page) -> list[dict]:
+    """Facebook Marketplace -- needs login (human via noVNC)."""
+    if not HEADED:
+        print("  SKIP Facebook Marketplace (requires login -- run in Docker with noVNC)")
+        return []
+
+    url = "https://www.facebook.com/marketplace/seattle/vehicles/?maxPrice=10000"
+    print(f"  Opening Facebook Marketplace...")
+    await page.goto(url, wait_until="networkidle", timeout=60000)
+
+    # Check if login is needed
+    if "login" in page.url.lower():
+        await pause_for_human(page, "Log into Facebook, then press Enter")
+
+    await page.wait_for_timeout(3000)
+    await pause_for_human(page, "Apply filters (hybrid, $10K max, 50mi) and press Enter when ready")
+
+    # Scrape whatever is visible
+    listings = []
+    cards = await page.query_selector_all("[class*='marketplace']")
+    for card in cards[:30]:
+        try:
+            text = await card.inner_text()
+            link_el = await card.query_selector("a[href*='/marketplace/']")
+            link = await link_el.get_attribute("href") if link_el else ""
+            listings.append({
+                "source": "facebook",
+                "title": text[:100].strip(),
+                "price": "see listing",
+                "url": f"https://www.facebook.com{link}" if link else "",
+            })
+        except Exception:
+            continue
+
+    print(f"  Found {len(listings)} Facebook listings")
+    return listings
+
+
 async def search_autotempest(page) -> list[dict]:
     """AutoTempest -- meta-aggregator, searches multiple sites."""
     url = (
@@ -132,7 +200,8 @@ async def search_autotempest(page) -> list[dict]:
 
 
 async def main():
-    print(f"Car Search -- {datetime.now(timezone.utc).strftime('%Y-%m-%d %H:%M UTC')}")
+    mode = "HEADED (noVNC)" if HEADED else "HEADLESS"
+    print(f"Car Search [{mode}] -- {datetime.now(timezone.utc).strftime('%Y-%m-%d %H:%M UTC')}")
     print(
         "Looking for: Used hybrids under $10K "
         "within 50 miles of 98077 (Woodinville WA)"
@@ -142,7 +211,7 @@ async def main():
     all_listings: list[dict] = []
 
     async with async_playwright() as p:
-        browser = await p.chromium.launch(headless=True)
+        browser = await p.chromium.launch(headless=not HEADED)
         context = await browser.new_context(
             user_agent=(
                 "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 "
@@ -152,12 +221,16 @@ async def main():
         )
         page = await context.new_page()
 
-        for search_fn in [search_craigslist, search_cargurus, search_autotempest]:
+        searches = [search_craigslist, search_cargurus, search_autotempest]
+        if HEADED:
+            searches.append(search_facebook)
+
+        for search_fn in searches:
             try:
                 results = await search_fn(page)
                 all_listings.extend(results)
             except Exception as e:
-                print(f"  ERROR: {e}")
+                print(f"  ERROR in {search_fn.__name__}: {e}")
 
         await browser.close()
 
@@ -168,6 +241,7 @@ async def main():
         json.dump(
             {
                 "search_date": datetime.now(timezone.utc).isoformat(),
+                "mode": mode,
                 "params": {
                     "zip": "98077",
                     "max_price": 10000,