add schemas and swap browser_use_runner in docker with browser and now agent use skill from the other container

browser_env/Dockerfile.browser

@@ -4,6 +4,8 @@ ENV DEBIAN_FRONTEND=noninteractive
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
     chromium \
+    python3 \
+    python3-pip \
     xvfb \
     fluxbox \
     x11vnc \
@@ -19,9 +21,14 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 WORKDIR /src
 RUN mkdir -p /src/browser_data
 
+RUN python3 -m pip install --no-cache-dir --break-system-packages \
+    "browser-use>=0.12.5" \
+    "langchain-openai>=0.3.0"
+
 COPY entrypoint.sh /entrypoint.sh
+COPY browser_use_runner.py /src/browser_use_runner.py
 RUN chmod +x /entrypoint.sh
 
-EXPOSE 6080 9222
+EXPOSE 6080 9222 8787
 
 ENTRYPOINT ["/entrypoint.sh"]

browser_env/browser_use_runner.py

@@ -0,0 +1,102 @@
+import asyncio
+import json
+import os
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from urllib import error, request
+
+from browser_use import Agent, Browser, ChatOpenAI
+
+
+def _json_response(handler, status_code, payload):
+    data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+    handler.send_response(status_code)
+    handler.send_header("Content-Type", "application/json; charset=utf-8")
+    handler.send_header("Content-Length", str(len(data)))
+    handler.end_headers()
+    handler.wfile.write(data)
+
+
+async def run_browser_task(task):
+    cdp_url = os.getenv("BROWSER_CDP_URL", "http://127.0.0.1:9222")
+    browser_view_url = os.getenv("BROWSER_VIEW_URL", "")
+
+    browser = Browser(cdp_url=cdp_url)
+
+    llm = ChatOpenAI(
+        model=os.getenv("MODEL_DEFAULT", "qwen3.5-122b"),
+        api_key=os.getenv("OPENAI_API_KEY"),
+        base_url=os.getenv("OPENAI_BASE_URL"),
+        temperature=0.0,
+    )
+
+    agent = Agent(task=task, llm=llm, browser=browser, use_vision=False)
+
+    try:
+        history = await agent.run()
+        return {
+            "success": True,
+            "result": history.final_result(),
+            "browser_view": browser_view_url,
+        }
+    except Exception as err:
+        return {"success": False, "error": f"Browser automation failed: {err}"}
+    finally:
+        try:
+            await browser.close()
+        except Exception:
+            pass
+
+
+class BrowserUseRPCHandler(BaseHTTPRequestHandler):
+    def do_GET(self):
+        if self.path != "/health":
+            _json_response(self, 404, {"success": False, "error": "Not found"})
+            return
+
+        try:
+            debug_url = os.getenv("BROWSER_HEALTH_URL", "http://127.0.0.1:9222/json/version")
+            with request.urlopen(debug_url, timeout=2):
+                pass
+            _json_response(self, 200, {"success": True})
+        except Exception as err:
+            _json_response(self, 503, {"success": False, "error": f"Browser is not ready: {err}"})
+
+    def do_POST(self):
+        if self.path != "/run":
+            _json_response(self, 404, {"success": False, "error": "Not found"})
+            return
+
+        try:
+            content_length = int(self.headers.get("Content-Length", "0"))
+            raw = self.rfile.read(content_length)
+            payload = json.loads(raw.decode("utf-8") if raw else "{}")
+            task = payload.get("task", "")
+            if not isinstance(task, str) or not task.strip():
+                _json_response(self, 400, {"success": False, "error": "Field 'task' is required"})
+                return
+
+            result = asyncio.run(run_browser_task(task.strip()))
+            code = 200 if result.get("success") else 500
+            _json_response(self, code, result)
+        except json.JSONDecodeError:
+            _json_response(self, 400, {"success": False, "error": "Invalid JSON payload"})
+        except error.URLError as err:
+            _json_response(self, 503, {"success": False, "error": f"Transport error: {err}"})
+        except Exception as err:
+            _json_response(self, 500, {"success": False, "error": f"Internal error: {err}"})
+
+    def log_message(self, format_str, *args):
+        return
+
+
+def main():
+    host = os.getenv("BROWSER_USE_RPC_HOST", "0.0.0.0")
+    port = int(os.getenv("BROWSER_USE_RPC_PORT", "8787"))
+    server = ThreadingHTTPServer((host, port), BrowserUseRPCHandler)
+    print(f"browser-use RPC listening on {host}:{port}")
+    server.serve_forever()
+
+
+if __name__ == "__main__":
+    main()
+

browser_env/entrypoint.sh

@@ -9,6 +9,7 @@ VNC_PORT="${VNC_PORT:-5900}"
 NOVNC_PORT="${NOVNC_PORT:-6080}"
 CHROME_LOCAL_DEBUG_PORT="${CHROME_LOCAL_DEBUG_PORT:-9223}"
 CHROME_PUBLIC_DEBUG_PORT="${CHROME_PUBLIC_DEBUG_PORT:-9222}"
+BROWSER_USE_RPC_PORT="${BROWSER_USE_RPC_PORT:-8787}"
 CHROME_PROFILE_DIR="${CHROME_PROFILE_DIR:-/src/browser_data}"
 
 MAX_RESTARTS="${MAX_RESTARTS:-10}"
@@ -119,6 +120,7 @@ start_bg fluxbox
 start_bg x11vnc -display "$DISPLAY" -rfbport "$VNC_PORT" -nopw -listen 0.0.0.0 -xkb -forever -shared
 start_bg websockify --web=/usr/share/novnc/ "$NOVNC_PORT" "localhost:${VNC_PORT}"
 start_bg socat "TCP-LISTEN:${CHROME_PUBLIC_DEBUG_PORT},fork,reuseaddr" "TCP:127.0.0.1:${CHROME_LOCAL_DEBUG_PORT}"
+start_bg python3 -u /src/browser_use_runner.py
 
 if ! wait_for_port 127.0.0.1 "$VNC_PORT" 20; then
   log "fatal: x11vnc did not open port ${VNC_PORT}"
@@ -128,8 +130,12 @@ if ! wait_for_port 127.0.0.1 "$NOVNC_PORT" 20; then
   log "fatal: websockify did not open port ${NOVNC_PORT}"
   exit 1
 fi
+if ! wait_for_port 127.0.0.1 "$BROWSER_USE_RPC_PORT" 20; then
+  log "fatal: browser-use RPC did not open port ${BROWSER_USE_RPC_PORT}"
+  exit 1
+fi
 
-log "browser infrastructure is ready (noVNC:${NOVNC_PORT}, DevTools proxy:${CHROME_PUBLIC_DEBUG_PORT})"
+log "browser infrastructure is ready (noVNC:${NOVNC_PORT}, DevTools proxy:${CHROME_PUBLIC_DEBUG_PORT}, browser-use RPC:${BROWSER_USE_RPC_PORT})"
 
 while true; do
   for pid in "${PIDS[@]}"; do