A0_browser_control w/ embedded noVNC browser

DockerfileLocal

@@ -33,4 +33,4 @@ EXPOSE 22 80 9000-9009
 RUN chmod +x /exe/initialize.sh /exe/run_A0.sh /exe/run_searxng.sh /exe/run_tunnel_api.sh
 
 # initialize runtime and switch to supervisord
-CMD ["/exe/initialize.sh", "$BRANCH"]
+CMD ["/exe/initialize.sh", "$BRANCH"]

agent.py

@@ -275,7 +275,6 @@ class AgentConfig:
     chat_model: models.ModelConfig
     utility_model: models.ModelConfig
     embeddings_model: models.ModelConfig
-    browser_model: models.ModelConfig
     mcp_servers: str
     profile: str = ""
     memory_subdir: str = ""
@@ -287,7 +286,12 @@ class AgentConfig:
     code_exec_ssh_user: str = "root"
     code_exec_ssh_pass: str = ""
     additional: Dict[str, Any] = field(default_factory=dict)
-
+    browser_control_headless: bool = False  # Browser GUI enabled for interaction (uses VNC if available, otherwise X11 forwarding)
+    browser_control_cdp_url: str = ""  # Chrome DevTools Protocol URL for native browser (e.g., "ws://host.docker.internal:9222/devtools/browser/..."), leave empty to use embedded browser with VNC
+    browser_control_start_url: str = "https://www.google.com"
+    browser_control_timeout: int = 5000 # milliseconds
+    # VNC is automatically enabled if available (configured in docker-compose.yml)
+    # Access browser control via noVNC when agent calls pause_for_user method
 
 @dataclass
 class UserMessage:
@@ -676,14 +680,6 @@ def get_utility_model(self):
             **self.config.utility_model.build_kwargs(),
         )
 
-    def get_browser_model(self):
-        return models.get_browser_model(
-            self.config.browser_model.provider,
-            self.config.browser_model.name,
-            model_config=self.config.browser_model,
-            **self.config.browser_model.build_kwargs(),
-        )
-
     def get_embedding_model(self):
         return models.get_embedding_model(
             self.config.embeddings_model.provider,

docker/base/Dockerfile

@@ -27,6 +27,12 @@ RUN bash /ins/install_base_packages4.sh
 # install python after packages to ensure version overriding
 RUN bash /ins/install_python.sh
 
+# install X11 support for browser display
+RUN bash /ins/install_x11_support.sh
+
+# install VNC server and noVNC for remote browser control
+RUN bash /ins/install_vnc.sh
+
 # install searxng
 RUN bash /ins/install_searxng.sh
 

docker/run/docker-compose.yml

@@ -1,8 +1,53 @@
 services:
   agent-zero:
     container_name: agent-zero
-    image: agent0ai/agent-zero:latest
+    # Use local development image (build with: docker build -f DockerfileLocal -t agent-zero-local --build-arg CACHE_DATE=$(date +%Y-%m-%d:%H:%M:%S) .)
+    image: agent-zero-local
+    # Use Docker Hub image for production deployments
+    # image: agent0ai/agent-zero:latest
     volumes:
-      - ./agent-zero:/a0
+      # Mount the actual project root (not the outdated copy in ./agent-zero)
+      # This allows live development - changes reflected immediately without rebuild
+      - ../..:/a0
+      # X11 socket for GUI display on macOS (auto-configured)
+      - /tmp/.X11-unix:/tmp/.X11-unix:rw
     ports:
-      - "50080:80"
+      - "55022:22"
+      - "50080:80"
+      - "56080:6080"  # noVNC web client for browser control
+      - "50090:9000"
+      - "50091:9001"
+      - "50092:9002"
+      - "50093:9003"
+      - "50094:9004"
+      - "50095:9005"
+      - "50096:9006"
+      - "50097:9007"
+      - "50098:9008"
+      - "50099:9009"
+    environment:
+      # X11 display forwarding via TCP (Docker Desktop on macOS uses VM)
+      - DISPLAY=host.docker.internal:0
+      - XAUTHORITY=/tmp/.Xauthority
+      # VNC configuration for remote browser control
+      - VNC_DISPLAY=:99
+      - VNC_RESOLUTION=1920x1080x24
+      - VNC_PORT=5900
+      - NOVNC_PORT=6080
+      - NOVNC_EXTERNAL_PORT=56080  # External port mapping for noVNC access
+      - VNC_PASSWORD=agent-zero
+    # Allow container to reach host for X11
+    extra_hosts:
+      - "host.docker.internal:host-gateway"
+    # Security options for X11
+    security_opt:
+      - seccomp:unconfined
+    # Shared memory for browser (required for Chromium)
+    shm_size: '2gb'
+    # Auto-check and setup display and VNC on startup
+    command: >
+      bash -c "
+        /exe/check_display.sh || true &&
+        /exe/start_vnc.sh || true &&
+        /exe/initialize.sh development
+      "

docker/run/fs/etc/supervisor/conf.d/vnc.conf

@@ -0,0 +1,13 @@
+[program:run_vnc]
+command=/exe/start_vnc.sh
+environment=
+user=root
+stopwaitsecs=10
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+autorestart=true
+startretries=3
+stopasgroup=true
+killasgroup=true

docker/run/fs/exe/check_display.sh

@@ -0,0 +1,82 @@
+#!/bin/bash
+# Automatic X11 display setup checker
+# Runs on container startup to verify display forwarding
+# No user interaction required - fully automatic
+
+set -e
+
+echo "========================================"
+echo "Agent Zero - Display Setup Check"
+echo "========================================"
+
+# Detect if running on macOS host
+IS_MACOS=false
+if [ -f /tmp/.X11-unix ] || [ "$DISPLAY" = "host.docker.internal:0" ]; then
+    IS_MACOS=true
+fi
+
+# Check if DISPLAY is set
+if [ -z "$DISPLAY" ]; then
+    echo "⚠️  No display configured (headless mode)"
+    echo "   Browser will run in headless mode (invisible)"
+    echo ""
+    echo "To enable visible browser on macOS:"
+    echo "  1. Install XQuartz: https://www.xquartz.org/"
+    echo "  2. Start XQuartz and restart Agent Zero"
+    exit 0
+fi
+
+# Display is configured - verify X11 libraries
+echo "✓ Display configured: $DISPLAY"
+
+# Check if X11 libraries are installed
+if ! dpkg -l | grep -q libx11-6; then
+    echo "Installing X11 libraries for browser display..."
+    apt-get update -qq
+    DEBIAN_FRONTEND=noninteractive apt-get install -y -qq \
+        libx11-6 libxcb1 libxcomposite1 libxcursor1 libxdamage1 \
+        libxext6 libxfixes3 libxi6 libxrandr2 libxrender1 libxss1 \
+        libxtst6 libgbm1 libasound2 libatk1.0-0 libatk-bridge2.0-0 \
+        libcups2 libdrm2 libgtk-3-0 libnspr4 libnss3 \
+        2>&1 | grep -v "^Reading" | grep -v "^Building" || true
+fi
+
+echo "✓ X11 libraries installed"
+
+# Test X11 connection
+if [ "$IS_MACOS" = true ]; then
+    echo "Testing X11 connection to macOS host..."
+
+    # Try to connect to X11
+    timeout 2 xdpyinfo -display "$DISPLAY" > /dev/null 2>&1 && {
+        echo "✓ X11 connection successful"
+        echo "✓ Browser will appear on your screen"
+        exit 0
+    } || {
+        echo ""
+        echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+        echo "⚠️  Cannot connect to X11 display"
+        echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+        echo ""
+        echo "To see the browser window, you need XQuartz:"
+        echo ""
+        echo "  1. Download and install XQuartz:"
+        echo "     https://www.xquartz.org/"
+        echo ""
+        echo "  2. Log out and log back in (required!)"
+        echo ""
+        echo "  3. Allow Docker connections:"
+        echo "     xhost +localhost"
+        echo ""
+        echo "  4. Restart Agent Zero:"
+        echo "     cd docker/run && docker-compose restart"
+        echo ""
+        echo "For now, browser will run in headless mode."
+        echo "━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━"
+        echo ""
+        exit 0
+    }
+fi
+
+echo "✓ Display setup complete"
+echo "========================================"

docker/run/fs/exe/initialize.sh

@@ -19,5 +19,8 @@ chmod 444 /root/.profile
 # update package list to save time later
 apt-get update > /dev/null 2>&1 &
 
+# Start VNC server in the background (for browser control feature)
+/exe/start_vnc.sh > /tmp/vnc_startup.log 2>&1 &
+
 # let supervisord handle the services
 exec /usr/bin/supervisord -c /etc/supervisor/conf.d/supervisord.conf