tests: add final-status execution tests (done token and fail status)

.github/workflows/ci.yml

@@ -0,0 +1,36 @@
+name: CI
+
+on:
+  push:
+    branches: [main, master]
+  pull_request:
+    branches: [main, master]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v4
+        with:
+          python-version: "3.11"
+
+      - name: Install dev requirements
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements-dev.txt
+
+      - name: Run linters
+        env:
+          PYTHONPATH: ${{ github.workspace }}
+        run: |
+          python -m black --check .
+          python -m isort --check-only .
+          python -m flake8 .
+
+      - name: Run tests
+        env:
+          PYTHONPATH: ${{ github.workspace }}
+        run: |
+          python -m pytest -q

.gitignore

@@ -161,4 +161,7 @@ cython_debug/
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
 #.idea/
 logs/
-.DS_Store
+.DS_Store
+
+# Local env file for secrets
+.env

.patches/0002-ci-use-pre-commit-mirrors-for-flake8-and-add-basic-h.patch

@@ -0,0 +1,33 @@
+From bd11c7fff67d0667527f5480bc2a71e8fbd197f9 Mon Sep 17 00:00:00 2001
+From: Vasyl198 <ananiivasilii@gmail.com>
+Date: Wed, 24 Dec 2025 14:29:28 +0200
+Subject: [PATCH 2/2] ci: use pre-commit mirrors for flake8 and add basic hooks
+
+---
+ .pre-commit-config.yaml | 11 ++++++++---
+ 1 file changed, 8 insertions(+), 3 deletions(-)
+
+diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
+index 8665a49..7e0e7f6 100644
+--- a/.pre-commit-config.yaml
++++ b/.pre-commit-config.yaml
+@@ -8,8 +8,13 @@ repos:
+     rev: 5.12.0
+     hooks:
+       - id: isort
+-  - repo: https://gitlab.com/pycqa/flake8
+-    rev: 6.0.0
++  - repo: https://github.com/pre-commit/mirrors-flake8
++    rev: 7.1.0
+     hooks:
+       - id: flake8
+-        args: ["--max-line-length=120"]
++        args: ["--max-line-length=88", "--extend-ignore=E203,W503"]
++  - repo: https://github.com/pre-commit/pre-commit-hooks
++    rev: v4.6.0
++    hooks:
++      - id: end-of-file-fixer
++      - id: trailing-whitespace
+-- 
+2.49.0.windows.1
+

.pre-commit-config.yaml

@@ -0,0 +1,20 @@
+repos:
+  - repo: https://github.com/psf/black
+    rev: 25.11.0
+    hooks:
+      - id: black
+        language_version: python3.11
+  - repo: https://github.com/PyCQA/isort
+    rev: 5.12.0
+    hooks:
+      - id: isort
+  - repo: https://github.com/pre-commit/mirrors-flake8
+    rev: 7.1.0
+    hooks:
+      - id: flake8
+        args: ["--max-line-length=88", "--extend-ignore=E203,W503"]
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.6.0
+    hooks:
+      - id: end-of-file-fixer
+      - id: trailing-whitespace

gui_agents/s3/cli_app.py

@@ -145,14 +145,23 @@ def show_permission_dialog(code: str, action_description: str):
     return False
 
 
+def execute_code(code_str: str):
+    """Execute the provided code string in a controlled namespace.
+
+    This helper centralizes code execution so it can be mocked in tests.
+    """
+    # Execute in globals so that imports persist if needed by subsequent steps.
+    exec(code_str, globals())
+
+
 def scale_screen_dimensions(width: int, height: int, max_dim_size: int):
     scale_factor = min(max_dim_size / width, max_dim_size / height, 1)
     safe_width = int(width * scale_factor)
     safe_height = int(height * scale_factor)
     return safe_width, safe_height
 
 
-def run_agent(agent, instruction: str, scaled_width: int, scaled_height: int):
+def run_agent(agent, instruction: str, scaled_width: int, scaled_height: int, require_exec_confirmation: bool = True):
     global paused
     obs = {}
     traj = "Task:\n" + instruction
@@ -182,8 +191,42 @@ def run_agent(agent, instruction: str, scaled_width: int, scaled_height: int):
 
         # Get next action code from the agent
         info, code = agent.predict(instruction=instruction, observation=obs)
+        print('DEBUG: agent.predict returned code:', repr(code))
+
+        # Normalize to (code_str, status) form. Some agents return [code, status]
+        code_str = None
+        status = None
+        try:
+            if isinstance(code, (list, tuple)) and len(code) >= 1:
+                code_str = code[0]
+                if len(code) > 1:
+                    status = str(code[1]).lower().strip()
+            elif isinstance(code, str):
+                code_str = code
+        except Exception:
+            code_str = str(code)
+
+        # Interpret explicit status when present, or exact token in code_str
+        # Semantics: if code_str itself is a terminal token ("done"/"fail"), stop without executing.
+        # If an explicit status is provided (e.g., [code, "done"]), execute the provided code_str once, then stop.
+        if isinstance(code_str, str) and code_str.strip().lower() in ("done", "fail"):
+            break
+        if status in ("done", "fail"):
+            # Execute the final code, then stop
+            execute_final = code_str
+            # fall through to execution branch below
+            final_exit_after_exec = True
+        else:
+            execute_final = None
+            final_exit_after_exec = False
 
-        if "done" in code[0].lower() or "fail" in code[0].lower():
+        # If execute_final is set, we want to execute it once and then exit the loop
+        if execute_final is not None:
+            code_to_run = execute_final
+            do_exit_after = True
+        else:
+            code_to_run = code_str
+            do_exit_after = False
             if platform.system() == "Darwin":
                 os.system(
                     f'osascript -e \'display dialog "Task Completed" with title "OpenACI Agent" buttons "OK" default button "OK"\''
@@ -205,14 +248,36 @@ def run_agent(agent, instruction: str, scaled_width: int, scaled_height: int):
 
         else:
             time.sleep(1.0)
-            print("EXECUTING CODE:", code[0])
+            print("EXECUTING CODE:", code_to_run)
 
             # Check for pause state before execution
             while paused:
                 time.sleep(0.1)
 
             # Ask for permission before executing
-            exec(code[0])
+            allowed = True
+            if require_exec_confirmation:
+                # Try platform GUI confirmation first
+                try:
+                    allowed = show_permission_dialog(code_to_run, "execute this action")
+                except Exception:
+                    allowed = False
+
+                if not allowed:
+                    try:
+                        print("Agent proposes to execute the following code:\n")
+                        print(code_to_run)
+                        resp = input("Execute this code? (y/N): ")
+                        if resp.lower().strip() == "y":
+                            allowed = True
+                    except Exception:
+                        allowed = False
+
+            if not allowed:
+                print("Execution denied by user; skipping this action.")
+            else:
+                execute_code(code_to_run)
+
             time.sleep(1.0)
 
             # Update task and subtask trajectories
@@ -224,6 +289,10 @@ def run_agent(agent, instruction: str, scaled_width: int, scaled_height: int):
                     + info["executor_plan"]
                 )
 
+            # If this was a final command (status indicated done/fail), exit after executing
+            if do_exit_after:
+                break
+
 
 def main():
     parser = argparse.ArgumentParser(description="Run AgentS3 with specified model.")
@@ -316,6 +385,13 @@ def main():
         help="Enable local coding environment for code execution (WARNING: Executes arbitrary code locally)",
     )
 
+    parser.add_argument(
+        "--allow_exec",
+        action="store_true",
+        default=False,
+        help="Allow executing agent-generated code without confirmation (dangerous).",
+    )
+
     args = parser.parse_args()
 
     # Re-scales screenshot size to ensure it fits in UI-TARS context limit
@@ -368,13 +444,16 @@ def main():
         enable_reflection=args.enable_reflection,
     )
 
+    # Determine whether user approval is required before executing agent code
+    require_exec_confirmation = not getattr(args, "allow_exec", False)
+
     while True:
         query = input("Query: ")
 
         agent.reset()
 
         # Run the agent on your own device
-        run_agent(agent, query, scaled_width, scaled_height)
+        run_agent(agent, query, scaled_width, scaled_height, require_exec_confirmation=require_exec_confirmation)
 
         response = input("Would you like to provide another query? (y/n): ")
         if response.lower() != "y":

requirements-dev.txt

@@ -0,0 +1,6 @@
+pytest
+pillow
+black
+flake8
+isort
+pre-commit

results/restaurant/index.html

@@ -0,0 +1,128 @@
+<!doctype html>
+<html lang="ru">
+<head>
+  <meta charset="utf-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1">
+  <title>Ресторан "У моря"</title>
+  <style>
+    :root { --accent: #ff6b6b; --muted: rgba(255,255,255,0.85) }
+    html,body { height:100%; margin:0 }
+    body {
+      font-family: Arial, Helvetica, sans-serif;
+      color: var(--muted);
+      background: url('https://images.unsplash.com/photo-1507525428034-b723cf961d3?w=1600&q=80&auto=format&fit=crop') center/cover fixed no-repeat;
+      -webkit-font-smoothing:antialiased;
+      -moz-osx-font-smoothing:grayscale;
+    }
+    .overlay { background: rgba(0,0,0,0.38); min-height:100%; }
+    header { display:flex; align-items:center; justify-content:space-between; padding:1rem 2rem; }
+    header h1{ margin:0; font-size:1.6rem }
+    nav a{ color:var(--muted); margin-left:1rem; text-decoration:none; font-weight:600 }
+
+    .hero { display:flex; gap:2rem; align-items:center; padding:6rem 2rem; max-width:1200px; margin:0 auto }
+    .hero .intro{ max-width:640px }
+    .hero h2{ font-size:2.4rem; margin:0 0 .5rem; text-shadow:0 6px 24px rgba(0,0,0,0.6) }
+    .hero p{ margin:.5rem 0 1rem; line-height:1.45 }
+    .btn{ display:inline-block; background:var(--accent); color:#fff; padding:.75rem 1.1rem; border-radius:999px; text-decoration:none; font-weight:700 }
+
+    .section{ padding:3rem 2rem; max-width:1100px; margin:0 auto }
+    .menu{ display:flex; gap:1rem; flex-wrap:wrap }
+    .menu-item{ background:rgba(255,255,255,0.04); padding:1rem; border-radius:10px; flex:1 1 240px; transition:transform .18s ease, box-shadow .18s ease }
+    .menu-item:hover{ transform:translateY(-8px) scale(1.02); box-shadow:0 12px 30px rgba(0,0,0,0.5) }
+    .menu-item h4{ margin:.2rem 0 }
+
+    footer{ text-align:center; padding:2rem; color:rgba(255,255,255,0.8) }
+
+    /* small screens */
+    @media (max-width:700px){
+      .hero{ padding:3rem 1rem; flex-direction:column; text-align:center }
+      nav{ display:none }
+    }
+
+    /* Reserve modal */
+    .modal{ position:fixed; left:0; top:0; right:0; bottom:0; display:flex; align-items:center; justify-content:center; background:rgba(0,0,0,0.6); opacity:0; pointer-events:none; transition:opacity .2s }
+    .modal.open{ opacity:1; pointer-events:auto }
+    .modal .card{ background:#fff; color:#222; padding:1.6rem; border-radius:8px; min-width:280px; max-width:420px }
+    .close{ background:#eee;border-radius:6px;padding:.4rem .6rem;cursor:pointer }
+  </style>
+</head>
+<body>
+  <div class="overlay">
+    <header>
+      <h1>Ресторан "У моря"</h1>
+      <nav>
+        <a href="#menu">Меню</a>
+        <a href="#contacts">Контакты</a>
+        <a href="#reserve" id="reserveBtn" class="btn">Забронировать</a>
+      </nav>
+    </header>
+
+    <main>
+      <section class="hero">
+        <div class="intro">
+          <h2>Свежая еда, уютная атмосфера и вид на море</h2>
+          <p>Насладитесь авторскими блюдами из местных ингредиентов, приготовленными с любовью нашим шеф-поваром.</p>
+          <p><strong>Часы работы:</strong> 10:00 — 23:00</p>
+          <p>Текущее время: <span id="now"></span></p>
+          <a href="#menu" class="btn">Посмотреть меню</a>
+        </div>
+        <div class="visual" aria-hidden="true">
+          <img src="https://images.unsplash.com/photo-1498654896293-37aacf113fd9?w=800&q=80&auto=format&fit=crop" alt="restaurant" style="width:320px;border-radius:8px;box-shadow:0 10px 30px rgba(0,0,0,0.6)">
+        </div>
+      </section>
+
+      <section id="menu" class="section">
+        <h3>Меню</h3>
+        <div class="menu">
+          <div class="menu-item"><h4>Салат из лосося</h4><p>Свежий лосось, микс салатов, цитрусовая заправка — 320 ₴</p></div>
+          <div class="menu-item"><h4>Паста морская</h4><p>Паста с морепродуктами в сливочном соусе — 380 ₴</p></div>
+          <div class="menu-item"><h4>Стейк</h4><p>Говяжий стейк с овощами — 420 ₴</p></div>
+          <div class="menu-item"><h4>Десерт дня</h4><p>Творожный чизкейк с ягодным соусом — 150 ₴</p></div>
+        </div>
+      </section>
+
+      <section id="contacts" class="section">
+        <h3>Контакты</h3>
+        <p>Адрес: Одесса, набережная, 1</p>
+        <p>Телефон: <a href="tel:+380000000000" style="color:var(--muted);text-decoration:underline">+38 0XX XXX XX XX</a></p>
+      </section>
+    </main>
+
+    <footer>
+      © Ресторан "У моря" — Приятного аппетита!
+    </footer>
+
+    <div id="reserveModal" class="modal" role="dialog" aria-hidden="true">
+      <div class="card">
+        <h4>Бронирование</h4>
+        <p>Пожалуйста, оставьте номер — мы свяжемся с вами для подтверждения.</p>
+        <div style="display:flex;gap:.5rem;margin-top:.6rem">
+          <input id="phone" placeholder="Ваш телефон" style="flex:1;padding:.5rem;border:1px solid #ddd;border-radius:6px">
+          <button id="sendReserve" class="btn">Отправить</button>
+        </div>
+        <div style="margin-top:.6rem;text-align:right"><button id="closeModal" class="close">Закрыть</button></div>
+      </div>
+    </div>
+  </div>
+
+  <script>
+    // Live clock
+    document.addEventListener('DOMContentLoaded', function(){
+      var nowEl = document.getElementById('now');
+      function tick(){ nowEl.textContent = new Date().toLocaleTimeString(); }
+      tick(); setInterval(tick,1000);
+
+      var reserveBtn = document.getElementById('reserveBtn');
+      var modal = document.getElementById('reserveModal');
+      var close = document.getElementById('closeModal');
+      var send = document.getElementById('sendReserve');
+      reserveBtn && reserveBtn.addEventListener('click', function(e){ e.preventDefault(); modal.classList.add('open'); modal.setAttribute('aria-hidden','false'); });
+      close && close.addEventListener('click', function(){ modal.classList.remove('open'); modal.setAttribute('aria-hidden','true'); });
+      send && send.addEventListener('click', function(){ alert('Спасибо! Мы свяжемся с вами для подтверждения брони.'); modal.classList.remove('open'); modal.setAttribute('aria-hidden','true'); });
+
+      // Smooth scroll for anchor links
+      document.querySelectorAll('a[href^="#"]').forEach(function(a){ a.addEventListener('click', function(e){ var target = document.querySelector(this.getAttribute('href')); if(target){ e.preventDefault(); target.scrollIntoView({behavior:'smooth'}); } }); });
+    });
+  </script>
+</body>
+</html>