Demo Debugging with AI

In this lesson, you’ll learn how to leverage AI (via Composer) to debug a Flask application. We’ll capture runtime errors in the browser or terminal, feed them into Composer alongside the relevant code, and apply the suggested fixes—all while applying your own programmer intuition.

1. Reproducing & Resolving a Jinja2 Template Error

1. Start your Flask development server and navigate to the login page.
2. Log in with valid credentials.

3. After logging in, you may encounter:

jinja2.exceptions.TemplateNotFound: dash.html
127.0.0.1 - - [20/Mar/2025 21:37:51] "GET /dashboard HTTP/1.1" 500 -

The Faulty Route

In app.py, the dashboard route is defined as:

@app.route('/dashboard')
@login_required
def dashboard():
    db = get_db()
    # ... build query ...
    return render_template('dash.html', tasks=tasks, users=users, statuses=statuses)

Feeding the Error and Code to Composer

jinja2.exceptions.TemplateNotFound: dash.html

# app.py (excerpt)
@app.route('/dashboard')
@login_required
def dashboard():
    db = get_db()
    # Build the SELECT query...
    return render_template('dash.html', tasks=tasks, users=users, statuses=statuses)

Ask Composer:

“Help me debug this error.”

Composer identifies:

Issue: You reference dash.html, but your templates directory contains dashboard.html.
Solution: Rename the template or update the call to render_template('dashboard.html', …).

Applying the Fix

-    return render_template('dash.html', tasks=tasks, users=users, statuses=statuses)
+    return render_template('dashboard.html', tasks=tasks, users=users, statuses=statuses)

Reload the page—the dashboard now renders correctly.

2. Debugging a Database Column Typo

Next, attempt to create a new task via the web interface:

Submitting the form triggers:

sqlite3.OperationalError: table tasks has no column named statu3

Traceback (most recent call last):
  File ".../flask/app.py", line 1478, in __call__
    response = self.handle_exception(e)
  ...
  File "app.py", line 193, in create_task
    db.execute(...)

The Faulty INSERT Statement

In app.py, the handler is:

@app.route('/task/create', methods=['GET', 'POST'])
@login_required
def create_task():
    if request.method == 'POST':
        title = request.form['title']
        description = request.form['description']
        status = request.form['status']
        assigned_to = request.form['assignee']

        db.execute(
            'INSERT INTO tasks (title, description, statu3, created_by, assigned_to) '
            'VALUES (?, ?, ?, ?, ?)',
            (title, description, status, session['user_id'], assigned_to)
        )
        db.commit()
        return redirect(url_for('dashboard'))
    return render_template('create_task.html')

Feeding Code and Error to Composer

sqlite3.OperationalError: table tasks has no column named statu3

db.execute(
    'INSERT INTO tasks (title, description, statu3, created_by, assigned_to) VALUES (?, ?, ?, ?, ?)',
    (title, description, status, session['user_id'], assigned_to)
)

Composer responds:

Cause: The column name statu3 is mistyped. It should be status.
Fix: Correct the column name in the SQL statement.

Applying the Fix

-    'INSERT INTO tasks (title, description, statu3, created_by, assigned_to) VALUES (?, ?, ?, ?, ?)',
+    'INSERT INTO tasks (title, description, status, created_by, assigned_to) VALUES (?, ?, ?, ?, ?)',

Save and refresh—the new task is created successfully.

3. Best Practices for AI-Assisted Debugging

1. Capture Complete Context

   • Include full tracebacks or error messages.
   • Paste only the relevant code where the error occurred.

2. Provide Schema or Definitions

   • Share your schema.sql or data models so the AI can verify table and column names.

3. Trust but Verify

   • Apply your own programming knowledge to confirm suggested changes align with your codebase.

4. Experiment with Multiple Models

   • If Composer’s recommendation isn’t clear, try another model (e.g., GPT-3.5, GPT-4, Gemini) or start a fresh session.

5. Use Logs as Supplemental Context

   • Include application or server logs (from SSH sessions or production environments) for deeper insights.

Links and References

• Flask Documentation
• Jinja2 Error Handling
• SQLite Documentation
• Composer AI Assistant Guide