DevFlow AI

Inspiration

We wanted to build an AI that goes beyond generating responses. Most tools stop at giving code or suggestions, but I thought: what if AI could actually execute workflows? That idea led to DevFlow AI — an agent that plans, acts, and completes tasks step-by-step.

What it does

DevFlow AI is an agentic AI system that converts natural language into real workflows. Users can describe tasks like cleaning data, generating emails, or automating processes, and the system executes them step-by-step with live feedback.

How we built it

We built DevFlow AI as a lightweight web app using HTML, CSS, and JavaScript. LLaMA 3.3 via Groq API is used for understanding user intent and generating structured workflows. The frontend interprets these workflows and executes them interactively with real-time updates.

Challenges we ran into

• Making the system truly “agentic” instead of just a chatbot
• Handling structured AI outputs (JSON parsing issues)
• Designing a smooth user experience for step-by-step execution
• Managing API limits and errors gracefully

Accomplishments that we're proud of

• Built a working agentic AI system fully in the browser
• Enabled real-time workflow execution visualization
• Integrated AI with practical use cases like data cleaning and email automation
• Designed a clean and intuitive UI for complex workflows

What we learned

Agentic AI requires both strong logic and good UI design and structured outputs from LLMs are critical for execution.Frontend-only systems can still demonstrate powerful AI workflows. How to integrate API key and handle the key securely.

What's next for DevFlow AI

• Add backend (AWS Lambda) for secure API handling
• Integrate real services (Gmail, GitHub, Slack)
• Enable multi-step automation pipelines Add workflow saving and scheduling features

Built With

• amazon-web-services
• async/await
• css
• flexbox
• git
• github
• grid
• html
• javascript
• llama3.3
• s3
• vscode