april 2024GitHub

AIr Chat – AI-Enhanced Conversational Assistant

Demonstration: https://youtu.be/ydZt61haShA

🚀 Inspiration

During project collaborations, live discussions with friends often turned chaotic—juggling between browser tabs, chat apps, technical issues, and solutions. We wondered:

How can we streamline access to intelligent answers without switching between applications?

Ever been in a critical conversation and wished for a smoother exchange? Or found yourself needing to reply on the go, without complete knowledge?

What if a system could guide us with intelligent, context-aware responses in real-time?

Finally, maintaining engaging conversations online isn't always easy. We asked:

How can we keep interactions lively, responsive, and human—even with limited input?

These questions shaped the foundation of AIr.

💡 What It Does

AIr is a chat application augmented with generative AI to enhance and streamline online conversations. By analyzing the ongoing dialogue, it suggests context-aware replies—ensuring seamless communication.

✨ Key Features

  • AI Response Suggestions
    Get 4 intelligent response suggestions for any message with a single click.

  • AI Assistant Chat
    Chat with an integrated AI assistant for research, summarization, or casual questions.

  • Image Analyzer
    Upload images within chat and receive detailed descriptions, insights, or answers using AI.

  • Sentiment Analyzer
    Analyzes mood, tone, and coherence in conversations for better emotional awareness.

Concerned about privacy? We’re already working on that — read on.

🛠 How We Built It

The stack powering AIr Chat:

  • Backend: Django
  • Frontend: React.js
  • Database: SQLite
  • AI Services: Google Gemini AI

🤖 Models Used

  • Text Generation: gemini-1.5-pro-latest
  • Image Analysis: gemini-pro-vision

🧗 Challenges We Faced

Every big idea comes with its own set of challenges. Here's what we encountered:

  • 🔄 Switching from a previously known AI model to Gemini required rethinking our architecture and integrations.
  • 🖼️ Integrating Gemini Pro Vision for image analysis demanded a temporary image storage solution and efficient retrieval mechanism.
  • 🧠 In the Sentiment Analyzer, negative or sensitive inputs were sometimes rejected by Gemini. We had to build fallbacks and input sanitization layers.

Despite the hurdles, these challenges led to new features and made the system more robust.

🏆 Accomplishments We're Proud Of

Taking a rough idea to a fully functioning demo during this hackathon was a major milestone. We built the core functionality exactly as we envisioned—and that’s something we’re truly proud of.

🔐 What’s Next for AIr Chat

Privacy-First AI

One common concern:

“Are my messages being sent to external AI servers?”

We’re addressing this with Gemini Nano – a local model that runs entirely on the user's device. No cloud transmission means full control stays with the user.

Expansion Possibilities

  • Slack plugin
  • LinkedIn message assistant
  • Gmail auto-reply helper
  • Standalone productivity chat app

The possibilities are endless with modular AI integration.

🧰 Built With

  • Django
  • React.js
  • Google Gemini (1.5 Pro, Pro Vision, Nano)
  • SQLite