EN
FRAI Portfolio Assistant with Real-Time Streaming
Traditional portfolios are static and cannot interact with visitors. Recruiters only see screenshots and text. I wanted to create a smart portfolio that can answer questions, explain my skills, and provide a real interactive experience.
I built an AI assistant integrated directly into my portfolio using Claude Haiku. The assistant supports real-time streaming conversations, multi-turn chat memory, terminal commands, and keyboard shortcuts. Users can: - Ask questions about my projects - Explore my AI skills - Interact with a terminal-style interface - Receive instant AI responses
I used Next.js Edge Runtime to reduce latency and improve streaming performance. Claude Haiku was selected because it provides fast responses at a low cost.
The architecture separates:
- Frontend chat interface
- Edge API route
- AI processing layer
Streaming responses improve user experience because users see the answer token by token in real time.
Prompt caching was added to reduce API costs and improve response speed.
Stack
Challenges
The hardest challenge was implementing real-time streaming correctly while keeping the UI smooth and responsive.
Other challenges included:
- Managing conversation history
- Reducing API costs
- Handling API failures gracefully
- Optimizing response latency
- Designing a terminal-style AI experience