Portfolio Agent Hub
OVERVIEW
A portfolio website with an embedded AI assistant that answers visitor questions about projects, skills, experience, and availability in real time. The assistant uses tool-calling to query structured data, providing accurate, contextual responses rather than generic AI output.
ARCHITECTURE
The app uses Next.js App Router for content delivery and a server-side chat endpoint powered by Vercel AI SDK. Typed tool interfaces expose project, skills, and availability data as structured functions. The client chat UI streams model output while preserving session context for follow-up questions.
FUNCTIONALITY
- Embedded AI chat interface with streaming responses
- Tool-calling architecture for structured data retrieval
- Context-aware responses grounded in actual portfolio data
- Availability and scheduling queries with calendar integration
- Conversation memory within session for natural follow-up questions
- Graceful fallbacks when questions fall outside the knowledge base
HOW IT WORKS
User prompts are sent to the server route, where the model can call typed tools instead of guessing. Tool results are merged into the response context, then streamed back token-by-token to the UI. The prompt scaffold is generated from current portfolio content so answers stay grounded and maintainable.
OUTCOMES
- Gives visitors immediate answers about projects, skills, and availability.
- Improves response quality by grounding outputs in structured portfolio data.
- Keeps chat responses maintainable through typed tool contracts and dynamic context composition.