Task Manager AIPortfolio
AI Engineer Portfolio · Next.js 16

Task Manager AI

A production-style task manager that showcases the patterns I reach for when building LLM features — streaming, tool-calling agents, structured outputs, and retrieval-augmented search over user data.

Streaming textTool callingStructured outputsRAG with pgvector
Open the appView source

What this demonstrates

Three AI surfaces, each implemented with a different production pattern.

🗂

Tasks

Create, browse, and delete tasks. Each one is auto-classified with structured outputs.

💬

AI Chat Assistant

Tool-calling agent that searches, reads, and creates tasks on your behalf.

📊

AI Task Analysis

Structured analysis over your tasks: priority, estimates, and recommendations.

Try it

Jump straight into the task list — create a task and watch it get classified, then ask the chat agent about it.

Open /tasks

How it works

A short tour of the architecture. Read the README for the full breakdown.

  1. 1

    Next.js 16 App Router

    Server Components by default; Server Actions for mutations; Route Handlers for streaming and JSON.

  2. 2

    AI SDK + OpenAI

    streamText for token streaming, generateText + Output.object for structured outputs, tool calls capped with stepCountIs.

  3. 3

    Supabase + pgvector

    Tasks live in Postgres; embeddings stored in a 1536-dim vector column and queried via cosine similarity for RAG.

  4. 4

    Feature-sliced layout

    features/tasks, features/ai, features/chat own their types, store, actions, and components. app/ stays focused on routing.