Alexa AI

Architecture

The assistant is built with a FastAPI backend and a Next.js frontend in TypeScript. Language models run locally through Ollama, and responses stream to the interface as they are generated.

• FastAPI backend
• Next.js + TypeScript frontend
• Ollama running local LLMs
• Streaming responses

Memory System

Memory is split into structured and semantic layers: SQLite stores structured memory while ChromaDB provides semantic recall. Memory is organized into categories so the assistant can distinguish what kind of thing it is remembering.

• Short-term memory
• Long-term memory
• Project memory
• User preference memory
• Task memory

Tools System

A tools registry gives the assistant controlled capabilities beyond conversation: file tools, system tools, project tools, and controlled terminal operations — each exposed deliberately rather than granting open system access.

Local-First Privacy

The local-first architecture is the point of the project: conversations, memory, and tool activity stay on the user’s machine. Privacy is a structural property, not a policy promise.

Current State

Alexa AI is a functional prototype under active development. The core loop — local models, streaming chat, persistent memory, and tools — works today and continues to be extended.