šŸ§  SmartPi Agentic Assistant

An LLM-Powered, Expandable Raspberry Pi Pico W Personal Assistant

šŸ§© Overview

Hand-drawn concept sketch of SmartPi Assistant showing dimensions and features
Conceptual design sketch showing the SmartPi device form factor with display showing sample data (Cambridge, MA, 27Ā°F, "Wear a Jacket", "Meeting at 10:30AM with John Doe")

SmartPi Agentic Assistant is a compact, low-power personal assistant built on the Raspberry Pi Pico W, using large language models (LLMs) to convert raw digital data ā€” from your calendar, email, weather feeds, or news sources ā€” into concise text for a 64Ɨ64 display, and natural speech via a small speaker.

It bridges the physical world of sensors, displays, and sound with the reasoning power of LLMs, creating a modular platform for context-aware, ambient intelligence at the edge. Think of it as a pocket-sized digital public infrastructure node ā€” open, extensible, and locally controllable.

Generated technical sketch of SmartPi hardware components and layout
Technical hardware layout showing component placement and connectivity for the SmartPi device

šŸ§® System Design (High-Level)

SmartPi Agentic Assistant high-level system design diagram showing hardware enclosure with display, Wi-Fi, laptop/USB, and software blocks (n8n, SmartPi MCP client, LLM, servers, agentic workflow)
Hardware, networking, and software flow: enclosure with display (SmartPi device) connects over Wi-Fi/USB; a local gateway orchestrates agentic workflows using LLMs, tools, and servers.

Hardware (HW)

  • SmartPi enclosure: Pico W carrier + 64Ɨ64 RGB matrix, mic, speaker (3D-printed box).
  • Connectivity: Wi-Fi ā†’ router; USB tether to user laptop for debug/flash.

Networking

  • Device ā†” local gateway over HTTP/WebSocket.
  • Gateway ā†” cloud services for LLMs and APIs.

Software (SW)

  • Agentic workflow: task orchestration (e.g., daily briefing pipeline).
  • n8n (or similar) for low-code glue and triggers.
  • SmartPi MCP client (Model Context Protocol) to expose device tools to LLMs.
  • LLM for summarization, intent routing, tool use.
  • Servers/tools: Calendar, Email, Weather, and other MCP tools.
  • Cloud LLMs: interchangeable providers (OpenAI, Google, etc.).

Everything is modular: swap LLM providers, add new tools/servers, or change workflows without touching device firmware.

šŸ§± Core Goals

  1. Useful Daily Information
    • Display calendar appointments, reminders, and tasks on a 64Ɨ64 RGB matrix.
    • Fetch weather from APIs and visualize with icons and short text.
    • Summarize news headlines or personal alerts and read them aloud via a speaker.
  2. Conversational & Context-Aware
    • Use a microphone + speaker for two-way interactions.
    • LLMs process queries and return outputs optimized for the small display and voice.
  3. Expandable & Modular
    • Plug in new "capability modules": email summarizer, IoT control, daily-focus coach.
    • Simple plugin architecture so developers can add new use-cases.

šŸ§° Hardware Setup

ComponentPurpose
Raspberry Pi Pico WWi-Fi-enabled MCU for data fetch + gateway integration
64Ɨ64 RGB LED MatrixDisplay for calendar, icons, and visual feedback
IĀ²S / PWM SpeakerAudio output for speech synthesis
Microphone (IĀ²S / analog)Optional voice input for interactions
Sensors (DHT11, LDR, PIR)Environmental awareness

šŸ§  Software Architecture

  1. Microcontroller Layer (Pico W)
    • MicroPython or C++ SDK: LED rendering, I/O, Wi-Fi client.
    • Talks to the gateway via lightweight HTTP/WebSocket.
  2. Integration Layer (Local Gateway)
    • Runs on Pi Zero/4 or a small server.
    • Handles OAuth/API auth for Calendar, Email, Weather.
    • Hosts agentic pipelines (e.g., n8n) and the SmartPi MCP client.
  3. LLM Layer
    • Interchangeable LLM providers for summarization, intent routing, tool calls.
    • Generates both short text for the LED matrix and TTS text for audio.

šŸŖ„ Example Use-Cases

šŸ—“ Calendar Display

šŸŒ¦ Weather Update

šŸ“° News Headlines

āš™ļø Future Expansion Ideas

šŸŒ Why It Matters

Applies Digital Public Infrastructure principles ā€” openness, modularity, interoperability ā€” to edge AI. It's a step toward human-centric, interpretable intelligence that lives in your space, not locked away in a cloud tab.

šŸ”Ž Open Questions / Next Steps