SAGE

Software Assistant for Generative Embedded

AI-Powered Code Generation
ESP32
CrewAI
Embedded Systems
Arduino
LLM Integration

An intelligent code generation and flashing tool for ESP32C6 development boards powered by local large language models. SAGE transforms natural language descriptions into complete, optimized Arduino code.

SAGE streamlines embedded systems development by generating Arduino code from natural language descriptions. Using a collaborative AI approach with CrewAI, SAGE simulates a development team with Senior and Junior Engineer agents working together.

The Senior Engineer analyzes requirements and creates technical specifications, while the Junior Engineer generates the actual code. This collaborative approach creates higher quality code than traditional code generation tools.

Documentation
SAGE-demo.jpg
SAGE Project

Key Features

AI-Powered Code Generation

Generate Arduino code from natural language descriptions like "make LEDs blink in sequence" or "create a WiFi status monitor."

Automated Compilation & Flashing

Build and upload code to ESP32C6 boards with a single command, streamlining your development workflow.

Intelligent Error Handling

Automatically analyze and fix compilation errors, reducing debugging time and frustration.

ESP32C6-Specific Optimizations

Code tailored specifically for ESP32C6 hardware, taking advantage of its unique capabilities.

Technical Architecture

SAGE uses a novel architecture based on the CrewAI framework to simulate a collaborative development team:

SAGE-architecture.jpg
SAGE Architecture

Senior Engineer Agent

  • Analyzes requirements
  • Creates technical specifications
  • Validates code quality
  • Produces documentation

Junior Engineer Agent

  • Generates code based on specifications
  • Implements hardware interfacing
  • Handles debugging and testing

Example Usage

The easiest way to use SAGE is through the interactive mode:

terminal.sh
$ python main.py --interactive

╔══════════════════════════════════════════════════════════════════════════╗
║ SAGE - Software Assistant for Generative Embedded                        ║
║ Generate Arduino code for ESP32C6 using AI with CrewAI Framework         ║
╚══════════════════════════════════════════════════════════════════════════╝

Enter project name: WiFiStatusMonitor

What would you like your ESP32C6 to do?
Connect to WiFi and display the connection status using different LED patterns
(solid GREEN for connected, blinking YELLOW for connecting, solid RED for disconnected).

Initializing SAGE engineering crew...

Generating Arduino code and documentation...
This might take a few minutes as the engineering team works...

Sample Requirements Inputs

LED Sequencing:

"Make the RED, GREEN, and YELLOW LEDs blink in sequence, with each LED staying on for 500ms."

Battery Monitor:

"Create a battery level monitor that displays the current battery level on Serial and shows different colors on the RGB LEDs depending on the battery level."

Development Process

✓

Phase 1: Setup

Command-line interface, Ollama client integration, Arduino CLI wrapper

⟳

Phase 2: Code Generation

LLM prompting strategy, code generation pipeline, CrewAI integration

⟳

Phase 3: Build System

Reliable compilation, cross-platform upload, error recovery mechanisms

Technologies Used

  • Python
  • CrewAI Framework
  • Ollama
  • Arduino CLI
  • ESP32C6
  • Rich (Terminal UI)

Requirements

  • Python 3.8+
  • Arduino CLI
  • Ollama with qwen2.5-coder:14b model
  • ESP32C6 development board
  • CrewAI (auto-installed)

Project Impact

SAGE significantly reduces development time for ESP32C6 projects by automating:

  • Boilerplate code generation
  • Hardware initialization
  • Peripheral configuration
  • Error handling implementation