Skip to content

LiminaLabsAI/edgestack

Repository files navigation

EdgeStack

Your private local-first AI agent business automation hub.

EdgeStack lets non-technical founders and solopreneurs build, run, and monitor AI-powered business workflows completely locally. No cloud accounts, no API subscriptions, no Docker, and no technical knowledge required.

Key Features

  1. Local-First Design: Powered by native Ollama and llama.cpp. Zero cloud dependency.
  2. Embedded Cloud Emulator (Floci): Emulates cloud APIs (S3, SQS, SES, DynamoDB, etc.) inside a lightweight local JVM subprocess.
  3. Bedrock AI Bridge: Intercepts cloud SDK Bedrock Runtime requests and routes them to your local Ollama instance on port 11434.
  4. Human-in-the-Loop (HITL) Gateways: Failsafe workflows that pause on error, analyze failures using local AI, notify you, and await explicit retry approval.
  5. Vibrant Telemetry & Cost Analytics: Real-time dashboard showing CPU/Memory/GPU usage alongside dynamic cost saving charts calculated using the local LLM.

Prerequisites

  • macOS 13.0 or higher
  • Java 21+ (For running the local cloud services emulator)
  • Ollama installed locally (Optional, EdgeStack can download models automatically)

Running Locally

To build and run the development version of the application:

# Install NPM dependencies
npm install

# Start Tauri development environment
export PATH=$PATH:/Users/sarang/.local/node/bin
npm run tauri dev

Production Build

To compile a distributable macOS Universal DMG package:

npm run tauri build -- --target universal-apple-darwin

Directory Structure

All application state, local databases, and assets are stored in the user's home directory:

  • SQLite Database: ~/edgestack/edgestack.db (stores workflows, execution history, human-in-the-loop gate approvals, and collected telemetry records)
  • Configuration: ~/edgestack/config.toml (stores general settings, resource allocations, notifications, and local model inference setup)
  • Pricing Records: ~/edgestack/bedrock_pricing.json (defines local cost tier metrics for AWS Bedrock model comparison)
  • Subdirectories:
    • ~/edgestack/vault/ (local vault storage objects)
    • ~/edgestack/floci/ (workspace for the Floci emulator subprocess)
    • ~/edgestack/workflows/ (workflow definitions and run history details)
    • ~/edgestack/logs/ (application execution and diagnostic logs)
    • ~/edgestack/bin/ (external helper binaries, including floci.jar)

Releases

No releases published

Packages

 
 
 

Contributors