EdgeOrchestra {Open Source Hackathon Submission}

[diyamajee-spec]

Open Source Hackathon 2026 Project Submission

Participant Details

Full Name: DIYA MAJEE

GitHub Username: diyamajee-spec

Team Name: Team CodeX

College/University: P. K. ROY MEMORIAL COLLEGE

---

Project Details

Project Title: EdgeOrchestra

Project Description:
EdgeOrchestra AI (or EdgeOrchestra) is a privacy-first, local-first multi-agent orchestration platform that lets you run a full swarm of specialized AI agents entirely on your own hardware.
It features agents like Vision, Planner, Research, Action, Creative, and Memory, all powered by local LLMs via Ollama (e.g., phi4-mini for reasoning and qwen2.5-vl for multimodal vision). There are no servers, no subscriptions, and no data leaving your device.
Key highlights:

Built with Tauri 2 (Rust backend for lightweight, secure desktop apps) + React 19 + TypeScript.
Uses Loro CRDT for conflict-free, offline memory/state management.
Interactive ReactFlow agent graph for visualization and drag-and-drop file/image handling.
Supports parallel "Orchestra Burst Mode," WASM plugin sandbox, webcam vision, voice I/O (STT/TTS), cinematic UI/animations, and more.
Designed as an "autonomous operating system for local AI workloads" for developers, researchers, and privacy-conscious power users. It's open-source under the MIT license.

The Problem It Solves:
Modern AI tools (especially multi-agent systems) typically rely on cloud services (e.g., OpenAI, Anthropic, or hosted agent platforms). This creates several issues:

Privacy & data sovereignty risks—sensitive prompts, files, images, or conversations are sent to third-party servers.
Ongoing costs — subscriptions, API usage fees, or rate limits.
Dependency on internet — no offline capability; performance tied to cloud latency.
Lack of control — black-box systems, limited customization, and vendor lock-in.
Resource underutilization — users with capable local hardware (GPUs/CPUs) still depend on remote inference.

EdgeOrchestra brings a complete multi-agent AI orchestra to your local machine:

Fully local inference via Ollama → zero cloud dependency.
Autonomous agent coordination — An intent router dynamically builds a dependency graph and delegates tasks across specialist agents.
Persistent memory with Loro CRDT for coherent, long-term context across sessions.
Rich interactivity—Visual agent graph, file/image dropping, real-time visualizations (memory graph with physics), plugins via WASM, and polished cinematic UI.
Extensibility & security — Sandboxed WASM plugins; Tauri for secure OS access.
Easy setup—Clone, install dependencies, pull Ollama models, and run with npm run tauri dev (one-click scripts available).

Tech Stack Used:

Layer	Technology	Role
Desktop Runtime	Tauri 2 (Rust)	Lightweight OS bridge, secure sandboxing, and zero-copy IPC
Frontend Framework	React 19 + TypeScript	Type-safe, component-driven UI architecture
Styling & Animation	Tailwind CSS v4 + Framer Motion	Glassmorphism design system with spring-physics animations
Local AI Inference	Ollama (phi4-mini, qwen2.5-vl)	100% on-device LLM and multimodal vision inference
State & Memory	Zustand + Loro CRDT	Conflict-free, offline-capable replicated memory store
Agent Visualization	ReactFlow / XYFlow	Interactive, draggable node-based agent graph
Memory Visualization	HTML5 2D Canvas Physics	Force-directed memory relationship network with spring layouts
Plugin Runtime	WebAssembly (wasm32-unknown-unknown)	Sandboxed, capability-restricted agent extension system
UI Primitives	Radix UI + Lucide Icons	Accessible, headless component foundation
Speech I/O	Web Speech API (STT + TTS)	Fully offline voice interaction layer

GitHub Repository Link: https://github.com/diyamajee-spec/EdgeOrchestra

Live Demo Link: https://edgeorchestra-1.vercel.app/

Presentation / Demo Video Link: https://gamma.app/docs/EdgeOrchestra-Harmonizing-the-Edge-p6yufjknuh2r3up

---

Open Source Readiness

• My project is public on GitHub
• My repository has a proper README.md
• I have added setup/installation instructions
• I have added screenshots/demo where possible
• I have added a license file
• My project is original and built/updated during the hackathon period

---

Memori Labs Sponsor Task

Please complete these before submitting:

• I have starred the Memori Labs GitHub repository.
  https://github.com/MemoriLabs/Memori

• I have followed Memori Labs on LinkedIn.
  https://www.linkedin.com/company/memorilabs/

• I have followed Memori Labs on X.
  https://x.com/memorilab

• I have checked Memori Labs' social links.
  https://linktr.ee/memorilabs

---

ID Card Verification

• I have generated my ID card from https://oshack.xyz
• If my ID was not verified, I completed the mandatory verification/giveaway form and tried again

---