Echo

An AI-powered workflow automation platform — create, record, and run desktop & browser workflows using voice, chat, or visual recording, powered by the EchoPrism vision-language agent.
Explore the docs »

View Demo · Report Bug · Request Feature

Table of Contents

1. About The Project
   • Architecture
   • Agent Diagram
   • Built With
2. Getting Started
   • Prerequisites
   • Phase 1: GCP Setup
   • Phase 2: Firebase Setup
   • Phase 3: IAM
   • Phase 4: Gemini API Key
   • Phase 5: Local Development
   • Phase 6: Deploy
3. Usage
4. Roadmap
5. Contributing
6. License
7. Contact
8. Acknowledgments

About The Project

Echo is an AI-powered workflow automation platform. Create and edit desktop and browser workflows (from recordings, voice, or chat), then run them via the EchoPrism vision-language agent — which executes steps (navigate, click, type, scroll) on your desktop. Use the web dashboard to manage workflows and runs, and the Electron desktop app for voice-driven control and running your workflows locally.

(back to top)

Architecture

(back to top)

Agent Diagram

(back to top)

Built With

(back to top)

Getting Started

To run the full stack locally or deploy from scratch, follow the phases below.

Prerequisites

Install the following tools before proceeding:

• Node.js 18+ — nodejs.org or nvm install 18
• pnpm — npm install -g pnpm
• Python 3.11+ — python.org or pyenv install 3.11
• Docker — for building and deploying images
• gcloud CLI — Install guide

  gcloud auth login
  gcloud auth application-default login

• Firebase CLI

  npm install -g firebase-tools
  firebase login

• Doppler (optional but recommended) — for secrets management

  brew install dopplerhq/cli/doppler
  doppler login

(back to top)

Phase 1: GCP Setup

1. Go to Google Cloud Console and create or select a project with billing enabled.

2. In APIs & Services → Enable APIs, enable:

   • Cloud Run API
   • Cloud Scheduler API
   • Firestore API
   • Cloud Storage API
   • Gemini API

3. Go to Cloud Storage → Buckets, create a bucket with Uniform bucket-level access, and note the name (e.g. echo-assets-prod).

(back to top)

Phase 2: Firebase Setup

1. Go to Firebase Console and create a new project or link your existing GCP project.

2. Enable authentication: Authentication → Sign-in method → enable Email/Password and Google.

3. Create Firestore: Firestore Database → Create database → choose Native mode.

4. Register your web app: Project Settings → Your apps → Add web app (</>) and copy the config object.

5. Deploy Firestore rules from the project root:

   cd firebase && firebase deploy --only firestore:rules

(back to top)

Phase 3: Service Accounts & IAM

Use the default compute service account for Cloud Run and ensure it has:

• Firestore: Cloud Datastore User (or Firestore roles)
• Storage: Storage Object Admin
• Cloud Run Jobs: Run Jobs Executor

(back to top)

Phase 4: Gemini API Key

1. Go to Google AI Studio
2. Sign in, select your GCP project, and create an API key
3. Copy the key — you'll need it for GEMINI_API_KEY

(back to top)

Phase 5: Local Development

Clone and install:

git clone https://github.com/JasonMun7/echo.git
cd echo
pnpm install
pnpm run install:backend

Option A: Doppler (recommended)

doppler setup   # select project and dev config

Then run each service in a separate terminal:

# Terminal 1 – backend
pnpm run dev:backend

# Terminal 2 – frontend
pnpm run dev

# Terminal 3 – desktop app
pnpm run dev:desktop

# Terminal 4 – EchoPrism Agent
pnpm run dev:agent

# Terminal 5 – LiveKit voice agent
pnpm run dev:livekit-agent

Set NEXT_PUBLIC_ECHO_AGENT_URL (web) and VITE_ECHO_AGENT_URL (desktop) to http://localhost:8081 in Doppler for local agent access.

Option B: .env files

# Web app
cd apps/web && cp .env.local.example .env.local
# Edit .env.local with Firebase config and NEXT_PUBLIC_API_URL=http://localhost:8000

# Backend
cd backend && cp .env.example .env
# Edit .env with ECHO_GCP_PROJECT_ID, ECHO_GCS_BUCKET, GEMINI_API_KEY

Local URLs:

• Frontend: http://localhost:3000
• Backend: http://localhost:8000
• EchoPrism Agent: http://localhost:8081

Environment Variables Reference:

Variable	Required	Description
ECHO_GCP_PROJECT_ID	Yes	GCP project ID
ECHO_GCS_BUCKET	Yes	GCS bucket name
GEMINI_API_KEY	Yes	Gemini API key
NEXT_PUBLIC_API_URL	Yes	Backend URL (web)
NEXT_PUBLIC_ECHO_AGENT_URL	Yes	EchoPrism Agent URL (web)
NEXT_PUBLIC_FIREBASE_*	Yes	Firebase config (web)
VITE_API_URL	Yes	Backend URL (desktop)
VITE_ECHO_AGENT_URL	Yes	EchoPrism Agent URL (desktop)
LIVEKIT_URL	Voice only	LiveKit server URL
LIVEKIT_API_KEY	Voice only	LiveKit API key
LIVEKIT_API_SECRET	Voice only	LiveKit API secret
ECHO_CLOUD_RUN_REGION	No	Default us-central1

See scripts/doppler-env-reference.md for the full reference.

(back to top)

Phase 6: Deploy to Cloud Run

pnpm run deploy
# or with explicit env:
GEMINI_API_KEY=your-key ECHO_GCS_BUCKET=your-bucket \
  ./scripts/deploy.sh YOUR_GCP_PROJECT_ID us-central1

The script builds and pushes Docker images, deploys frontend and backend as Cloud Run services, and deploys the EchoPrism agent as a Cloud Run Job.

To deploy the LiveKit voice agent (optional):

pnpm run deploy:livekit-agent

Requires LIVEKIT_URL, LIVEKIT_API_KEY, LIVEKIT_API_SECRET, LIVEKIT_AGENT_SECRET, ECHOPRISM_AGENT_URL, and GEMINI_API_KEY.

(back to top)

Usage

Visit the live demo to check out our web app. Make sure to follow the instructions in our releases page to ensure the desktop app can be ran.

1. Create a workflow — record a screen capture, describe steps via chat, or use voice on the desktop app
2. Edit steps — review and modify the auto-generated workflow steps in the dashboard
3. Run — trigger a run from the desktop app; EchoPrism executes each step via vision-language grounding
4. Monitor — watch the execution and click Ctrl + Shift + V to interrupt for user steering

(back to top)

Roadmap

• Mobile app automation — Allow Echo to automate tasks on phones as well
• Fine tuning — Improve model accuracy by training on user data with Vertex AI
• Expanded integrations — Add third-party app connectors like Slack, Notion, and G-Suite
• Workflow marketplace — Create a library of community-shared automations users can install and customize
• Schedule workflows — Allow users to schedule workflows to run at specific times
• Reduce costs — Optimize calls to OmniParser service to reduce time and monetary costs

See the open issues for a full list of proposed features and known issues.

(back to top)

Contributing

Contributions are what make the open source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

If you have a suggestion that would make this better, please fork the repo and create a pull request. You can also open an issue with the tag "enhancement". Don't forget to give the project a star!

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

(back to top)

Top Contributors

License

Distributed under the MIT License. See LICENSE.txt for more information.

(back to top)

Contact

Jason Mun — jason.mun484@gmail.com · LinkedIn

Andrew Cheung — andrewcheung360@gmail.com · LinkedIn

Project Link: https://github.com/JasonMun7/echo

(back to top)

Acknowledgments

• OmniParser — UI element grounding for vision-based automation
• LiveKit — Real-time voice and video infrastructure
• Gemini — Vision-language model powering EchoPrism
• UI-TARS — GUI agent model for automated UI interaction
• Best-README-Template

(back to top)