eVera — Your Sovereign AI Agent

eVera is a fully autonomous personal AI agent that works completely offline, on your local network, or with the full internet. It controls your computer, runs 160+ local AI models, speaks with a 3D animated avatar, and is available on every platform — Web, Desktop, Mobile, and VS Code.

---

Download

Pre-built installers are published automatically on every tagged release via GitHub Actions CI. No Python, Node.js, or any other dependency needs to be installed manually.

Platform	Download	Install
Windows	Vera-Setup-*.exe	Run the installer — everything is bundled
macOS	Vera-*.dmg	Open DMG → drag Vera to Applications
Linux (AppImage)	Vera-*.AppImage	chmod +x Vera-*.AppImage && ./Vera-*.AppImage
Linux (deb)	Vera-*.deb	sudo dpkg -i Vera-*.deb
Android	Vera-*.apk	Enable "Install from unknown sources" → tap APK
Android Auto	Vera-Auto-*.apk	Install on phone → connect to car
Wear OS	Vera-Wear-*.apk	Sideload or install via Play Store
iOS	Vera.xcarchive.zip	Open in Xcode → distribute to device
Web (PWA)	Open in browser	Install from browser address bar (Chrome/Edge/Safari)
VS Code	evera-ai.vsix	code --install-extension evera-ai.vsix

All binaries are built and published automatically by GitHub Actions on every v* tag push. See Actions for live build status.

---

One-Command Install (from source)

Linux / macOS

curl -fsSL https://raw.githubusercontent.com/embeddedos-org/eVera/master/install.sh | bash

This single command clones the repo, installs Python 3.12, creates a virtual environment, installs all Python packages, installs system libraries (tesseract, ffmpeg, portaudio, xdotool), installs Playwright Chromium, installs Ollama, pulls a default offline model, and installs Electron dependencies. Nothing else is needed.

Windows (PowerShell)

irm https://raw.githubusercontent.com/embeddedos-org/eVera/master/setup.ps1 | iex

After Setup

source .venv/bin/activate      # Linux/Mac
# .venv\Scripts\activate       # Windows

python main.py --mode server
# Open http://localhost:8000

For full manual installation steps, see INSTALL.md.

---

How eVera Compares

Feature	eVera	Manus	ChatGPT	Claude	OpenClaw	Cursor	AnythingLLM	AutoGen
Fully offline (no internet)	✅	❌	❌	❌	❌	❌	✅	❌
Runs on your hardware	✅	❌	❌	❌	❌	❌	✅	✅
3D animated avatar	✅	❌	❌	❌	❌	❌	❌	❌
Multi-accent voice (TTS)	✅	❌	❌	❌	❌	❌	❌	❌
Computer / OS control	✅	✅	❌	❌	❌	❌	❌	❌
LAN / org network access	✅	❌	❌	❌	❌	❌	❌	❌
LOCAL / LAN / WWW modes	✅	❌	❌	❌	❌	❌	❌	❌
Virtual LLM router	✅	❌	❌	❌	❌	❌	❌	❌
RAG knowledge base	✅	❌	❌	❌	❌	❌	✅	❌
Multi-agent crew	✅	✅	❌	❌	❌	❌	❌	✅
Workflow automation	✅	✅	❌	❌	❌	❌	❌	❌
VS Code extension	✅	❌	❌	❌	❌	✅	❌	❌
WhatsApp integration	✅	❌	❌	❌	✅	❌	❌	❌
Telegram integration	✅	❌	❌	❌	✅	❌	❌	❌
Discord integration	✅	❌	❌	❌	✅	❌	❌	❌
Web PWA	✅	✅	✅	✅	✅	❌	✅	❌
Desktop app	✅	❌	❌	✅	❌	✅	✅	❌
Mobile app (Android/iOS)	✅	❌	✅	✅	✅	❌	❌	❌
Android Auto	✅	❌	❌	❌	❌	❌	❌	❌
Wear OS	✅	❌	❌	❌	❌	❌	❌	❌
54+ specialized agents	✅	✅	❌	❌	❌	❌	❌	✅
160+ LLM models	✅	❌	❌	❌	❌	❌	❌	❌
Ollama offline models	✅	❌	❌	❌	❌	❌	✅	❌
LM Studio / Jan / llama.cpp	✅	❌	❌	❌	❌	❌	✅	❌
Open source	✅	❌	❌	❌	✅	❌	✅	✅
Free (no subscription)	✅	❌	❌	❌	❌	❌	✅	✅
Data stays on your machine	✅	❌	❌	❌	❌	❌	✅	✅

---

Operating Modes

eVera works in three distinct modes, selectable from the header dropdown at any time.

Mode	Internet Required	LLMs Available	Computer Control	LAN Access
LOCAL	No	Ollama, LM Studio, Jan, llama.cpp	Full	No
LAN	No	All offline + LAN-hosted servers	Full	Yes
WWW	Yes	All 100+ models including cloud	Full	Yes

Switch modes without restarting — the agent adapts its tool set and model selection automatically.

---

Offline LLMs

eVera works completely without internet using Ollama. Pull any model after setup:

# Recommended — best balance of quality and speed
ollama pull qwen3:8b          # 5 GB
ollama pull llama3.2:3b       # 2 GB — fastest
ollama pull deepseek-r1:7b    # 5 GB — best reasoning
ollama pull qwen2.5-coder:7b  # 5 GB — best code

# Ultra-low RAM (under 4 GB)
ollama pull qwen3:0.6b        # 400 MB
ollama pull llama3.2:1b       # 700 MB

# High-end hardware
ollama pull qwen3:32b         # 20 GB
ollama pull llama3.3:70b      # 40 GB

eVera also auto-detects and integrates with:

Tool	Default Port	Notes
LM Studio	1234	Load any GGUF model, start local server
Jan AI	1337	Load any model, enable API server
llama.cpp	8080	./llama-server -m model.gguf --port 8080

All three appear in the model selector under green 🟢 groups, sorted before cloud providers.

---

Features

Computer Control (LOCAL mode)

eVera can control any function of the running computer system:

• Screen: take screenshots, read screen content, find UI elements
• Mouse & Keyboard: click, type, drag, scroll — full desktop automation
• Files: read, write, move, search, compress, extract
• Applications: launch, close, interact with any app
• System: manage processes, services, network interfaces, power state
• Browser: full web automation via Playwright (Chromium)
• Terminal: execute shell commands, run scripts, manage environments

3D Avatar

A real-time 3D holographic avatar (Three.js WebGL) is displayed alongside the chat panel. The avatar:

• Animates through 8 expressions: idle, thinking, speaking, happy, excited, sad, error, listening
• Responds to voice amplitude in real time (mouth movement)
• Falls back to an animated 2D canvas face if WebGL is unavailable
• Is visible on desktop, web, and mobile (bottom tab on mobile)

Voice — Multi-Accent TTS

All browser-native TTS voices are available in the voice selector, grouped by language and accent. Local (offline) voices are marked with ●, network voices with ○. Selection is saved across sessions.

Virtual LLM Router

eVera's Virtual LLM Router automatically selects the best available model for each task — no manual model selection needed:

Task	Preferred Offline Model	Cloud Fallback
Coding	qwen2.5-coder:7b	gpt-4o
Reasoning / math	qwq:32b	o3-mini
Vision	llava:13b	gpt-4o
Fast / chat	qwen3:0.6b	gpt-4o-mini
Long context	qwen3:14b	gemini-2.5-pro
Embedding	mxbai-embed-large	(offline only)

The router checks model availability in real time and falls back automatically if a model is not pulled.

VS Code Extension

Install the eVera extension to get AI assistance directly in your editor:

# From release download
code --install-extension evera-ai.vsix

• Inline chat — Ctrl+Shift+V to open eVera chat in the sidebar
• Explain code — right-click any selection → "eVera: Explain"
• Fix bug — right-click → "eVera: Fix Bug"
• Generate code — right-click → "eVera: Generate"
• Inline completion — Copilot-style suggestions as you type
• Works offline — uses your local Ollama models, no cloud required

WhatsApp Integration

Two modes — no account needed for local mode:

# Local (scan QR once, no API key)
cd vera/messaging && npm install whatsapp-web.js qrcode-terminal express
node whatsapp_bridge.js

# Or Twilio (production)
TWILIO_ACCOUNT_SID=ACxxx TWILIO_AUTH_TOKEN=xxx python main.py

54+ Specialized Agents

eVera routes every request to the best-suited agent automatically:

Category	Agents
Research	Web search, academic papers, news, Wikipedia
Code	Write, review, debug, execute, explain
Files	Read, write, convert, summarize, extract
System	Shell, process manager, service control
Browser	Navigate, scrape, fill forms, screenshot
Memory	Working memory, episodic memory, fact extraction
LAN	Network scanner, SSH, file shares, org data
Creative	Writing, brainstorming, translation, summarization
Messaging	WhatsApp, Telegram, Discord, Slack

Memory System

• Working memory: current conversation context
• Episodic memory: past conversations, searchable
• Fact memory: extracted facts about the user and their environment, persisted across sessions

---

Architecture

eVera/
├── main.py                  ← Entry point (server / desktop / cli modes)
├── config.py                ← All settings (Pydantic, .env-driven)
├── setup.sh                 ← One-command Linux/macOS installer
├── install.sh               ← curl-pipe remote installer
├── setup.ps1                ← One-command Windows installer
├── INSTALL.md               ← Full installation guide
├── vera/
│   ├── brain/               ← Agent orchestration, memory, planning
│   │   ├── agents/          ← 39+ specialized agents
│   │   ├── computer_use/    ← Desktop control (mouse, keyboard, screen)
│   │   ├── lan_agent.py     ← LAN network discovery and access
│   │   └── orchestrator.py  ← Multi-agent routing
│   ├── providers/
│   │   ├── models.py        ← 100+ model definitions across 12+ providers
│   │   └── manager.py       ← LiteLLM routing, health checks, streaming
│   ├── operating_mode.py    ← LOCAL / LAN / WWW mode enforcement
│   └── static/              ← Web UI (PWA)
│       ├── index.html
│       ├── app.js           ← Main UI logic, voice, mobile nav
│       ├── face.js          ← Three.js 3D avatar + 2D fallback
│       ├── style.css        ← Glassmorphism design system
│       ├── manifest.json    ← PWA manifest
│       └── sw.js            ← Service worker (offline cache)
├── electron/                ← Desktop app wrapper
│   ├── main.js              ← Electron main process
│   └── package.json         ← electron-builder config
├── mobile/                  ← React Native mobile app
│   ├── android/             ← Android (Phone + Auto + Wear)
│   └── ios/                 ← iOS
└── .github/workflows/
    └── build.yml            ← CI: builds all platforms on every v* tag

---

CI / Build Pipeline

Every push to a v* tag triggers a full build across all platforms simultaneously:

Job	Runner	Output
Pre-flight	ubuntu-latest	Lint (ruff), pytest, frontend validation
Desktop — Windows	windows-latest	Vera-Setup-*.exe (NSIS installer)
Desktop — macOS	macos-latest	Vera-*.dmg
Desktop — Linux	ubuntu-latest	Vera-*.AppImage, Vera-*.deb
Android	ubuntu-latest	Vera-*.apk, Vera-Auto-*.apk, Vera-Wear-*.apk
iOS	macos-latest	Vera.xcarchive.zip (unsigned, for distribution)
VS Code	ubuntu-latest	evera-ai.vsix
GitHub Release	ubuntu-latest	Creates release, attaches all 9 artifacts

The Python backend is bundled into a self-contained binary via PyInstaller before Electron packages it. Users need no Python, Node.js, or any runtime installed.

To trigger a release manually:

git tag v2.x.x
git push origin v2.x.x

---

Configuration

All settings are in .env (copy from .env.example). No setting is required — eVera runs fully offline with zero configuration.

Variable	Default	Description
VERA_LLM_OLLAMA_URL	http://localhost:11434	Ollama server
VERA_LLM_LM_STUDIO_URL	http://localhost:1234	LM Studio server
VERA_LLM_JAN_URL	http://localhost:1337	Jan AI server
VERA_LLM_LLAMACPP_URL	http://localhost:8080	llama.cpp server
VERA_LLM_OPENAI_API_KEY	(empty)	OpenAI (optional)
VERA_LLM_ANTHROPIC_API_KEY	(empty)	Anthropic (optional)
VERA_LLM_GEMINI_API_KEY	(empty)	Google Gemini (optional)
VERA_SERVER_PORT	8000	Web server port
VERA_SERVER_API_KEY	(empty)	Auth key for LAN/WWW mode

---

Development

git clone https://github.com/embeddedos-org/eVera.git
cd eVera
bash setup.sh
source .venv/bin/activate
python main.py --mode server

Run tests:

pytest tests/ -v

Build desktop app locally (Linux):

bash setup.sh --build-desktop
# Output: electron/dist/Vera-*.AppImage

---

License

MIT License. See LICENSE.

---

eVera is built by embeddedOS — open, local-first, and fully under your control.