A no-build web UI for chatting with multiple AI providers, with persistent memory, emotion tracking, and multi-agent orchestration. Pure HTML, CSS, and JavaScript.

Recommended: select Eva (AIG) in the model dropdown for the full experience.

Run the installer first. It detects and installs every dependency Eva needs on your machine (Linux or macOS), self-updates the checkout, and can rebuild the AppImage:

./install.sh            # check dependencies, install missing (asks first)
./install.sh --check    # report only, install nothing
./install.sh --yes      # install everything missing, no prompts

Then download or build the standalone Linux AppImage and run it. The bridge starts automatically on a private localhost port.

cd standalone
npm install
npm run dist
./dist/'Eva Standalone-5.2.3.AppImage'

Prereqs on the host: Node.js 24+, Python 3.12+, GitHub Copilot CLI authenticated (copilot auth login). See standalone/README.md.

Tip: For the full Eva experience (persistent memory, emotion tracking, knowledge graph), point Settings > MCP at an Azure Data Explorer cluster you can sign in to. The bridge uses azure-identity device code or interactive login on first use, so no keys are stored. Browser-only setup and the manual ACP bridge launch are documented in README-2.md.

Semantic recall: When an OpenAI API key is set in Settings > Auth, Eva ranks stored facts by meaning (OpenAI text-embedding-3-small, cached on disk) so relevant memories surface even when worded differently. Without a key, recall falls back to synonym-expanded keyword matching.

Skills: Import a skill from pasted text, a URL, a GitHub repo/SKILL.md, or a file in Settings > Models > Skills. Eva normalizes ("Eva'rises") it into her own format, stores it in ADX, and applies the matching skill automatically when a request fits.

• README-2.md: architecture, MCP, ACP, browser-only setup, roadmap
• standalone/README.md: AppImage build and runtime