// the find
EKKOLearnAI/hermes-web-ui
Web dashboard for Hermes Agent — multi-platform AI chat, session management, scheduled jobs, usage analytics
A Vue 3 + Koa web dashboard, desktop app, and CLI wrapper for the Hermes Agent runtime. It handles chat sessions, multi-platform channel configuration (Telegram, Discord, Slack, WhatsApp, etc.), scheduled jobs, file browsing, and usage analytics — all pointed at a locally running Hermes Agent process. Think of it as the control plane for a self-hosted AI agent, not a standalone product.
- Deployment options are well thought out: npm CLI, Docker Compose, and an Electron desktop app all ship from the same codebase, and the auto-update path has a CDN-first fallback to GitHub Releases. That's real operational care.
- Development velocity is visible: the docs/chat-chain-changes directory has 40+ dated entries from just the last two weeks, covering real issues like bridge abort timeouts, Windows PTY decoding, and TTS playback state. Active maintenance, not a one-shot release.
- The BFF namespacing decision — all Hermes-specific code under hermes/ directories across API, components, views, and stores — is the right call for a project that claims multi-agent extensibility. It's not just marketing; the structure enforces the boundary.
- Voice I/O support goes further than most: server-side TTS with stored secrets (so the browser never touches your API key), five TTS providers including a custom voice-clone path, and a barge-in model that stops playback before starting new voice input.
- The license is BSL-1.1, not open source. Production or commercial use is restricted until the change date. This will bite teams that deploy it internally and assume 'self-hosted' means 'free to use however you want' — it does not.
- Default credentials are admin / 123456 with a post-login prompt to change them. Shipping a known default password for a tool that exposes a web terminal and file browser to the local network is a real risk if someone skips that prompt or deploys it in Docker without changing the token.
- The whole stack is useless without a running Hermes Agent process. The README buries this — the architecture diagram makes it clear, but the feature table makes it look like a self-contained product. Anyone who installs it expecting a generic AI dashboard will hit a confusing dependency wall.
- WeChat QR login routes through Tencent iLink API, and channel credentials write to ~/.hermes/.env. There's no mention of what data Tencent receives or retains during that QR flow, which matters if you're using this in any context with privacy requirements.