praveen-ks-2001/hermes-agent-template

Hermes Agent – Railway Template

What it is – A ready‑to‑deploy Railway template that runs the open‑source Hermes Agent (an autonomous LLM‑powered AI assistant from Nous Research). The template adds a web‑based admin UI that lets you configure LLM providers, messaging channels (Telegram, Discord, Slack, …), tools, and manage the Hermes gateway without editing any config files.

Key capabilities

• One‑click deployment on Railway via the badge button.
• Admin dashboard (/setup) for selecting an LLM provider, enabling channels, toggling tools, and controlling the gateway (start/stop/restart).
• Integrated Hermes UI (/) – the native Hermes web interface (chat, keys, skills, analytics, console) is proxied behind the same login.
• User pairing – incoming messages generate pairing requests that you can approve/deny from the dashboard.
• Live status & logs – see gateway uptime, model, pending requests, and stream logs in‑browser.
• Backup & restore – snapshot the entire /data/.hermes directory (config, credentials, chat history, skills) as a zip that can be downloaded for 6 hours.
• Password‑protected admin access (cookie‑based login, default user admin).
• Persistent storage – attach a Railway volume at /data so configuration survives redeploys.

How to get started

1. Obtain an LLM key – sign up at OpenRouter (or any other supported provider) and copy an API key.
2. Create a Telegram bot – talk to @BotFather, run /newbot, and copy the bot token.
3. Deploy – click the Deploy on Railway button, set ADMIN_PASSWORD (or let the system generate one), and attach a volume mounted at /data.
4. Configure – open the app URL, log in (admin / your password), go to /setup and:
   • Choose the provider (e.g., OpenRouter) and paste the key/model.
   • Enable Telegram and paste the bot token.
   • Click Save & Start.
5. Use – message the Telegram bot; the first message appears as a pairing request in the dashboard. Approve it and start chatting.

Supported providers & channels

• Providers – OpenRouter, Anthropic, Google AI Studio, xAI, DeepSeek, Qwen Cloud, GLM, Kimi, NVIDIA NIM, Fireworks AI, AWS Bedrock, Azure Foundry, Ollama Cloud, 9Router, OmniRoute, any OpenAI‑compatible endpoint, and many more.
• Channels – Telegram, Discord, Slack, WhatsApp, Email, Mattermost, Matrix.
• Tool integrations – web search (Parallel, Tavily, Perplexity, etc.), scraping (Firecrawl), image generation (FAL), browser automation (Browserbase), GitHub, voice (Whisper/TTS), memory (Honcho), and others.

Architecture at a glance

Railway container (port $PORT)
 └─ server.py (Starlette + Uvicorn) – handles /login, /setup, /health, proxies / to the native Hermes UI
     ├─ hermes‑dashboard (listening on 127.0.0.1:9119) – only reachable via the proxy
     └─ hermes‑gateway – the actual agent process that talks to Telegram/Discord/etc.

The gateway is supervised by server.py; if it crashes, it is automatically restarted.

Running locally

docker build -t hermes-agent .
docker run --rm -it -p 8080:8080 \
  -e ADMIN_PASSWORD=changeme -v hermes-data:/data hermes-agent

Visit http://localhost:8080 and log in with admin / changeme.

Updating – The template pins a specific Hermes release via the HERMES_REF build argument (default v0.22.2). To upgrade, deploy a matching release/<version>/<n> branch or override HERMES_REF (the latter may break existing volumes). Always back up the /data volume before changing versions.

Who might use this – Developers or hobbyists who want a self‑hosted autonomous AI assistant with minimal ops effort, especially those who prefer Railway’s simple deployment model and a web UI for managing channels, tools, and user access.

---

Credits – Hermes Agent by Nous Research; UI inspired by the OpenClaw Railway admin template.