Deploy SillyTavern 1.19 AI Chat and Character UI

Deploy and Host SillyTavern with Railway

Run SillyTavern, the open-source chat interface for power users of large language models, on Railway with a login from the first minute. Bring any model provider or your own endpoint, keep characters, lorebooks, chats and extensions on a persistent volume, and reach it from any browser.

About Hosting SillyTavern

SillyTavern is a single Node.js server: a web UI for chatting with LLMs through character cards, World Info lorebooks, group chats, presets and extensions. It stores everything in files, including your provider API keys, so a hosted copy must never be reachable without a login. Upstream's Docker setup assumes a private network. This community template runs the official ghcr.io/sillytavern/sillytavern:1.19.0 image as an unprivileged user, forces user accounts on, creates a generated admin account before the server starts, hides account names on the login page, and keeps config, data, plugins and extensions on one Railway volume.

Common Use Cases

• Long-form roleplay and collaborative fiction with character cards and lorebooks that follow you between phone and desktop
• One interface for many model providers (OpenAI, Anthropic, Google, OpenRouter, DeepSeek, NovelAI, AI Horde) with saved presets per model
• A front end for your own model server (Ollama, KoboldCpp, vLLM or any OpenAI-compatible API) running elsewhere or on Railway
• A shared instance for a small group: each person gets an account with separate chats, characters and API keys
• Trying community extensions for image generation, text-to-speech, translation and vector memory

Dependencies for SillyTavern Hosting

• An API key or endpoint for at least one model provider, entered in SillyTavern after the first login (AI Horde works without an account). Not needed to deploy.
• A Railway volume at /data (included)

Deployment Dependencies

• SillyTavern repository and releases: https://github.com/SillyTavern/SillyTavern
• Container image: https://github.com/SillyTavern/SillyTavern/pkgs/container/sillytavern
• Documentation: https://docs.sillytavern.app/
• Docker installation and config.yaml reference: https://docs.sillytavern.app/installation/docker/ and https://docs.sillytavern.app/administration/config-yaml/
• Multi-user accounts: https://docs.sillytavern.app/administration/multi-user/
• Extensions: https://docs.sillytavern.app/extensions/

Implementation Details

First login

1. Open the service's public URL once the deployment is healthy.
2. The login page asks for a handle. Enter ST_ADMIN_HANDLE and ST_ADMIN_PASSWORD from the service's Variables tab. Both are generated at deploy; the handle is random and the login page does not list accounts, so nobody can target the admin's password recovery.
3. Pick a persona name, open API Connections (plug icon) and add a provider key or endpoint.
4. Import a character card or open the bundled one and start chatting.
5. Other people: User Settings → Admin Panel creates their accounts. Give them hard-to-guess handles.

Security notes: the login cannot be switched off in this template. User accounts are pinned on; whitelist mode, basic auth (it would block Railway's healthcheck) and SillyTavern's security override are pinned off. A variable or config.yaml value that tries to change them is logged and ignored. API keys are stored on the server, in each user's secrets.json on the volume, and are never sent back to the browser. UI extensions are third-party JavaScript running with your session, so install them only from sources you trust. Server plugins run inside the server with access to every user's data and keys; they are off (SILLYTAVERN_ENABLESERVERPLUGINS=false) until you enable them for code you have reviewed.

Password: change ST_ADMIN_PASSWORD and redeploy to reset it. A password changed inside SillyTavern is kept until you change the variable.

Configuration: config.yaml is created on the first boot at /data/config/config.yaml and is never rewritten. Edit it with railway ssh and redeploy, or set any key as a variable: SILLYTAVERN_ plus the key in upper case, dots as underscores (for example SILLYTAVERN_SESSIONTIMEOUT=604800).

Scaling: SillyTavern keeps its state in files and is a single writer. Keep one replica and scale vertically. It idles at roughly 0.2-0.3 GB RAM; built-in local models for vector memory, captioning or speech add more while used.

Pinning and upgrades: the Dockerfile pins ghcr.io/sillytavern/sillytavern:1.19.0 by tag and digest. To upgrade, change both to a newer release from https://github.com/SillyTavern/SillyTavern/releases and redeploy. SillyTavern migrates its data and adds new config keys on start; everything on the volume is kept. Download a backup from the app first.

Why Deploy SillyTavern on Railway?

SillyTavern is most useful when it is always there: the same characters, lorebooks and chats on every device. Railway keeps it running with HTTPS, a persistent volume, automatic restarts, logs and SSH access, without a server to maintain. You pay for the small amount of RAM and CPU it uses; model usage is billed by your provider.