Deploy and Host Open Notebook on Railway

Open Notebook is an open-source, self-hosted alternative to Google's NotebookLM. You collect sources (PDFs, web pages, YouTube, audio, text), chat with them with citations, keep notes, run reusable transformations, search across everything, and generate multi-speaker podcasts. It works with OpenAI, Anthropic, Google, Groq, Mistral, OpenRouter, Ollama, and any OpenAI-compatible endpoint, configured from the UI.

About Hosting Open Notebook

Hosting Open Notebook means running its container (Next.js UI, FastAPI backend, and a background worker under supervisord) next to a SurrealDB instance. This template mirrors the upstream two-service docker-compose.yml: the app image lfnovo/open_notebook:1.14.0 with a volume at /app/data for uploads, podcasts, and caches, and surrealdb/surrealdb:v2.6.5 with a volume at /data for the RocksDB datastore. Only the UI port (8502) is public; the frontend proxies /api/* to the backend, so a single Railway domain covers both. API_URL is pinned to the generated public domain because the frontend would otherwise auto-detect a :5055 URL that Railway does not expose. SurrealDB binds [::]:8000 for IPv6 private networking and runs with RAILWAY_RUN_UID=0 so the non-root image can write the root-mounted volume. Both services have healthchecks with PORT set to their listen ports.

Common Use Cases

Research workspace: organize papers, articles, and transcripts into notebooks and ask grounded questions
Private NotebookLM: keep sources and notes on your own infrastructure with your own provider keys
Podcast generation: turn a notebook into a multi-speaker episode with configurable speaker profiles
Content transformations: summaries, key insights, study guides, and custom prompts over any source
Team knowledge base: password-protected instance with search across sources, notes, and insights

Dependencies for Open Notebook Hosting

SurrealDB 2.x: required, included as an internal-only service with its own volume
AI provider API key: needed for chat, embeddings, transformations, and podcasts; added after deploy in Manage → Models (stored encrypted in the database)
No Redis, Postgres, or GPU required

Deployment Dependencies

Implementation Details

First steps after deploy:

Open the generated domain and sign in with OPEN_NOTEBOOK_PASSWORD (Open Notebook service → Variables)
Go to Manage → Models → Add Credential, pick a provider, paste the key, Test Connection, Discover Models
Create a notebook, add a source, start chatting

Key variables:

# Open Notebook
PORT=8502
API_URL=https://${{Open Notebook.RAILWAY_PUBLIC_DOMAIN}}
OPEN_NOTEBOOK_ENCRYPTION_KEY=${{secret(64)}}
OPEN_NOTEBOOK_PASSWORD=${{secret(24)}}
SURREAL_URL=ws://${{SurrealDB.RAILWAY_PRIVATE_DOMAIN}}:8000/rpc
SURREAL_USER=${{SurrealDB.SURREAL_USER}}
SURREAL_PASSWORD=${{SurrealDB.SURREAL_PASS}}
SURREAL_NAMESPACE=open_notebook
SURREAL_DATABASE=open_notebook

# SurrealDB (start command: /surreal start --log info)
PORT=8000
SURREAL_BIND=[::]:8000
SURREAL_USER=root
SURREAL_PASS=${{secret(32)}}
SURREAL_PATH=rocksdb:/data/mydatabase.db
RAILWAY_RUN_UID=0

Do not rotate OPEN_NOTEBOOK_ENCRYPTION_KEY after saving provider credentials, and do not change SURREAL_PASS after first boot (the root user is only created when none exists). Provider keys via env (OPENAI_API_KEY, etc.) still work but are deprecated upstream; the UI is the source of truth. Optional OPEN_NOTEBOOK_ENABLE_DOCLING / OPEN_NOTEBOOK_ENABLE_CRAWL4AI install heavy extraction runtimes on the next boot and cache them on the volume.

Why Deploy Open Notebook on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Open Notebook on Railway, you get a password-protected research workspace with SurrealDB, generated secrets, persistent volumes for both services, healthchecks, managed SSL on one public domain, and private networking between the app and its database.