How to Deploy LobeHub on Railway

How to Deploy LobeHub on Railway

I finally figured out how to deploy LobeHub on Railway. LobeHub is an open-source AI agent platform that helps me build my own knowledgebase. Railway is my main SaaS platform for deploying Docker apps.

I did find some templates on the Railway Marketplace. However, I didn't find the right one for me, so I chose to create one myself. I used LobeHub(v2.1.56) to build it.

Choosing the database was a bit troublesome. I couldn't use the Railway official PostgreSQL template, because it doesn't support the pg_search extension. After some research, I chose ParadeDB for my PostgreSQL database, since it supports. I use the Docker image lobehub/lobehub for faster build times. The variables are listed below:

APP_URL="https://${{RAILWAY_PUBLIC_DOMAIN}}"
DATABASE_URL="postgres://${{ParadeDB.PGUSER}}:${{ParadeDB.PGPASSWORD}}@${{ParadeDB.PGHOST}}:${{ParadeDB.PGPORT}}/${{ParadeDB.PGDATABASE}}"
QSTASH_TOKEN=""
## deployed a SearXNG instance on Railway
SEARCH_PROVIDERS="searxng"
SEARXNG_URL=""
## the following can be obtained from https://lobehub.com/docs/self-hosting/platform/docker#deploying-on-a-linux-server
KEY_VAULTS_SECRET=""
AUTH_SECRET=""
JWKS_KEY=""
## chose Cloudflare R2
S3_ACCESS_KEY_ID=""
S3_SECRET_ACCESS_KEY=""
S3_ENDPOINT=""
S3_BUCKET=""

Then set the port in the networking settings to 8080 .

After the above setup, you will have a LobeHub instance up and running!

However, I can't run commands on my self-hosted instance because the market authorization failed:

Market authorization expired. An authorization dialog has been shown to the user. Please wait for the user to complete authorization and then retry the current task.

I'm thinking about how to remove that part so that I can run cmds freely. I'm working on it.

Add third-party sandbox(AIO Sandbox) to LobeHub

I'm not going to remove the official cloud-sandbox – it's too much trouble. I deployed AIO sandbox on Railway, accessing it through MCP(streamable HTTP). But it needs a lot of memory to run, so I need to enable serverless mode to reduce resource usage. And it takes about 20 minutes to deploy, so don't worry if it doesn't deploy quickly.