Deploy RocketRide

AI pipeline engine with visual builder, SDKs, MCP and pgvector

Deploy RocketRide

Just deployed

/var/lib/postgresql/data

/opt/data

Deploy and Host self hosted RocketRide (AI/ML) on Railway

This template deploys the RocketRide engine, an open-source AI pipeline engine, together with a bundled pgvector database.

About Hosting RocketRide open-source software on Railway (self hosted RocketRide template)

This template runs the RocketRide engine on Railway so the pipelines, uploaded files and vector data stay in services you control.

This is independent community packaging of rocketride-org/rocketride-server 3.6.0. It is not affiliated with or endorsed by RocketRide or Aparavi Software AG.

Why Deploy RocketRide, the self-hosted AI pipeline engine on Railway

Stitching model APIs together by hand means you maintain the glue yourself, from retries and chunking to embeddings and agent loops, and you pay each provider separately.

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 RocketRide on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.

Railway vs Other Hosting Providers and VPS for RocketRide self hosting

ProviderWhat You Get with RailwayWhat You Get with the Other Provider
DigitalOceanBoth services deployed from one template, with a public domain and attached volumes.A Linux droplet you configure yourself, including Docker, volumes and TLS.
AWSThe rocketride and pgvector services wired together, with health checks set for you.Broad networking and storage control, but you assemble the container runtime yourself.
HetznerA managed template deploy, so you skip server setup and hardening.Low-cost bare servers where you handle OS updates, Docker and backups.

Choose Railway for a fast deploy, or a VPS when you want to manage the host.

Common Use Cases for hosted RocketRide

  • Document ingestion for retrieval: Build a pipeline that reads PDFs and images, runs OCR, chunks the text, embeds it and writes vectors into the bundled pgvector database.
  • Scripted pipelines from an app: Call the same pipeline from a Python or TypeScript service through the SDK, so your code sends a file in and gets structured output back without managing the model plumbing yourself.
  • Agent workflows with tools: Wire an agent node to LLM and vector-store nodes so it can look up stored context, reason over it and call other nodes in a loop until the task is finished.
  • Editor-driven development: Compose and debug pipelines in the RocketRide VS Code extension, then point the same engine at your public domain when you are ready to share a working version with teammates.
  • Assistant-accessible pipelines: Expose pipelines to an AI assistant through an MCP client that presents your API key, so the assistant can run pipelines you already built instead of rebuilding them each time.

Dependencies for RocketRide Docker hosted on Railway

This template runs two services: a public rocketride engine and a private pgvector database.

Deployment Dependencies for Managed RocketRide Service

The rocketride service is public on port 8080, mounts a volume at /opt/data, checks /version with a 300 second timeout, and uses the Always restart policy. The pgvector service is private, uses image pgvector/pgvector:0.8.0-pg16, stores files under PGDATA, uses database name postgres, and restarts On Failure. Keep one replica of each service. Optional Milvus, etcd, MinIO and ChromaDB from the upstream dev compose file are not included.

Implementation Details for RocketRide

The image is ghcr.io/onepush-dev/rocketride-engine:3.6.0, the unmodified upstream ghcr.io/rocketride-org/rocketride-engine:3.6.0 engine plus railway/entry.sh. The entrypoint creates and chowns /opt/data, warns if ROCKETRIDE_APIKEY is empty, then runs as uid 1000 and starts the engine on 0.0.0.0:8080. Baked values are ROCKETRIDE_PORT=8080 and ROCKETRIDE_LOG_LEVEL=info. /version is public; /ping, /task and the /task/service WebSocket require the API key. POSTGRES_URL is provided for pgvector store nodes (host pgvector.railway.internal, port 5432, user postgres, database postgres, plus the pgvector POSTGRES_PASSWORD); the engine does not use it as a global setting. The first run of a node installs that node's Python packages, so the first pipeline run can take a minute. This packaging describes 50+ nodes, matching the pinned 3.6.0 packaging README, not later unpinned counts. Model provider keys are entered per node and are not needed to boot.

Two template variables on rocketride are generated or referenced, with no inputs:

  • ROCKETRIDE_APIKEY is a 40 character key the UI, SDKs, VS Code extension and MCP clients must present. If it is empty, anyone who finds the URL can run pipelines.
  • POSTGRES_URL is the bundled pgvector DATABASE_URL, for pgvector store nodes.

How does RocketRide compare against other tools

RocketRide vs stitching model APIs by hand (Hand-Rolled Pipelines Alternative)

  • Ingestion and retrieval: Direct model API calls give full control over each request, while RocketRide offers 50+ nodes for LLMs, embeddings, OCR, parsers, chunkers, vector stores and agents in one pipeline.
  • Pipeline definition: Hand-written glue code lives in your repository, whereas RocketRide pipelines are portable JSON files you build visually.
  • Where hand-rolled wins: A single model call with light logic is simpler as plain SDK code than as an engine.

RocketRide vs a local Docker Compose install of RocketRide (Local Compose Alternative)

  • Setup effort: A local Compose install gives full control of network and storage, while this template provisions the engine and pgvector with no inputs needed.
  • Persistence and access: Local data stays on your machine, whereas here state lives on the /opt/data volume and the API is reachable over HTTPS.
  • Where local Compose wins: Local runs keep data off third-party infrastructure and let you add Milvus or ChromaDB, which this template does not include.

RocketRide with Docker vs Railway (Railway Hosting Alternative)

  • Operations: Running the Docker image yourself means managing hosts, TLS and restarts, while Railway handles the domain, the /version health check and restarts.
  • Scaling and replicas: Each service stays at one replica here because local persistent state requires it, so this is not a horizontal scaling setup.
  • Where plain Docker wins: A host you run gives direct control over the machine.

How to use RocketRide on Railway?

  1. Open the rocketride service, go to the Variables tab, and copy the value of ROCKETRIDE_APIKEY.
  2. Open the public domain for the rocketride service. When the UI asks for a key, paste ROCKETRIDE_APIKEY. You can also point the RocketRide VS Code extension, an SDK or an MCP client at the domain with that key. Use a wss URI, for example wss://your-domain.
  3. Add model provider keys only in the pipeline nodes that need them. The engine boots without any of them.
  4. In any pgvector store node, set the host to pgvector.railway.internal, the port to 5432, the user to postgres, the database to postgres, and the password to the POSTGRES_PASSWORD value from the pgvector service.
  5. Expect the first run of each node to take a minute, because the node installs its own Python packages.
  6. Keep one replica of each service. The health check hits /version, which is public.
Health check: GET /version (timeout 300 seconds)

How to self host RocketRide on other VPS Services (RocketRide self hosting guide)

Clone the Repository

Clone the packaging repository on your VPS.

git clone https://github.com/onepush-dev/rocketride
cd rocketride

Install Dependencies

Install Docker and Docker Compose on the host.

Configure Environment Variables

Set an API key and a database password before you start.

export ROCKETRIDE_APIKEY=replace-with-a-long-random-key
export POSTGRES_PASSWORD=replace-with-a-strong-password

Start the RocketRide Application

Start the engine with a persistent data directory and the API key, and add a pgvector container if pipelines need a vector store.

docker compose up -d
curl http://your-domain:8080/version

Railway handles the public domain, TLS, volumes and health checks for you.

Pricing and License of RocketRide (RocketRide pricing)

RocketRide is MIT-licensed open-source software, and you can self-host it at no licence cost.

RocketRide hosted vs self hosted comparison (Pricing, features, costs, and more)

This template packages the MIT engine (server-v3.6.0).

Monthly cost of self hosting RocketRide on Railway

Railway bills on usage, so cost depends on RAM, CPU, volume size and other services in the project.

System Requirements for Hosting RocketRide on a VPS

A VPS needs enough memory for the engine, Python node environments and pgvector, plus disk for the data volume.

Frequently Asked Questions (FAQs)

How do I sign in to my RocketRide instance?

Copy ROCKETRIDE_APIKEY from the rocketride service's Variables tab and enter it when the web UI asks for it.

Is /version protected by the API key?

No. /version is public and is what the health check uses. /ping, /task and the /task/service WebSocket all require the key.

Do I need a model provider key to boot the engine?

No.

How do I use the bundled pgvector database?

In any pgvector store node, set the host to pgvector.railway.internal, the port to 5432, the user and database to postgres, and the password to the POSTGRES_PASSWORD value from the pgvector service.

Are Milvus and ChromaDB included in this template?

No.

What should I back up?

Back up the volume mounted at /opt/data on the rocketride service and the pgvector database volume.


Template Content

More templates in this category

View Template
Chat Chat
Chat Chat, your own unified chat and search to AI platform.

okisdev
117
View Template
Weaviate
Vector database with hybrid search and an API key-protected API

GitHub Locker
1
View Template
Möbius Personal AI OS
Self-hosted AI agent that builds and runs apps for you, with memory.

bento
1