Deploy and Host R2R on Railway

Deploy R2R on Railway to run your own production-ready RAG engine with a RESTful API, hybrid search, knowledge graphs, and a management dashboard. Self-host R2R and keep full control of your data and retrieval pipeline.

This template pre-configures the R2R API server (sciphiai/r2r:latest), R2R Dashboard (sciphiai/r2r-dashboard:1.0.3), PostgreSQL with pgvector (pgvector/pgvector:pg16), and a graph-clustering service (ragtoriches/cluster-prod). Internal services communicate over Railway's private network.

Getting Started with R2R on Railway

After deploying, open the R2R Dashboard URL. Sign in with the default credentials from your environment variables (NEXT_PUBLIC_R2R_DEFAULT_EMAIL / NEXT_PUBLIC_R2R_DEFAULT_PASSWORD). Add your LLM provider API key (OPENAI_API_KEY or ANTHROPIC_API_KEY) on the R2R service in Railway. Upload your first document via the dashboard or the REST API:

curl -X POST https://your-r2r-domain.up.railway.app/v3/documents \
  -H "Content-Type: multipart/form-data" \
  -F "[email protected]"

Query your ingested documents with a RAG request:

curl -X POST https://your-r2r-domain.up.railway.app/v3/retrieval/rag \
  -H "Content-Type: application/json" \
  -d '{"query": "What are the key findings?", "search_settings": {"use_hybrid_search": true}}'

About Hosting R2R

R2R (Retrieval to Response) is an open-source (MIT) agentic RAG engine by SciPhi with 18k+ GitHub stars, backed by Y Combinator. It provides a complete API-first platform for production-grade retrieval.

Key features:

• Multimodal ingestion -- PDFs, Markdown, images, audio, and JSON
• Hybrid search -- semantic vector search plus keyword search with reciprocal rank fusion
• Knowledge graphs -- automatic entity and relationship extraction via Graph RAG
• Deep Research API -- multi-step reasoning across your knowledgebase and the web
• User and document management -- built-in auth, collections, per-user scoping
• RESTful API -- Python and TypeScript SDKs included

Why Deploy R2R on Railway

• One-click deploy with Postgres (pgvector), dashboard, and graph clustering pre-wired
• Private network keeps database and clustering services unexposed
• All configuration via environment variables -- no YAML or config files to manage
• Bring your own LLM provider key (OpenAI, Anthropic, or others)
• MIT license with no usage restrictions

Common Use Cases for Self-Hosted R2R

• Internal knowledge base -- ingest company docs and query them with natural language
• AI-powered product search -- add semantic search to your SaaS using R2R's retrieval endpoints
• Document Q&A chatbot -- power a customer-facing chatbot grounded in your documentation
• Research automation -- run multi-step queries across large document collections

Dependencies for R2R on Railway

This template deploys four services:

• R2R -- sciphiai/r2r:latest -- main RAG API server (Python, FastAPI)
• R2R-Dashboard -- sciphiai/r2r-dashboard:1.0.3 -- Next.js management UI
• Postgres -- pgvector/pgvector:pg16 -- PostgreSQL with pgvector for vector embeddings
• graph-clustering -- ragtoriches/cluster-prod -- knowledge graph clustering service

Environment Variables Reference for R2R

Deployment Dependencies for R2R

• Runtime: Python 3.10+, Node.js (Next.js dashboard)
• GitHub: SciPhi-AI/R2R
• Docker Hub: sciphiai/r2r, sciphiai/r2r-dashboard
• Docs: r2r-docs.sciphi.ai

Hardware Requirements for Self-Hosting R2R

R2R delegates inference to external LLM providers, so no GPU is needed. On Railway's 8 GB plan, all four services run comfortably.

Self-Hosting R2R with Docker

Clone the repository and run all services with Docker Compose:

git clone https://github.com/SciPhi-AI/R2R.git
cd R2R
docker compose -f compose.full.yaml --profile postgres up -d

Or run the core services individually with Docker:

docker run -d --name r2r-postgres \
  -e POSTGRES_USER=postgres -e POSTGRES_PASSWORD=changeme \
  -e POSTGRES_DB=postgres pgvector/pgvector:pg16

docker run -d --name r2r -p 7272:7272 \
  -e R2R_HOST=0.0.0.0 -e R2R_PORT=7272 \
  -e R2R_POSTGRES_HOST=r2r-postgres \
  -e R2R_POSTGRES_USER=postgres \
  -e R2R_POSTGRES_PASSWORD=changeme \
  sciphiai/r2r:latest

Access the API at http://localhost:7272.

Is R2R Free to Self-Host on Railway?

R2R is fully open-source under the MIT license. SciPhi offers a managed cloud with a free tier (300 RAG requests/month) and enterprise plans. Self-hosting on Railway means you pay only for infrastructure plus your LLM provider API costs.

FAQ: Self-Hosting R2R on Railway

What is R2R and why should I self-host it? R2R is an open-source RAG engine by SciPhi providing hybrid search, knowledge graphs, and document management via a REST API. Self-hosting gives you full data control and lets you use any LLM provider.

What does this Railway template deploy for R2R? Four services: the R2R API server, a Next.js dashboard, PostgreSQL with pgvector for vector embeddings, and a graph-clustering service. Database and clustering run on Railway's private network.

Why does R2R need PostgreSQL with pgvector on Railway? R2R stores vector embeddings in PostgreSQL using pgvector. Railway's managed Postgres lacks this extension, so the template uses pgvector/pgvector:pg16 with a dedicated volume.

How do I add my OpenAI or Anthropic API key to R2R on Railway? Open the R2R service in Railway, go to Variables, and add OPENAI_API_KEY or ANTHROPIC_API_KEY. Redeploy the service. R2R supports multiple providers simultaneously.

Can I use R2R without an LLM API key on Railway? The API server starts and accepts document uploads without an LLM key. RAG queries require at least one provider key, but hybrid search and document management work independently.

How do I scale R2R for larger document collections on Railway? Increase R2R_POSTGRES_MAX_CONNECTIONS if you see pool exhaustion. For heavier workloads, scale the R2R service vertically in Railway (more CPU/RAM).