Deploy and Host Weaviate on Railway

This Railway template deploys a Weaviate vector database — ready for semantic search, RAG pipelines, and AI-powered apps — with a single click.

Getting Started with Weaviate on Railway

Once your Railway deploy is live, confirm the instance is healthy by visiting https://your-url.railway.app/v1/.well-known/ready — it should return {}. Your API key is auto-generated during deployment; copy it from the AUTHENTICATION_APIKEY_ALLOWED_KEYS variable in your Railway service environment tab.

Connect using the official Python client:

import weaviate

client = weaviate.connect_to_custom(
    http_host="your-url.railway.app",
    http_port=443,
    http_secure=True,
    grpc_host="your-url.railway.app",
    grpc_port=443,
    grpc_secure=True,
    auth_credentials=weaviate.auth.AuthApiKey("your-api-key"),
)

# Create a collection and insert an object
collection = client.collections.create("Article")
collection.data.insert({"title": "Hello Weaviate", "body": "First object!"})

Or connect via the JavaScript client: npm install weaviate-client, then use weaviate.connectToCustom(...) with the same host and API key.

About Hosting Weaviate

Weaviate is an open-source vector database built for AI and machine learning workloads. It stores data objects alongside their vector embeddings, enabling semantic search, Retrieval-Augmented Generation (RAG), and recommendation systems without complex integration layers.

Key features:

• Hybrid search — combine vector similarity search with BM25 keyword search using a single alpha parameter
• GraphQL + REST + gRPC APIs — query your data from any language or toolchain
• Schema-based data model — define collections with typed properties for rich filtering alongside vector search
• Built-in vectorizer modules — optionally let Weaviate call OpenAI, Cohere, or Hugging Face to generate embeddings at import time
• RBAC + API key auth — fine-grained access control out of the box
• HNSW indexing — fast approximate nearest-neighbour search that scales logarithmically

This template ships with DEFAULT_VECTORIZER_MODULE="none", meaning you bring your own vectors. This keeps the setup self-contained and works with any embedding model or provider. See the environment variable section below for how to add an external vectorizer.

Why Deploy Weaviate on Railway

Running Weaviate on a raw VPS means managing Docker, configuring volumes, handling TLS, setting up restart policies, and monitoring disk pressure. Railway eliminates all of that. The volume is attached and mounted automatically; environment variables are managed in a UI with secret generation built in; and redeploys are zero-downtime.

Common Use Cases

• RAG pipelines — store chunked document embeddings and retrieve the top-k most relevant chunks at query time to ground LLM responses in your own data
• Semantic search — replace keyword search with meaning-based search over product catalogues, knowledge bases, or support articles
• Recommendation engines — find similar items (articles, products, users) by storing entity embeddings and querying by vector proximity
• Multi-modal AI apps — store and search across text and image embeddings in the same collection using Weaviate's multi-vector support

Dependencies for Weaviate

• Weaviate DB — cr.weaviate.io/semitechnologies/weaviate:1.36.2 (also available as semitechnologies/weaviate on Docker Hub)

Adding an Embedder (Optional)

This template ships without a vectorizer — you provide vectors yourself. To let Weaviate auto-generate embeddings at import time, update these variables:

OpenAI:

DEFAULT_VECTORIZER_MODULE="text2vec-openai"
ENABLE_MODULES="text2vec-openai"
OPENAI_APIKEY="sk-..."

Cohere:

DEFAULT_VECTORIZER_MODULE="text2vec-cohere"
ENABLE_MODULES="text2vec-cohere"
COHERE_APIKEY="..."

AWS Bedrock:

DEFAULT_VECTORIZER_MODULE="text2vec-aws"
ENABLE_MODULES="text2vec-aws"
AWS_ACCESS_KEY="..."
AWS_SECRET_KEY="..."

After updating the variables, redeploy the service and specify the vectorizer in your collection schema at creation time.

Deployment Dependencies

• Docker image: cr.weaviate.io/semitechnologies/weaviate
• Official docs: docs.weaviate.io
• Client libraries: Python (pip install weaviate-client), JS/TS (npm install weaviate-client), Go, Java

Minimum Hardware Requirements for Weaviate

Weaviate's memory usage is directly driven by dataset size — the HNSW index must fit in RAM. CPU determines query and import speed; more cores = more concurrent queries.

RAM estimate: roughly 3 GB per 1 million 384-dimension vectors. For 768-dimension vectors (typical for OpenAI text-embedding-3-small), plan ~6 GB per million objects. On Railway, scale your service's memory allocation in Settings → Resources as your dataset grows.

Self-Hosting Weaviate

To run Weaviate outside Railway, on any Linux host with Docker:

# docker-compose.yml
services:
  weaviate:
    image: cr.weaviate.io/semitechnologies/weaviate:1.36.2
    ports:
      - "8080:8080"
      - "50051:50051"
    volumes:
      - weaviate_data:/var/lib/weaviate
    environment:
      PERSISTENCE_DATA_PATH: "/var/lib/weaviate"
      CLUSTER_HOSTNAME: "node1"
      AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED: "false"
      AUTHENTICATION_APIKEY_ENABLED: "true"
      AUTHENTICATION_APIKEY_ALLOWED_KEYS: "your-secret-key"
      AUTHENTICATION_APIKEY_USERS: "admin"
      AUTHORIZATION_ENABLE_RBAC: "true"
      AUTHORIZATION_RBAC_ROOT_USERS: "admin"
      DEFAULT_VECTORIZER_MODULE: "none"
      QUERY_DEFAULTS_LIMIT: "25"

volumes:
  weaviate_data:

docker compose up -d
# Weaviate is now available at http://localhost:8080

For Kubernetes deployments at scale, use the official Helm chart.

Is Weaviate Free?

Weaviate is open-source under the BSD-3 licence — the database itself is completely free to self-host. On Railway, your only cost is infrastructure (compute + storage). The managed cloud offering, Weaviate Cloud, has a free sandbox tier (30-day expiry) and paid tiers starting at $45/month (Flex) based on vector dimensions stored. Self-hosting on Railway is typically the most cost-effective option for teams who want persistent, production-grade infrastructure without per-query or per-dimension fees.

Weaviate vs Pinecone vs Qdrant

Weaviate is the best choice when your data has rich relationships, you want plug-in embedding modules, or your team is already familiar with GraphQL. Qdrant edges it out on raw throughput; Pinecone is the lowest-ops option if you don't need self-hosting.

FAQ

What is Weaviate? Weaviate is an open-source vector database that stores data objects alongside their vector embeddings. It supports semantic search, hybrid search, RAG, and recommendation systems via REST, GraphQL, and gRPC APIs.

What does this Railway template deploy? A single Weaviate service running the official semitechnologies/weaviate Docker image, with API key authentication enabled, RBAC configured, and a Railway persistent volume mounted at /var/lib/weaviate. Your API key is auto-generated at deploy time.

Why does the template use RAILWAY_RUN_UID="0"? Railway runs containers as a non-root user by default. Weaviate needs root permissions to write to its data directory. Without this variable set to 0, the service crashes on startup with a permission error when trying to create files in /var/lib/weaviate.

Does Weaviate support GraphQL? Yes. Weaviate exposes a GraphQL endpoint at /v1/graphql for querying objects, running semantic search (nearVector, nearText), and applying filters. The REST API at /v1 handles schema management and CRUD operations.

Can I use this Weaviate template in production? Yes, with some care. The template is configured with API key auth and RBAC enabled by default. For production, pin a specific image version (e.g. 1.36.2 rather than latest), monitor Railway's memory usage as your dataset grows, and ensure RAILWAY_DEPLOYMENT_DRAINING_SECONDS is set so Weaviate can flush data to disk cleanly before restarts.

How do I add an embedding model like OpenAI? Set DEFAULT_VECTORIZER_MODULE="text2vec-openai", ENABLE_MODULES="text2vec-openai", and OPENAI_APIKEY="sk-..." in your Railway environment variables, then redeploy. Weaviate will call the OpenAI Embeddings API automatically when you insert objects into collections configured with that vectorizer.

Can I browse my data without writing code? Yes. Open console.weaviate.cloud in your browser and connect it to your Railway instance URL with your API key. This gives you a full GUI for browsing collections, running queries, and inspecting objects — no Weaviate Cloud subscription required.