Deploy and Host LangWatch on Railway

Run LangWatch agent testing and LLM observability with durable traces, datasets, evaluations, and background workers. This template deploys the pinned community release with PostgreSQL, Redis, ClickHouse analytics, and Railway object storage.

About Hosting LangWatch

LangWatch receives application traces, turns production examples into datasets, and runs evaluators and agent tests. Railway hosts the web application and workers alongside private persistent data services, so the deployment survives restarts without placing databases on the public network.

Why Deploy LangWatch on Railway

The template generates its secrets, wires private service discovery, provisions three explicit volumes and a Bucket, and publishes only the authenticated LangWatch application. That removes the manual Compose translation while preserving the upstream service boundaries.

Common Use Cases

Inspect OpenTelemetry and SDK traces from LLM applications.
Build datasets from real interactions and run regression evaluations.
Test agent behavior with dedicated NLP and evaluator workers.
Keep prompts, traces, scores, and artifacts in a private deployment.

Dependencies for LangWatch Hosting

Deployment Dependencies

LangWatch 3.9.0 application, workers, NLP engine, and evaluators
PostgreSQL for accounts, projects, and configuration
Redis for asynchronous jobs
ClickHouse for trace and evaluation analytics
Railway Bucket for datasets and stored objects

The deployment is a single-node small-team baseline rather than a highly available cluster. Configure model-provider credentials after first login when an evaluator requires them.