Deploy and Host Latitude on Railway

Latitude is the open-source observability platform for AI agents and LLMs — "Sentry, but for agents." It groups failing traces into tracked issues, runs human-aligned evaluations, and makes every trace semantically searchable. Send OpenTelemetry traces from any model provider or framework and see what will break next before your users do.

About Hosting Latitude

This template deploys the complete Latitude stack: five application services (web UI, public API / MCP server, OTLP trace ingestion, background workers, and Temporal workflow workers), a one-shot database-migrations job, and all backing infrastructure — Postgres with pgvector, ClickHouse, two Redis instances (cache + durable queue), a Postgres-backed Temporal server, and a SeaweedFS object store. Everything runs from Latitude's published Docker images over Railway's private network. On the first deploy the services initialize the databases and Temporal schema and then settle. The encryption keys and infrastructure passwords are generated for you; you add an email transport and optional AI provider keys.

Common Use Cases

• Monitor and debug production AI agents — failing traces grouped into tracked issues with status, size, and trend, plus alerts when new issues appear or escalate.
• Evaluate and improve LLM output quality with human-aligned evals and an alignment score that tracks drift from human judgment over time.
• Explore agent traces by meaning with semantic search across 100% of traffic, and query everything from your coding agent through the built-in MCP server.

Dependencies for Latitude Hosting

• PostgreSQL (with the pgvector extension) — primary store
• ClickHouse — span/telemetry OLAP store
• Redis ×2 — cache + durable BullMQ queue
• Temporal — workflow engine (Postgres-backed, no Elasticsearch/Cassandra)
• SeaweedFS — S3-compatible object storage

All of the above are bundled in this template. Optional external dependencies you provide: an email transport (SMTP, Mailgun, or SendGrid) for magic-link sign-in, and AI provider keys (e.g. Voyage, OpenAI, Amazon Bedrock) to enable AI features.

Deployment Dependencies

• Latitude documentation
• Self-host guide
• Configuration reference
• Published Docker images (Docker Hub)
• Source on GitHub

Implementation Details

After the stack is up:

1. Add an email transport on the web, api, and ingest services (e.g. LAT_MAILGUN_* or LAT_SMTP_*) — magic-link sign-in needs it.
2. Open the web service's public URL and register; the first account creates your organization.
3. (Optional) Add AI provider keys to turn on AI features — LAT_VOYAGE_API_KEY (default embeddings + reranking), plus a generation provider such as LAT_OPENAI_API_KEY with LAT_AI_GENERATION_PROVIDER / LAT_AI_GENERATION_MODEL. Core trace ingestion and viewing work without any keys.

Why Deploy Latitude on Railway?

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 Latitude 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.