Deploy Arize Phoenix | (Just Updated) LLM Tracing & Evals, Data Survives Redeploys
LLM tracing and evals, traces kept on a volume
phoenix
Just deployed
/root/.phoenix
Deploy and Host Arize Phoenix on Railway
Arize Phoenix is an open-source AI observability platform for LLM applications. It collects OpenTelemetry traces from your app, shows every prompt, retrieval and tool call in a trace view, and runs evaluations, datasets and experiments so you can compare prompt and model changes.
This template runs Phoenix 20.21.1 as a single service from a digest-pinned official image, with the web UI and the OTLP/HTTP collector on one HTTPS domain and the working directory on a Railway volume.
About Hosting Arize Phoenix
Phoenix needs a few things handled before it keeps your data on Railway:
- Data survives redeploys. Traces, datasets and experiments are stored in a SQLite database under
/root/.phoenix. The template mounts a Railway volume there, so redeploying does not wipe your history. - One URL for the UI and the collector. Port 6006 serves the web UI and the OTLP/HTTP endpoint
(
/v1/traces), so you point your instrumented app at the same domain you open in the browser. - Pinned image. The official image is pinned by digest, so a redeploy does not silently pull a different build.
- Authentication is off by default, as in upstream. Anyone with the URL can view and write traces. Add
PHOENIX_ENABLE_AUTH=trueandPHOENIX_SECRET(a long random string) in the service variables to turn on logins before sending real data.
Common Use Cases
- Tracing LLM apps built with OpenAI, Anthropic, LangChain, LlamaIndex or the Vercel AI SDK through OpenInference
- Debugging RAG pipelines by inspecting retrieved documents next to the generated answer
- Running evals and experiments on datasets to compare prompts and models
- Keeping a shared prompt and trace history for a team without a paid observability SaaS
Dependencies for Arize Phoenix Hosting
- One Railway volume for the Phoenix working directory (created by the template)
Deployment Dependencies
- Phoenix documentation: https://arize.com/docs/phoenix
- Official image: https://hub.docker.com/r/arizephoenix/phoenix
- Source: https://github.com/Arize-ai/phoenix
Implementation Details
Point an OpenTelemetry exporter at the deployed domain:
PHOENIX_COLLECTOR_ENDPOINT=https://
Traces are accepted at https:///v1/traces over OTLP/HTTP. The template ships no required
variables, so the deploy form is empty.
Why Deploy Arize Phoenix 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 Arize Phoenix 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.
Template Content
phoenix
arizephoenix/phoenix:20.21.1