Deploy and Host Local Deep Research on Railway

Local Deep Research (LDR) is an open-source AI research assistant: give it a question and it runs multi-step web research across many sources and writes a report with citations, using the LLM of your choice and a private search engine. This is a community-maintained template; it is not affiliated with the Local Deep Research or SearXNG projects.

About Hosting Local Deep Research

LDR is a multi-user application that encrypts each user's data with their own password, and it needs a search engine (SearXNG) alongside it, with SearXNG's JSON API enabled. Out of the box LDR's user registration is open and, behind a TLS-terminating proxy, it needs the right forwarded headers or its real-time features break.

This template runs the whole thing on Railway as three services behind one password: LDR itself, a private SearXNG with the JSON API turned on, and a Caddy front-door that adds HTTP basic authentication so the instance is private by default and forwards the browser's HTTPS scheme so LDR's WebSocket progress streaming works. Bring your own LLM key (OpenAI, Anthropic, OpenRouter, or any OpenAI-compatible endpoint), configured inside LDR after you sign in.

Common Use Cases

A private deep-research assistant for your team, behind a shared password.
Cited literature and web research reports generated with your own LLM key.
A self-hosted alternative to hosted "deep research" tools, with your data encrypted per user.

Dependencies for Local Deep Research Hosting

An LLM: OpenAI, Anthropic, OpenRouter, or any OpenAI-compatible endpoint (configured in LDR's settings).
Nothing else — SearXNG is bundled and private.

Deployment Dependencies

Local Deep Research: https://github.com/LearningCircuit/local-deep-research (MIT)
SearXNG: https://github.com/searxng/searxng (AGPL-3.0)
Caddy: https://caddyserver.com (Apache-2.0)
Template repository, images and tests: https://github.com/youssefsiam38/ldr-railway

Implementation Details

LDR runs upstream's own published image, pinned by digest (its maintainers already build non-root, capability-dropped, with pinned dependencies). SearXNG runs the stock image with only its JSON output format and secret configured. A Caddy front-door adds HTTP basic authentication over everything except the health check, routes to LDR, and forces the HTTPS scheme LDR needs for secure cookies and its WebSocket same-origin check. Ollama (local models) is omitted as impractical on CPU instances.

Tested in CI and on a live deployment of this template: all three services start, authentication gates the app, LDR is reachable through the front-door, SearXNG's JSON API answers, the Socket.IO WebSocket connects through basic auth, and a redeploy keeps the data volume.

The deploy form generates OWNER_PASSWORD (front-door username owner). Copy it from the caddy service's variables, open the domain, then create your LDR account and add an LLM key inside the app.

Why Deploy Local Deep Research 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 Local Deep Research 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.