Graph Viewer Deployment

Production Server

The graph viewer ships with server.mjs — a lightweight Node.js HTTP server that serves the Vite-built static files from dist/. It handles:

Content-type detection for all common web assets
Immutable cache headers (max-age=31536000) on /assets/* (Vite hashed filenames)
no-cache for HTML and non-hashed files
SPA fallback — unknown routes serve index.html for client-side routing
HEAD request support
Path traversal protection

The server listens on 0.0.0.0:3000 by default. Override with HOST and PORT environment variables.

Docker

The included multi-stage Dockerfile builds and serves the viewer:

FROM node:20-slim AS builder
WORKDIR /app
COPY package*.json ./
RUN npm ci
COPY . .
RUN npm run build

FROM node:20-slim
WORKDIR /app
COPY --from=builder /app/dist ./dist
COPY --from=builder /app/server.mjs ./
COPY --from=builder /app/package*.json ./
ENV HOST=0.0.0.0
EXPOSE 3000
CMD ["node", "server.mjs"]

Build and run:

docker build -t automem-graph-viewer .
docker run -p 3000:3000 automem-graph-viewer

If you run the viewer locally alongside the AutoMem Docker stack, note that FalkorDB’s built-in browser also defaults to port 3000. Set PORT=3001 (or similar) on one service to avoid a collision.

Railway

Recommended: deploy the published image. Point your Railway service at the public GHCR image instead of building from source:

ghcr.io/verygoodplugins/automem-graph-viewer:stable

This avoids spending Railway compute rebuilding the frontend on every deploy. The viewer service typically needs no custom environment variables in production.

After the viewer’s domain is created, set these variables on the automem API service (not the viewer service) so browser traffic can reach it:

GRAPH_VIEWER_URL=https://<viewer-domain>
VIEWER_ALLOWED_ORIGINS=https://<viewer-domain>

Browser-to-API traffic must use the API’s public domain — Railway private domains aren’t reachable from user browsers. The viewer itself stores only browser-side config (server URL, token) and should never receive database credentials.

Recommended Railway layout

Service	Source	Public?	Notes
`automem`	`ghcr.io/verygoodplugins/automem:stable`	Yes	Auth, API routes, data access, `/viewer/*` bootstrap
`automem-graph-viewer`	`ghcr.io/verygoodplugins/automem-graph-viewer:stable`	Yes	Static React/Three.js UI; no DB secrets
`mcp-automem`	`ghcr.io/verygoodplugins/mcp-automem:stable`	Yes	Remote MCP bridge for hosted clients
`falkordb`	`falkordb/falkordb:latest`	No	Private graph database
`qdrant`	`qdrant/qdrant:latest` or Qdrant Cloud	No/External	Vector store

Keep the standalone viewer service served at /. Only set VITE_BASE_PATH=/viewer/ when the built assets are actually served from the AutoMem API origin under /viewer/.

Source-linked deploys. The repository’s railway.toml is kept for preview/source deployments and mirrors the Dockerfile used to publish the GHCR image:

[build]
builder = "DOCKERFILE"
dockerfilePath = "Dockerfile"

[deploy]
healthcheckPath = "/"
restartPolicyType = "ON_FAILURE"
restartPolicyMaxRetries = 5

To deploy from source:

Push the repository to GitHub
Connect it to a Railway project
Railway builds the Dockerfile and deploys automatically on push

The PORT environment variable is set automatically by Railway. Configure the API server URL via the viewer’s settings UI or localStorage after deployment.

Static Hosting

Since the viewer is a static SPA after build, you can also host it on any static file server (Nginx, Cloudflare Pages, Vercel, etc.):

npm run build
# Upload dist/ to your hosting provider

Requirements for the hosting provider:

Serve index.html for all unknown routes (SPA fallback)
Set appropriate cache headers on /assets/*

The viewer will use relative URLs by default. Users configure the AutoMem API URL through the settings panel on first visit.