DataFlow Web

Web UI and server for managing DataFlow Operator: view and edit DataFlow manifests, processor logs, and metrics.

Structure

• web/ — Vue 3 + Vite frontend (dashboard, manifests, logs, metrics). Build: cd web && npm install && npm run build; output in web/dist/, copied into the Docker image as /app/static.
• cmd/server/ — GUI server entrypoint.
• internal/gui/ — HTTP server, static file serving and API (/api/dataflows, /api/logs, /api/status, /api/namespaces).

The image is built separately from the operator image. On docker build, the frontend (Node) is built first, then the Go server; static assets come from the Vue build.

Building the image

The github.com/dataflow-operator/dataflow dependency is fetched from GitHub (main/master branch), no local replace. Build from the dataflow-web directory:

# from dataflow-web directory
cd dataflow-web
docker build -t dataflow-web:local .

# or from repo root
docker build -f dataflow-web/Dockerfile -t dataflow-web:local dataflow-web

Image contents: binary /app/server, static files in /app/static. Server listens on port 8080.

Taskfile (local build and dev)

The repo includes a Taskfile. After installing Task (go install github.com/go-task/task/v3/cmd/task@latest or via brew/apt):

task install   # Go deps + npm in web/
task build     # build frontend + static/ + server binary
task run       # run ./server (after build)
task dev       # dev mode: Go server + Vite dev in parallel
task web:test  # frontend tests
task docker    # build Docker image
task --list    # list tasks

Running locally

Option 1 — Go only (requires pre-built static):

cd dataflow-web
# Build frontend and copy to static (or use the image)
cd web && npm install && npm run build && cp -r dist/* ../static/ 2>/dev/null || true
cd ..
go run ./cmd/server --bind-address=:8080

Option 2 — frontend dev with hot reload:

# Terminal 1: backend
cd dataflow-web && go run ./cmd/server --bind-address=:8080

# Terminal 2: frontend (proxy to :8080)
cd dataflow-web/web && npm install && npm run dev

Open http://localhost:8080 (option 1) or http://localhost:5173 (option 2). Cluster access requires KUBECONFIG or in-cluster config.

Deploying to cluster

Deploy via Helm with gui.enabled=true. See Web GUI documentation for configuration and deployment details.