Sepex Viewer

Prototype Streamlit dashboard for Sepex

Jobs overview with per-job details

Quick Start

1. Environment variables

copy .env.example .env

The defaults work for local development (postgres + minio + auth disabled).

2. Create the shared docker network

The compose file expects an external network named process_api_net. Create it once per machine:

docker network create process_api_net

3. Start the stack

docker compose up -d

The viewer is at http://localhost:8501. First boot pulls the sepex API image and initializes postgres + minio.

---

Docker Commands

Stop the application:

docker compose down

Start the application:

docker compose up -d

Rebuild the viewer image:

docker compose build viewer

Rebuild and restart:

docker compose up -d --build

View logs:

docker compose logs -f viewer

Clear all data (postgres, minio, api state):

docker compose down
rmdir /s /q .data

Prune volumes:

docker system prune -a --volumes

---

Services

Service	Image	Purpose
viewer (app/)	Streamlit, built locally	Dashboard UI for sepex
sepex	ghcr.io/dewberry/sepex/api	Process execution API
postgres	postgres:17.2-alpine3.20	Job and process metadata
minio	quay.io/minio/minio	S3-compatible object storage

The viewer hits the sepex API over HTTP (SEPEX_BASE_URL). The API uses postgres for job state and minio for storage. To run against AWS S3 instead, set STORAGE_SERVICE=aws-s3 in .env.

The viewer container mounts the repo at /app, so changes under app/ hot-reload without a rebuild.

Access Points

• Viewer: http://localhost:8501
• sepex API: http://localhost:5050
• sepex API docs (Swagger): http://localhost:5050/swagger/index.html
• MinIO console: http://localhost:9001 (user / password from .env)
• Postgres: localhost:5432

Configuration toggles (db, storage, auth) are documented inline in .env.example.

Registering Processes

The sepex API registers processes from yaml files at startup. Point PLUGINS_LOAD_DIR in .env at a directory of yaml configs and restart the sepex container — every .yaml / .yml in there is read and registered.

examples/register-processes/pyecho.yaml is a sample config — a Python plugin that echoes its input. Drop it (and any others) into your plugins directory, restart, and the processes show up in the Processes column of the viewer.