Configuration

Settings come from environment variables with the prefix ONEOPEN_ (via pydantic-settings).

Environment variables

Variable	Default	Description
ONEOPEN_DATA_DIR	~/.oneopen	Root data directory
ONEOPEN_HOST	127.0.0.1	Default bind host for serve
ONEOPEN_PORT	8765	Default bind port
ONEOPEN_DEBUG	false	Debug flag
ONEOPEN_APP_NAME	OneOpen ML Studio	Application display name

Examples

Windows (PowerShell):

$env:ONEOPEN_DATA_DIR = "D:\oneopen-data"
$env:ONEOPEN_PORT = "9000"
oneopen serve

Unix:

export ONEOPEN_DATA_DIR=/data/oneopen
export ONEOPEN_PORT=9000
oneopen serve

CLI flags (--host, --port, init --data-dir) override the corresponding defaults for that process.

Derived paths

Given data_dir:

Path	Purpose
{data_dir}/oneopen.db	SQLite database
{data_dir}/uploads/	Fallback upload area
{data_dir}/exports/	Export artifacts
{data_dir}/models/	Weights, SAM checkpoints, runs
{data_dir}/projects/{id}/	Default project location
{data_dir}/training_jobs/	Training job JSON state

See Data layout for more detail.

Binding beyond localhost

You can pass --host 0.0.0.0, but the product is designed as a local tool without authentication. Do not expose it on untrusted networks.