Web UI REST API Reference¶

The srunx Web UI exposes a REST API at http://127.0.0.1:8000/api/. All responses are JSON.

Jobs¶

Method	Endpoint	Description
GET	`/api/jobs`	List all SLURM jobs in the queue
GET	`/api/jobs/{job_id}`	Get detailed status for a specific job
POST	`/api/jobs`	Submit a new job
DELETE	`/api/jobs/{job_id}`	Cancel a running or pending job
GET	`/api/jobs/{job_id}/logs`	Get stdout/stderr log contents

GET /api/jobs¶

Returns a list of all jobs from squeue.

Response:

[
  {
    "name": "train-resnet",
    "job_id": 18431,
    "status": "RUNNING",
    "depends_on": [],
    "command": [],
    "resources": {
      "nodes": 1,
      "gpus_per_node": 8,
      "partition": "gpu",
      "time_limit": "8:00:00"
    },
    "partition": "gpu",
    "nodes": 1,
    "gpus": 8,
    "elapsed_time": "1:30:00"
  }
]

POST /api/jobs¶

Submit a new job with a SLURM script.

Request body:

{
  "name": "my-job",
  "script_content": "#!/bin/bash\n#SBATCH --gpus=1\npython train.py",
  "job_name": "training-run"
}

Response (201):

{
  "name": "my-job",
  "job_id": 18500,
  "status": "PENDING",
  "depends_on": [],
  "command": [],
  "resources": {}
}

DELETE /api/jobs/{job_id}¶

Cancel a job. Returns 204 No Content on success.

GET /api/jobs/{job_id}/logs¶

Response:

{
  "stdout": "Epoch 1/10: loss=0.85...",
  "stderr": "WARNING: GPU memory high"
}

Resources¶

Method	Endpoint	Description
GET	`/api/resources`	Get GPU and node availability per partition

GET /api/resources¶

Query parameters:

partition (optional) — Filter to a specific partition

Response:

[
  {
    "timestamp": "2026-03-30T09:00:00+00:00",
    "partition": "gpu",
    "total_gpus": 32,
    "gpus_in_use": 24,
    "gpus_available": 8,
    "jobs_running": 3,
    "nodes_total": 4,
    "nodes_idle": 1,
    "nodes_down": 0,
    "gpu_utilization": 0.75,
    "has_available_gpus": true
  }
]

Note

Multi-node jobs are correctly accounted for: gpus_in_use = gpus_per_node * num_nodes.

Workflows¶

Method	Endpoint	Description
GET	`/api/workflows`	List all workflow definitions
GET	`/api/workflows/{name}`	Get a specific workflow with its jobs
POST	`/api/workflows/validate`	Validate YAML content
POST	`/api/workflows/upload`	Upload and save a workflow YAML
POST	`/api/workflows/create`	Create a workflow from structured JSON (DAG builder)
DELETE	`/api/workflows/{name}`	Delete a workflow YAML file
POST	`/api/workflows/{name}/run`	Run a workflow (sync mounts, submit jobs, monitor)
GET	`/api/workflows/runs`	List workflow run records
GET	`/api/workflows/runs/{run_id}`	Get run status with live job statuses
POST	`/api/workflows/runs/{run_id}/cancel`	Cancel all jobs in a run

POST /api/workflows/upload¶

Request body:

{
  "yaml": "name: my-pipeline\njobs:\n  - name: step1\n    command: ['echo', 'hello']",
  "filename": "my-pipeline.yaml"
}

Validation rules:

Filename must be alphanumeric with hyphens/underscores only
File extension must be .yaml or .yml
Content size limit: 1MB
python: args are rejected (security)

POST /api/workflows/validate¶

Request body:

{"yaml": "name: test\njobs: []"}

Response:

{"valid": true}

// or
{"valid": false, "errors": ["Duplicate job name: step1"]}

POST /api/workflows/create¶

Create a workflow from a structured JSON payload. Used by the DAG builder.

Request body:

{
  "name": "ml-pipeline",
  "jobs": [
    {
      "name": "preprocess",
      "command": ["python", "preprocess.py"],
      "depends_on": [],
      "resources": {"nodes": 1},
      "work_dir": "/home/researcher/ml-project"
    },
    {
      "name": "train",
      "command": ["python", "train.py", "--epochs", "100"],
      "depends_on": ["preprocess"],
      "resources": {"nodes": 1, "gpus_per_node": 4, "time_limit": "4:00:00"},
      "environment": {"conda": "ml_env"}
    }
  ]
}

Job fields:

Field	Required	Description
`name`	Yes	Job name (alphanumeric, hyphens, underscores)
`command`	Yes	Command as a list of strings
`depends_on`	No	List of upstream job names, optionally prefixed with dependency type (e.g., `afternotok:preprocess`)
`resources`	No	Object with `nodes`, `gpus_per_node`, `ntasks_per_node`, `cpus_per_task`, `memory_per_node`, `time_limit`, `partition`, `nodelist`
`environment`	No	Object with `conda`, `venv`, `container`, `env_vars`
`work_dir`	No	Working directory on the remote cluster
`log_dir`	No	Log output directory
`retry`	No	Number of retry attempts
`retry_delay`	No	Delay between retries in seconds

Response (200):

{
  "name": "ml-pipeline",
  "jobs": [
    {
      "name": "preprocess",
      "job_id": null,
      "status": "UNKNOWN",
      "depends_on": [],
      "command": ["python", "preprocess.py"],
      "resources": {"nodes": 1, "gpus_per_node": null, "partition": null, "time_limit": null}
    }
  ]
}

Error responses:

409 — Workflow with the same name already exists
422 — Validation error (invalid name, duplicate job names, dependency cycle, Pydantic validation failure)

DELETE /api/workflows/{name}¶

Delete a workflow YAML file from the workflow directory.

Path parameters:

name — Workflow name (alphanumeric, hyphens, underscores)

Response (200):

{"status": "deleted", "name": "ml-pipeline"}

Error responses:

404 — Workflow not found
422 — Invalid workflow name

POST /api/workflows/{name}/run¶

Run a workflow end-to-end: identify and sync referenced mounts, render SLURM scripts, submit jobs in topological order with --dependency flags, and start a background monitor that polls sacct every 10 seconds.

Path parameters:

name — Workflow name

Response (202):

{
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "workflow_name": "ml-pipeline",
  "started_at": "2026-03-30T12:00:00+00:00",
  "completed_at": null,
  "status": "running",
  "job_ids": {
    "preprocess": "18500",
    "train": "18501",
    "evaluate": "18502"
  },
  "job_statuses": {
    "preprocess": "PENDING",
    "train": "PENDING",
    "evaluate": "PENDING"
  },
  "error": null
}

The status field transitions through: syncing, submitting, running, then a terminal state (completed, failed, or cancelled).

Error responses:

404 — Workflow not found
422 — Invalid workflow name
500 — Script rendering failed
502 — Mount sync failed or sbatch submission failed

GET /api/workflows/runs/{run_id}¶

Get the current status and job-level details for a single workflow run. Job statuses are updated by the background monitor every 10 seconds.

Path parameters:

run_id — UUID of the run (returned by the POST /{name}/run endpoint)

Response (200):

{
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "workflow_name": "ml-pipeline",
  "started_at": "2026-03-30T12:00:00+00:00",
  "completed_at": null,
  "status": "running",
  "job_ids": {"preprocess": "18500", "train": "18501"},
  "job_statuses": {"preprocess": "COMPLETED", "train": "RUNNING"},
  "error": null
}

Error responses:

404 — Run not found

POST /api/workflows/runs/{run_id}/cancel¶

Cancel all SLURM jobs associated with a workflow run. Each submitted job is cancelled via scancel. The run status is set to cancelled.

Path parameters:

run_id — UUID of the run

Response (200):

{"status": "cancelled", "run_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"}

If some jobs fail to cancel (e.g., already completed), the response includes a warnings array:

{
  "status": "cancelled",
  "run_id": "a1b2c3d4-...",
  "warnings": ["evaluate: Job 18502 not found"]
}

Error responses:

404 — Run not found
422 — Run is already in a terminal state (completed, failed, or cancelled)

Sweep Runs¶

Parameter sweeps expand a workflow into N independent cells across a matrix of axes. Each cell is a regular workflow_runs row that inherits the parent sweep's sweep_run_id; the parent sweep_runs row tracks aggregate counters (cells_pending, cells_running, cells_completed, cells_failed, cells_cancelled) that are updated atomically on every cell state transition.

Method	Endpoint	Description
GET	`/api/sweep_runs`	List all sweep runs (most recent first, capped at 200)
GET	`/api/sweep_runs/{id}`	Get one sweep run with counters and matrix spec
GET	`/api/sweep_runs/{id}/cells`	List every cell (workflow_run) linked to the sweep
POST	`/api/sweep_runs/{id}/cancel`	Request sweep cancellation (drain pending + scancel running)

Per-cell cancellation reuses the existing workflow-run endpoint — POST /api/workflows/runs/{cell_id}/cancel, where cell_id is the workflow_runs.id returned by GET /api/sweep_runs/{id}/cells.

GET /api/sweep_runs¶

List every sweep run. Returns an array ordered by started_at descending, bounded to 200 rows.

Response (200):

[
  {
    "id": 42,
    "name": "train-lr-seed-sweep",
    "workflow_yaml_path": "/workflows/train.yaml",
    "status": "running",
    "matrix": {"lr": [0.001, 0.01, 0.1], "seed": [1, 2, 3]},
    "args": {"dataset": "cifar10"},
    "fail_fast": false,
    "max_parallel": 4,
    "cell_count": 9,
    "cells_pending": 3,
    "cells_running": 2,
    "cells_completed": 4,
    "cells_failed": 0,
    "cells_cancelled": 0,
    "submission_source": "web",
    "started_at": "2026-04-22T09:00:00+00:00",
    "completed_at": null,
    "cancel_requested_at": null,
    "error": null
  }
]

GET /api/sweep_runs/{id}¶

Get the full row for a single sweep run. Fields mirror the list response.

Path parameters:

id — Sweep run ID (integer)

Example:

GET /api/sweep_runs/42

Response (200):

{
  "id": 42,
  "name": "train-lr-seed-sweep",
  "workflow_yaml_path": "/workflows/train.yaml",
  "status": "completed",
  "matrix": {"lr": [0.001, 0.01, 0.1], "seed": [1, 2, 3]},
  "args": {"dataset": "cifar10"},
  "fail_fast": false,
  "max_parallel": 4,
  "cell_count": 9,
  "cells_pending": 0,
  "cells_running": 0,
  "cells_completed": 9,
  "cells_failed": 0,
  "cells_cancelled": 0,
  "submission_source": "web",
  "started_at": "2026-04-22T09:00:00+00:00",
  "completed_at": "2026-04-22T10:15:00+00:00",
  "cancel_requested_at": null,
  "error": null
}

Error responses:

404 — Sweep run not found

GET /api/sweep_runs/{id}/cells¶

List every cell belonging to the sweep. Each cell is a workflow_runs row and carries the resolved args used for its Jinja render. Ordering is (status, started_at ASC, id ASC) — grouped by status so callers see running/pending cells together.

Path parameters:

id — Sweep run ID

Response (200):

[
  {
    "id": "c1f2e3d4-0001-4e00-a000-000000000001",
    "workflow_name": "train",
    "status": "completed",
    "started_at": "2026-04-22T09:00:00+00:00",
    "completed_at": "2026-04-22T09:20:00+00:00",
    "args": {"lr": 0.001, "seed": 1, "dataset": "cifar10"},
    "error": null,
    "triggered_by": "web"
  },
  {
    "id": "c1f2e3d4-0002-4e00-a000-000000000002",
    "workflow_name": "train",
    "status": "running",
    "started_at": "2026-04-22T09:05:00+00:00",
    "completed_at": null,
    "args": {"lr": 0.01, "seed": 1, "dataset": "cifar10"},
    "error": null,
    "triggered_by": "web"
  }
]

triggered_by is web, cli, mcp, or schedule (post-V4 migration). To cancel a single cell, use POST /api/workflows/runs/{cell_id}/cancel with the cell's id as cell_id.

Error responses:

404 — Sweep run not found

POST /api/sweep_runs/{id}/cancel¶

Request sweep cancellation. Stamps cancel_requested_at on the sweep row, drains all pending cells, and fires scancel for any running cells. The sweep transitions to draining until every cell reaches a terminal state, then to cancelled via the aggregator. A sweep_run.status_changed event is emitted for any subscriber watches.

Calling this endpoint a second time on the same sweep is a no-op: the existing cancel_requested_at timestamp is preserved and the current row is returned.

Path parameters:

id — Sweep run ID

Example:

POST /api/sweep_runs/42/cancel

Response (202):

{
  "id": 42,
  "name": "train-lr-seed-sweep",
  "workflow_yaml_path": "/workflows/train.yaml",
  "status": "draining",
  "matrix": {"lr": [0.001, 0.01, 0.1], "seed": [1, 2, 3]},
  "args": {"dataset": "cifar10"},
  "fail_fast": false,
  "max_parallel": 4,
  "cell_count": 9,
  "cells_pending": 0,
  "cells_running": 2,
  "cells_completed": 4,
  "cells_failed": 0,
  "cells_cancelled": 3,
  "submission_source": "web",
  "started_at": "2026-04-22T09:00:00+00:00",
  "completed_at": null,
  "cancel_requested_at": "2026-04-22T09:30:00+00:00",
  "error": null
}

Error responses:

404 — Sweep run not found

Files¶

Method	Endpoint	Description
GET	`/api/files/mounts`	List configured mount points (name and remote only)
GET	`/api/files/mounts/config`	List mounts with full details (name, local, remote)
POST	`/api/files/mounts`	Add a mount to the current SSH profile
DELETE	`/api/files/mounts/{mount_name}`	Remove a mount from the current SSH profile
GET	`/api/files/browse`	Browse local filesystem under a mount's local root
GET	`/api/files/read`	Read text file contents from a mount
POST	`/api/files/sync`	Sync a mount's local directory to the remote via rsync

GET /api/files/mounts¶

Returns the list of mount points from the current SSH profile. Only mount names and remote prefixes are returned; local paths are never exposed.

Response:

[
  {
    "name": "ml-project",
    "remote": "/home/researcher/ml-project"
  }
]

Returns an empty list if no SSH profile is configured or the profile has no mounts.

GET /api/files/mounts/config¶

Returns all mounts with full details including local paths. Used by the mount management UI.

Response:

[
  {
    "name": "ml-project",
    "local": "/home/user/projects/ml-project",
    "remote": "/home/researcher/ml-project"
  }
]

Returns an empty list if no SSH profile is configured or the profile has no mounts.

POST /api/files/mounts¶

Add a new mount to the current SSH profile. The mount is persisted to the profile configuration file.

Request body:

{
  "name": "ml-project",
  "local": "/home/user/projects/ml-project",
  "remote": "/home/researcher/ml-project"
}

Response (200):

{
  "name": "ml-project",
  "local": "/home/user/projects/ml-project",
  "remote": "/home/researcher/ml-project"
}

Error responses:

409 — A mount with the same name already exists
422 — Validation error (missing required fields or invalid values)
503 — No SSH profile configured

DELETE /api/files/mounts/{mount_name}¶

Remove a mount from the current SSH profile.

Path parameters:

mount_name — Mount name to remove

Response (200):

{"status": "deleted", "name": "ml-project"}

Error responses:

404 — Mount not found
503 — No SSH profile configured

GET /api/files/browse¶

Browse directory contents under a mount's local root, returning entries with their corresponding remote paths.

Query parameters:

mount (required) — Mount name (must match a configured mount)
path (optional) — Relative path within the mount root (default: root directory)

Response:

{
  "entries": [
    {"name": "train.py", "type": "file", "size": 2048},
    {"name": "data", "type": "directory", "size": null},
    {"name": "latest", "type": "symlink", "size": null, "accessible": true, "target_kind": "directory"}
  ],
  "remote_prefix": "/home/researcher/ml-project/src",
  "mount_name": "ml-project"
}

Entry types: file, directory, symlink. Symlinks include an accessible field indicating whether the link target is within the mount boundary, and a target_kind field ("file" or "directory") indicating the type of the link target. Accessible symlinks to directories can be expanded in the file explorer.

Warning

Security: The resolved path must stay within the mount's local root. Path traversal attempts (e.g., ../../etc/passwd) return 403 Forbidden. Symlinks pointing outside the mount boundary are marked accessible: false and cannot be followed. Local filesystem paths are never included in the response.

Error responses:

400 — Path is not a directory
403 — Path outside mount boundary or permission denied
404 — Mount not found, directory not found, or no SSH profile configured

GET /api/files/read¶

Read text file contents from a mount's local root. Used by the file explorer to preview scripts before submission.

Query parameters:

mount (required) — Mount name
path (required) — Relative path within the mount root

Response:

{
  "content": "#!/bin/bash\n#SBATCH --gpus=1\npython train.py",
  "path": "scripts/train.sh",
  "mount": "ml-project"
}

Error responses:

400 — path parameter missing or path is not a file
403 — Path outside mount boundary
404 — File not found or no SSH profile configured
413 — File exceeds 1 MB size limit

POST /api/files/sync¶

Sync a mount's local directory to the remote server via rsync.

Request body:

{"mount": "ml-project"}

Response (200):

{"status": "synced", "mount": "ml-project"}

Error responses:

404 — Mount not found
502 — rsync command failed (includes missing rsync binary)
503 — No SSH profile configured

Config¶

Method	Endpoint	Description
GET	`/api/config`	Get current merged configuration
PUT	`/api/config`	Update user configuration
GET	`/api/config/paths`	List config file paths with existence status
POST	`/api/config/reset`	Reset user config to defaults
GET	`/api/config/ssh/profiles`	List SSH profiles and current active profile
POST	`/api/config/ssh/profiles`	Add a new SSH profile
PUT	`/api/config/ssh/profiles/{name}`	Update an existing SSH profile
DELETE	`/api/config/ssh/profiles/{name}`	Delete an SSH profile
POST	`/api/config/ssh/profiles/{name}/activate`	Set profile as current active
POST	`/api/config/ssh/profiles/{name}/mounts`	Add a mount to a profile
DELETE	`/api/config/ssh/profiles/{name}/mounts/{mount_name}`	Remove a mount from a profile
GET	`/api/config/env`	List active SRUNX* environment variables
GET	`/api/config/projects`	List projects from current profile's mounts
GET	`/api/config/projects/{mount_name}`	Read project config (srunx.json)
PUT	`/api/config/projects/{mount_name}`	Update project config
POST	`/api/config/projects/{mount_name}/init`	Initialize project config with example values

GET /api/config¶

Returns the current merged configuration (system + user + project).

Response:

{
  "resources": {
    "nodes": 1,
    "gpus_per_node": 0,
    "ntasks_per_node": 1,
    "cpus_per_task": 1,
    "memory_per_node": null,
    "time_limit": null,
    "partition": null,
    "nodelist": null
  },
  "environment": {
    "conda": null,
    "venv": null,
    "container": null,
    "env_vars": {}
  },
  "notifications": {
    "slack_webhook_url": null
  },
  "log_dir": "logs",
  "work_dir": null
}

PUT /api/config¶

Validate and save configuration to the user config file (~/.config/srunx/config.json).

Request body: A full or partial SrunxConfig object (same shape as GET response).

Response: The updated merged configuration.

Error responses:

422 — Validation error
500 — Failed to write config file

GET /api/config/paths¶

Returns all config file paths with their existence status and source label.

Response:

[
  {"path": "/etc/srunx/config.json", "exists": false, "source": "system"},
  {"path": "/home/user/.config/srunx/config.json", "exists": true, "source": "user"},
  {"path": "srunx.json", "exists": false, "source": "project"}
]

POST /api/config/reset¶

Reset user config to defaults. Overwrites ~/.config/srunx/config.json with a fresh SrunxConfig.

Response: The default configuration.

GET /api/config/ssh/profiles¶

Returns all SSH profiles and identifies the current active profile.

Response:

{
  "current": "dgx-server",
  "profiles": {
    "dgx-server": {
      "hostname": "dgx.example.com",
      "username": "researcher",
      "key_filename": "~/.ssh/id_ed25519",
      "port": 22,
      "description": "Main DGX cluster",
      "ssh_host": null,
      "proxy_jump": null,
      "mounts": [],
      "env_vars": {}
    }
  }
}

POST /api/config/ssh/profiles¶

Add a new SSH profile.

Request body:

{
  "name": "dgx-server",
  "hostname": "dgx.example.com",
  "username": "researcher",
  "key_filename": "~/.ssh/id_ed25519",
  "port": 22,
  "description": "Main DGX cluster",
  "ssh_host": null,
  "proxy_jump": null
}

Response: The created profile object.

Error responses:

409 — Profile with the same name already exists

PUT /api/config/ssh/profiles/{name}¶

Update an existing SSH profile. Only valid ServerProfile fields are applied: hostname, username, key_filename, port, description, ssh_host, proxy_jump, env_vars.

Request body: Object with fields to update (partial update).

Response: The updated profile object.

Error responses:

404 — Profile not found

DELETE /api/config/ssh/profiles/{name}¶

Delete an SSH profile and all its mounts.

Response:

{"ok": true}

Error responses:

404 — Profile not found

POST /api/config/ssh/profiles/{name}/activate¶

Set a profile as the current active profile.

Response:

{"ok": true}

Error responses:

404 — Profile not found

POST /api/config/ssh/profiles/{name}/mounts¶

Add a mount point to a profile.

Request body:

{
  "name": "ml-project",
  "local": "/home/user/projects/ml-project",
  "remote": "/home/researcher/ml-project"
}

Response: The created mount object.

Error responses:

404 — Profile not found
422 — Validation error

DELETE /api/config/ssh/profiles/{name}/mounts/{mount_name}¶

Remove a mount from a profile.

Response:

{"ok": true}

Error responses:

404 — Profile or mount not found

GET /api/config/env¶

Returns all SRUNX_* and SLACK_WEBHOOK_URL environment variables currently set in the server process.

Response:

[
  {
    "name": "SRUNX_DEFAULT_PARTITION",
    "value": "gpu",
    "description": "Default SLURM partition"
  },
  {
    "name": "SRUNX_SSH_PROFILE",
    "value": "dgx-server",
    "description": "SSH profile for web server"
  }
]

Returns an empty list if no SRUNX_* variables are set.

GET /api/config/projects¶

List projects derived from the current SSH profile's mounts.

Response:

[
  {
    "mount_name": "ml-project",
    "local_path": "/home/user/projects/ml-project",
    "remote_path": "/home/researcher/ml-project",
    "config_exists": true,
    "config_path": "/home/user/projects/ml-project/srunx.json"
  }
]

Returns an empty list if no SSH profile is active.

GET /api/config/projects/{mount_name}¶

Read srunx.json from a mount's local directory.

Response:

{
  "mount_name": "ml-project",
  "local_path": "/home/user/projects/ml-project",
  "config_path": "/home/user/projects/ml-project/srunx.json",
  "exists": true,
  "config": {
    "resources": {"gpus_per_node": 4, "time_limit": "8:00:00"},
    "environment": {"conda": "ml_env"}
  }
}

Error responses:

400 — No active SSH profile
404 — Mount not found in active profile

PUT /api/config/projects/{mount_name}¶

Save srunx.json to a mount's local directory.

Request body: A SrunxConfig object.

Response: The saved project config response.

Error responses:

400 — No active SSH profile
404 — Mount not found
422 — Validation error
500 — Failed to write file

POST /api/config/projects/{mount_name}/init¶

Initialize srunx.json with example values in a mount's local directory.

Response: The created project config response with example values.

Error responses:

400 — No active SSH profile
404 — Mount not found
409 — srunx.json already exists

History¶

Method	Endpoint	Description
GET	`/api/history`	Get recent job execution history
GET	`/api/history/stats`	Get aggregate job statistics

GET /api/history/stats¶

Query parameters:

from (optional) — Start date (ISO format)
to (optional) — End date (ISO format)

Response:

{
  "total": 42,
  "completed": 35,
  "failed": 4,
  "cancelled": 3,
  "avg_runtime_seconds": 3600.0
}

Error Responses¶

All errors follow this format:

{"detail": "Error message description"}

Status codes:

400 — Invalid input (e.g., negative job ID, path is not a directory)
403 — Path outside mount boundary or permission denied
404 — Resource not found (job, workflow, mount, directory, run)
409 — Resource already exists (e.g., workflow or mount with duplicate name)
413 — YAML content too large
422 — Validation error or invalid state transition (e.g., cancelling an already-terminal run)
500 — Internal error (e.g., script rendering failure)
502 — SLURM command, rsync command, or sbatch submission failed
503 — SSH connection not configured or rsync not installed

Configuration¶

The Web UI is configured via environment variables:

Variable	Default	Description
`SRUNX_SSH_PROFILE`	(current profile)	srunx SSH profile name
`SRUNX_SSH_HOSTNAME`	—	Direct SSH hostname
`SRUNX_SSH_USERNAME`	—	Direct SSH username
`SRUNX_SSH_KEY`	—	Path to SSH private key
`SRUNX_SSH_PORT`	22	SSH port