syft-bg Python API

import syft_bg

Init

syft_bg.init(do_email="you@example.com")

Creates the config file. Optionally pass syftbox_root and token_path.

Ensure running

syft_bg.ensure_running(["notify", "approve"])

Starts the listed services if they aren't already running. Pass restart=True to force a restart.

• services — list of service names, or a dict mapping service names to config overrides
• restart — if True, restart services even if already running (default False)

Status

syft_bg.status

Shows email, services, auto-approval objects, and environment info. No parentheses needed.

Auto-approve

syft_bg.auto_approve(
    contents=["code/main.py", "run.sh"],
    file_paths=["code/params.json", "config.yaml"],
    peers=["charlie@org.com"],
    base_dir="~/SyftBox/datasites/me@org.com/apis/jobs/inbox/ds@org.com/my-job",
)

Registers files for auto-approval. Jobs matching these files from listed peers get approved automatically.

A path is relative to the job submission root, so code sits under code/ and the script the runner executes is run.sh. The two lists together must name every file of the submission, and run.sh must be in contents, or the rule matches nothing. config.yaml belongs in file_paths: its bytes carry the job name and the time it was submitted, so hashing them matches one job.

• contents — files (or directories) to approve by content
• file_paths — files to allow by name only (e.g. data files)
• peers — restrict to these emails. Omit to allow any peer
• name — optional name for the approval object
• base_dir — the job submission directory, which contents is resolved against. file_paths is stored as written, so write those relative to the same root

Auto-approve from job

from syft_bg import auto_approve_job

job = do_manager.jobs[0]

# Default: every file matched by content, except config.yaml and
# code/params.json, which are matched by name
auto_approve_job(job)

# Match data.json by name too; config.yaml and code/params.json stay by name
auto_approve_job(job, file_paths=["data.json"])

# Content-match main.py and the script that runs, allow the rest by name
auto_approve_job(
    job,
    contents=["main.py", "run.sh"],
    file_paths=["data.json", "config.yaml"],
)

contents must name run.sh, and the two lists together must name every file of the job, or auto_approve_job refuses to write an object that could never approve anything.

Creates an auto-approval config from an existing job's files. Calls auto_approve() internally.

• job — JobInfo object to use as template
• contents — filenames from the job to match by name AND content. Default (None): every file is content-matched except config.yaml and code/params.json, which are matched by name
• file_paths — filenames from the job to match by name only. With contents left unset, the other files are content-matched, except config.yaml and code/params.json, which stay matched by name. With both set, the two lists must name every file of the job between them
• peers — restrict to these emails. Defaults to the job's submitter
• name — optional name for the approval object (defaults to job name)

Service control

syft_bg.start()                              # start all services
syft_bg.stop()                               # stop all services
syft_bg.restart()                            # restart all services
syft_bg.ensure_running(["notify", "approve"])  # start listed services
syft_bg.logs("approve")                      # last 50 lines of approve service log
syft_bg.logs("notify")                       # last 50 lines of notify service log

Typical notebook flow

import syft_bg

# Create config
syft_bg.init(do_email="you@example.com")

# Start services
syft_bg.ensure_running(["notify", "approve"])

# Check everything
syft_bg.status

# Auto-approve future runs of this job
job = do_manager.jobs[0]
syft_bg.auto_approve_job(job)