Python SDK

datamaker-forms reads a .dmf, validates, sealed-box encrypts, and posts. Depends only on PyNaCl. Python ≥ 3.9.

Install

pip install datamaker-forms

The import name is datamaker.

Submit

import datamaker as dm

with open("contact.dmf", "rb") as fh:
    form = dm.read_form(fh.read())   # verifies signature + form hash

result = dm.submit(
    form=form,
    values={"email": "ada@example.com", "full_name": "Ada"},
)
# → {"submission_id": ..., "edit_token": ..., "form_id": "contact_form"}

Or pass the bytes: dm.submit(dmf=open("contact.dmf","rb").read(), values=...).

Building the values dict

values is keyed by each field’s key (its storage name — not the label). Inspect the form to see the names + kinds you need:

for f in form.fields:
    print(f["key"], "·", f["kind"], "(required)" if f["required"] else "")
# email · email (required)
# full_name · text
# age · number
# subscribed · boolean
# plan · choice
# interests · multi-choice
# signup_date · date
# budget · money

Then map your data, using the value type each kind expects:

values = {
    "email":       "ada@example.com",  # email        -> str
    "full_name":   "Ada Lovelace",     # text         -> str
    "age":          37,                # number        -> int  (37, not "37")
    "subscribed":   True,              # boolean       -> bool
    "plan":        "pro",              # choice        -> str (one of the choice values)
    "interests":   ["news", "beta"],   # multi-choice  -> list[str]
    "signup_date": "2026-05-29",       # date          -> "YYYY-MM-DD"
    "budget":       1999.99,           # money         -> float
}

dm.submit(form=form, values=values)

For image / attachment fields, upload the bytes first and pass the returned ref as the value — see Files & attachments.

The full value type for every kind is in the field kinds reference.

Forgiving, but be intentional

By default submit coerces ("37" → 37, "yes" → True, a single choice → ["x"] for multi-choice) and validates: it raises ValidationError (with .issues) for a missing required field, an unknown key, a value for a read-only kind, or an out-of-list choice. Pass validate=False to skip, or allow_unknown=True to ignore extra keys.

try:
    dm.submit(form=form, values=values)
except dm.ValidationError as e:
    for i in e.issues:
        print(f"{i['field']}: {i['message']}")

API

• read_form(dmf_bytes, verify=True) -> FormDescriptor — verifies the Ed25519 signature + form hash; raises DmfError on tampering.
• submit(form=… | dmf=…, values=…, **opts) -> dict — options: api_base_url, submitter_id (default None), validate (default True), allow_unknown, verify, poster (inject an HTTP poster for tests).
• build_submission(form, values, **opts) -> dict — seal without sending.
• post_submission(envelope, **opts) -> dict — post a pre-built envelope.
• validate_values(fields, input, allow_unknown=False) -> (values, issues).

submit/build_submission raise ValidationError (with .issues); a non-2xx raises SubmissionError (with .status).

CLI

datamaker inspect contact.dmf
datamaker submit contact.dmf --field email=ada@example.com --field full_name=Ada
datamaker submit contact.dmf --data-file answers.json --dry-run

The wire format is identical to the JS and .NET SDKs — a .dmf signed anywhere verifies here, and submissions interoperate.