ncmec-cybertip

A fully typed, async Python client for the NCMEC CyberTipline Reporting API (ispws), built on httpx and pydantic-xml.

The CyberTipline API lets Electronic Service Providers (ESPs) submit reports of apparent child sexual exploitation to the National Center for Missing & Exploited Children (NCMEC) programmatically, instead of via the web form. This library wraps all seven endpoints and models the full XML report / file-details schema as type-checked Python objects.

The XML documents produced by this library are built from the published technical documentation. NCMEC recommends validating submissions against the authoritative XSD (GET /xsd) — see Validation.

Features

• Async client (httpx.AsyncClient) covering all endpoints: status, xsd, submit, upload, fileinfo, finish, retract.
• Fully typed report and file-details models (pydantic-xml), with field validation (lengths, ranges, patterns, enums) mirroring the schema.
• Presence flags (<sextortion/>, <viral/>, …) modeled as plain booleans.
• Structured errors: non-zero response codes raise ApiError carrying the code, description, report ID, and Request-ID header.
• Ships with py.typed; 100% test coverage; ruff + ty clean.

Installation

uv add ncmec-cybertip
# or
pip install ncmec-cybertip

Requires Python 3.13+.

Quick start

import asyncio
from datetime import datetime, timezone

from ncmec_cybertip import (
    CyberTiplineClient,
    Details,
    Email,
    EventName,
    FileAnnotations,
    FileDetails,
    FileRelevance,
    IncidentSummary,
    IncidentType,
    InternetDetails,
    IpCaptureEvent,
    NameValuePair,
    OriginalFileHash,
    Person,
    Report,
    Reporter,
    TESTING_URL,
    WebPageIncident,
    file_hashes,
)


async def main() -> None:
    report = Report(
        incident_summary=IncidentSummary(
            incident_type=IncidentType.CHILD_PORNOGRAPHY,
            incident_date_time=datetime(2012, 10, 15, 8, tzinfo=timezone.utc),
        ),
        internet_details=[
            InternetDetails(
                web_page_incident=WebPageIncident(url=["http://badsite.example/x"])
            )
        ],
        reporter=Reporter(
            reporting_person=Person(
                first_name="John",
                last_name="Smith",
                email=[Email(value="jsmith@example.com")],
            )
        ),
    )

    async with CyberTiplineClient(
        username="usr123",
        password="pswd123",
        base_url=TESTING_URL,  # omit for production
    ) as client:
        await client.status()                       # verify connectivity + auth

        opened = await client.submit(report)        # 1. open the report
        report_id = opened.report_id

        uploaded = await client.upload(report_id, "evidence.jpg")  # 2. add a file
        file_id = uploaded.file_id

        # 3. optional: attach details/metadata for the uploaded file
        details = FileDetails(
            report_id=report_id,
            file_id=file_id,
            original_file_name="evidence.jpg",
            file_relevance=FileRelevance.REPORTED,
            file_viewed_by_esp=True,
            exif_viewed_by_esp=True,
            file_annotations=FileAnnotations(viral=True, generative_ai=True),
            # Compute MD5 + SHA1 + SHA256 from the file with the standard library,
            # or hand-build OriginalFileHash(value=..., hash_type=...) yourself.
            original_file_hash=file_hashes("evidence.jpg"),
            ip_capture_event=IpCaptureEvent(
                ip_address="63.116.246.17",
                event_name=EventName.UPLOAD,
                date_time=datetime(2011, 10, 31, 12, tzinfo=timezone.utc),
                port=443,
            ),
            details=[
                Details(
                    name_value_pair=[
                        NameValuePair(name="Make", value="Canon"),
                        NameValuePair(name="Model", value="EOS 5D"),
                    ]
                )
            ],
            additional_info=["File was originally posted with 6 others"],
        )
        await client.file_info(details)

        done = await client.finish(report_id)       # 4. finish (files it!)
        print("Filed report", done.report_id, "with files", done.file_ids)


asyncio.run(main())

The report lifecycle

1. submit(report) — opens a report, returns the assigned report_id.
2. upload(report_id, file) — optional, repeatable; returns a file_id + MD5.
3. file_info(file_details) — optional per file; adds metadata/annotations.
4. finish(report_id) — files the report with NCMEC (irreversible).
5. retract(report_id) — cancels an unfinished report.

An unfinished report is automatically deleted by NCMEC 24 hours after it is opened, or 1 hour after the last modification, whichever is later.

Authentication

GET/POST over HTTPS with HTTP Basic auth. Credentials are issued by NCMEC and differ between the testing (https://exttest.cybertip.org/ispws) and production (https://report.cybertip.org/ispws) environments — exposed as TESTING_URL and PRODUCTION_URL. base_url defaults to production.

Error handling

from ncmec_cybertip import ApiError, ResponseCode

try:
    await client.submit(report)
except ApiError as err:
    print(err.response_code)   # e.g. ResponseCode.VALIDATION_FAILED (4100)
    print(err.description)     # server-provided description
    print(err.request_id)      # Request-ID header — include when contacting NCMEC

Documented response codes are available as ResponseCode constants (the list is non-exhaustive, so unknown codes still round-trip).

Uploading files

upload() accepts a path, raw bytes, or an open binary file object:

await client.upload(report_id, "photo.jpg")             # path
await client.upload(report_id, raw_bytes, filename="x") # bytes
with open("photo.jpg", "rb") as fh:
    await client.upload(report_id, fh, filename="photo.jpg")

File hashes

Original file hashes for FileDetails can be computed from the standard library (hashlib) or built by hand. file_hashes() accepts a path, bytes, or a binary file object and returns OriginalFileHash entries; it defaults to MD5 + SHA1 + SHA256:

from ncmec_cybertip import OriginalFileHash, file_hashes

details.original_file_hash = file_hashes("evidence.jpg")             # md5, sha1, sha256
details.original_file_hash = file_hashes(raw_bytes, algorithms=["sha256"])

# Single algorithm, or supply a precomputed digest yourself:
OriginalFileHash.compute(raw_bytes, "sha1")
OriginalFileHash(value="fafa5efeaf3cbe3b23b2748d13e629a1", hash_type="MD5")

The hashType label is the conventional upper-case form (MD5/SHA1/SHA256) for known algorithms; any other hashlib name is upper-cased.

Industry classification

Files may optionally carry an ESP-designated IndustryClassification (A1/A2/B1/B2), where the letter is the apparent age of the minor (A = prepubescent, B = pubescent) and the digit is the content rank (1 = sex act, 2 = lascivious exhibition):

from ncmec_cybertip import FileDetails, FileRelevance, IndustryClassification

details = FileDetails(
    report_id=report_id,
    file_id=file_id,
    file_relevance=FileRelevance.REPORTED,
    industry_classification=IndustryClassification.A1,
)

Per the schema, only Reported content may be classified: setting an industry_classification (or a potential_meme annotation) on a Supplemental Reported file raises ValidationError locally, before it would be rejected by NCMEC at /finish.

Bring your own httpx client

Pass an existing httpx.AsyncClient (for custom transports, proxies, retries). When you do, the library will not close it for you:

async with httpx.AsyncClient(...) as http:
    client = CyberTiplineClient(username="u", password="p", client=http)
    ...

Validation

Model construction enforces the documented field rules (max lengths, numeric ranges, regex patterns, enum membership). For authoritative structural validation, fetch and validate against the live schema:

xsd = await client.get_xsd()

Development

uv sync
uv run ruff format .
uv run ruff check .
uv run ty check
uv run pytest --cov --cov-fail-under=100

Integration tests

Live tests are marked integration and skip unless credentials are provided via environment variables. They are never run by default.

# Full submit -> retract cycle against the NCMEC TEST environment (exttest).
# Never calls finish(), so nothing is filed.
NCMEC_TEST_USER=... NCMEC_TEST_PASS=... uv run pytest -m integration

# Read-only auth smoke test against PRODUCTION /status (no report is submitted).
NCMEC_PROD_USER=... NCMEC_PROD_PASS=... uv run pytest -m integration \
    tests/test_integration.py::test_status_live_prod

Releasing

Releases are automated via GitHub Actions + PyPI Trusted Publishing (OIDC — no API tokens). To cut a release:

1. Bump version in pyproject.toml (e.g. 0.1.0 -> 0.1.1). At the first stable release, also change the Development Status classifier from 4 - Beta to 5 - Production/Stable.

2. Commit, then push a matching tag:

   git tag v0.1.1
   git push origin v0.1.1

The Release workflow then builds the sdist + wheel, verifies the tag matches the package version, publishes to PyPI, and creates a GitHub Release with generated notes and the built artifacts attached. A mismatch between the tag and pyproject.toml version fails the build before anything is published.

Disclaimer

This is an unofficial client and is not affiliated with or endorsed by NCMEC. Reporting obligations and the meaning of each field are governed by NCMEC's documentation and applicable law (e.g. 18 U.S.C. § 2258A). Always validate against the official XSD before submitting in production.

License

MIT