API2Convert Python SDK

The official Python client for the API2Convert file-conversion API. Convert, compress and transform images, documents, audio, video, ebooks, archives and CAD — and run operations like OCR, merge, thumbnail and website capture — in one line of code.

from api2convert import Api2Convert

client = Api2Convert("YOUR_API_KEY")

client.convert("invoice.docx", "pdf").save("invoice.pdf")

That single call creates a job, uploads your file, starts it, waits for it to finish and gives you back a result you can save. No polling loops, no manual upload handling.

Requirements

• Python 3.10+
• httpx (installed automatically)

Install

pip install api2convert

Get an API key from the API2Convert dashboard / documentation.

Quick start

from api2convert import Api2Convert

# Reads the API2CONVERT_API_KEY environment variable when no key is passed.
client = Api2Convert("YOUR_API_KEY")

# 1) From a local file
client.convert("photo.png", "jpg").save("photo.jpg")

# 2) From a URL
client.convert("https://example.com/photo.png", "jpg").save("photo.jpg")

# 3) With conversion options (discover them via client.options("jpg"))
client.convert("photo.png", "jpg", {"quality": 85, "width": 1280, "height": 720}).save("out/")

convert(source, to, options=None, ...) — source is a local path, a public URL, or an open binary stream; to is the target format; options is the conversion options map for that target. Less-common controls are keyword-only arguments: category, timeout, output_index, filename, download_password. The returned ConversionResult lets you:

result = client.convert("report.docx", "pdf")

result.save("report.pdf")       # stream to a file
result.save("downloads/")       # ...or a directory (keeps the server filename)
data = result.contents()        # ...or get the raw bytes
url = result.url()              # ...or just the download URL

Password-protect the result

Pass download_password and the output is locked behind it. The SDK remembers the password and sends it automatically when you download — you don't pass it again:

result = client.convert("statement.docx", "pdf", download_password="hunter2")

result.save("statement.pdf")    # the password is applied for you

The download URL still needs the password from anywhere else (a browser, curl, another process), via the X-Api2convert-Download-Password header. When you already hold an OutputFile — e.g. from the Jobs API — hand the password to download():

client.download(output, "hunter2").save("out/")

Asynchronous conversions & webhooks

For long-running jobs, start the conversion and get notified via a webhook instead of waiting:

job = client.convert_async("movie.mov", "mp4", callback="https://your-app.example.com/webhooks/api2convert")

In your webhook handler, verify and parse the callback:

from api2convert import Api2Convert
from api2convert import SignatureVerificationError

payload = request.body                       # the RAW body (bytes or str)
signature = request.headers.get("X-Oc-Signature")

try:
    event = Api2Convert.webhooks().construct_event(payload, signature, "YOUR_WEBHOOK_SECRET")
    job = event.job
    # ... react to job.status.code ...
except SignatureVerificationError:
    ...  # respond 400

Signed webhooks are being rolled out. Until they are enabled for your account no signature is sent — call Api2Convert.webhooks().parse(payload) (or pass an empty secret) to deserialize the callback without verifying.

Cloud storage

Read an input straight from your own cloud storage, and/or deliver the converted output into a bucket — no local upload or download in between. Supports Amazon S3, Azure and Google Cloud.

Import an input with the per-provider CloudInput factory (keys are flat and lowercase, exactly as the API expects) and convert it like any other source:

from api2convert import Api2Convert, CloudInput

client = Api2Convert("YOUR_API_KEY")

s3_input = CloudInput.amazon_s3(
    bucket="my-bucket",
    file="invoices/report.docx",
    accesskeyid="AKIA...",
    secretaccesskey="...",
)

client.convert(s3_input, "pdf").save("report.pdf")

To deliver the result into a bucket instead, attach an OutputTarget via output_targets. When any output target is set the conversion delivers straight to your storage and produces no local file — so there is nothing to save():

from api2convert import Api2Convert, CloudProvider, OutputTarget

client = Api2Convert("YOUR_API_KEY")

target = OutputTarget.of(
    CloudProvider.AMAZON_S3,
    parameters={"bucket": "my-bucket", "file": "out/report.pdf"},
    credentials={"accesskeyid": "AKIA...", "secretaccesskey": "..."},
)

# Delivered to the bucket — the job completes without a downloadable output.
client.convert("report.docx", "pdf", output_targets=[target])

azure(...) and google_cloud(...) build inputs the same way; output uses the generic OutputTarget for every provider. Credentials ride in the request body but are redacted in exception messages and object inspection (repr) as [REDACTED] — they are never logged or printed.

Error handling

Every failure is a typed exception extending api2convert.Api2ConvertError:

from api2convert import (
    Api2Convert,
    AuthenticationError,
    ConversionFailedError,
    RateLimitError,
    ValidationError,
)

try:
    Api2Convert("KEY").convert("photo.png", "jpg").save("photo.jpg")
except ValidationError as e:
    ...  # bad target / option — str(e) explains
except AuthenticationError:
    ...  # bad or missing API key
except RateLimitError as e:
    ...  # too many requests — retry after e.retry_after seconds
except ConversionFailedError as e:
    ...  # the job failed — inspect e.errors()

Exception	When
AuthenticationError	401 / 403 — bad or missing key
PaymentRequiredError	402 — no remaining quota
ValidationError	400 / 422 — invalid request (e.g. unknown target)
NotFoundError	404 — resource doesn't exist
RateLimitError	429 — exposes .retry_after
ServerError	5xx
NetworkError	transport failure / non-JSON success body
ConversionFailedError	the job reached failed; exposes .job and .errors()
ConversionTimeoutError	the job didn't finish within the poll timeout
SignatureVerificationError	a webhook payload failed verification

Transient failures (429, 5xx, network errors) are retried automatically with exponential backoff.

Power user: the full job API

convert() is sugar over the Jobs API. Drop down to it for compound jobs, merges, presets, custom polling or job chaining:

job = client.jobs.create({
    "process": False,
    "conversion": [{"target": "pdf", "options": {"pdf_a": True}}],
})

client.jobs.upload(job, "contract.docx")                          # local file
client.jobs.add_input(job.id, {"type": "remote", "source": "https://example.com/appendix.docx"})

client.jobs.start(job.id)
done = client.jobs.wait(job.id, timeout_seconds=120)

for output in done.output:
    client.download(output).save("out/")

Available resources: client.jobs, client.conversions (the catalog + option discovery), client.presets, client.stats, client.contracts.

Discover the valid options for any target:

options = client.options("jpg")            # -> {"quality": {...}, "width": {...}, ...}

Configuration

client = Api2Convert(
    "YOUR_API_KEY",
    timeout=30,             # per-request network timeout (seconds)
    max_retries=2,          # automatic retries for transient failures
    poll_interval=1.0,      # first poll interval when waiting (seconds)
    poll_max_interval=5.0,  # backoff cap (seconds)
    poll_timeout=300,       # give up waiting after this many seconds
)

Bring your own configured httpx.Client by passing http_client=....

Security — never publish your API key

• Never hard-code or commit your API key. Load it from the environment (API2CONVERT_API_KEY) or a secrets manager.
• In CI, store it as a masked & protected variable and never print it to logs.
• Treat the per-job upload token and your webhook signing secret with the same care.
• The SDK never logs your key/token and never puts them in exception messages.
• If a key is ever exposed, revoke and rotate it in the API2Convert dashboard immediately.

Development

python -m venv .venv && . .venv/bin/activate
pip install -e '.[dev]'

ruff check src tests examples && ruff format --check src tests examples   # lint + format
mypy src examples                                                         # static typing (strict)
pytest -m "not live"                                     # offline unit tests

Live conformance tests run against the real API when API2CONVERT_API_KEY is set:

API2CONVERT_API_KEY=... pytest -m live

Examples

Runnable, self-contained programs live in examples/ — one per documented guide. Each reads the key from API2CONVERT_API_KEY (and honours API2CONVERT_BASE_URL). Run any of them with, e.g., API2CONVERT_API_KEY=your-key python examples/quickstart.py.

Example	Guide
quickstart.py	Convert a remote file, look the job up, download the output
convert_files.py	Browse the conversions catalog, then convert
uploading_files.py	One-call upload + convert of a local file
job_lifecycle.py	Drive create → add input → start → wait → outputs by hand
add_watermark.py	Stamp a PNG onto a PDF (multi-input job)
create_thumbnails.py	Render a document page preview
compress_files.py	Shrink a file with the compress operation
create_archives.py	Bundle several files into a ZIP
create_hashes.py	Compute a SHA-256 checksum
extract_assets.py	Extract embedded assets from a document
file_analysis.py	Extract file metadata as JSON
compare_files.py	Diff two images
capture_website.py	Screenshot a URL to PNG
audio_operations.py	Re-encode audio (WAV → AAC)
image_operations.py	Resize an image
webhooks.py	Start a conversion asynchronously with a callback
presets.py	List saved conversion presets
statistics.py	Read API usage for a month
rate_limits.py	Inspect the account's contracts
authentication.py	Verify the API key with an authenticated call

The live conformance suite mirrors these examples 1:1 — each test performs the same operation and asserts success — plus two negative tests (an unknown target is a typed ValidationError; a bad key is a typed AuthenticationError that never leaks the key).

It runs automatically against the real API on every release tag (see .github/workflows/live-conformance.yml), so a published version is always verified end to end.

This SDK is hand-written and kept in sync with the API by an AI agent — see AGENTS.md and docs/SDK_CONTRACT.md. Notable changes are recorded in docs/CHANGELOG.md.

License

MIT — see LICENSE.