typefold-client

Python client for the Typefold API. Needs Python 3.10 or later; its only dependency is httpx. It has a blocking TypefoldClient and an asyncio AsyncTypefoldClient with the same methods.

pip install typefold-client

Create one client per API key and reuse it. The key needs render:create, render:read and template:read, and render:cancel to cancel renders.

from typefold import TypefoldClient

client = TypefoldClient("https://pdf.example.com", api_key)

Close it with client.close() when your application stops, or use it as a context manager.

Render a report

render_report renders a published report or saved template by name or ID and returns the PDF or XLSX bytes:

from typefold import (
    Orientation,
    PageSize,
    PdfMetadata,
    RenderReportRequest,
    ReportFormat,
    ReportPdfOptions,
    TemplateRef,
)

pdf = client.render_report(
    RenderReportRequest(
        template=TemplateRef.by_name("SoldGoodsForWarehouseReportPdf"),
        data={"Data": {"reportRows": report_rows, "summaryRows": summary_rows}},
        pdf=ReportPdfOptions(
            page_size=PageSize.A4,
            orientation=Orientation.LANDSCAPE,
            metadata=PdfMetadata(title="Sold goods, week 39"),
        ),
    )
)

xlsx = client.render_report(
    RenderReportRequest(
        format=ReportFormat.XLSX,
        template=TemplateRef.by_name("SoldGoodsForWarehouseReportXlsx"),
        data=data,
    )
)

data is sent as written: dictionary keys keep their names, and dates, decimals, UUIDs and dataclasses are written as JSON.

Templates imported from jsreport keep their shortids, so stored shortids still work: TemplateRef.by_shortid("rkJTnK2ce"). list_reports returns each template's shortid.

A retry with the same idempotency_key returns the render the key already created. get_report_data_schema returns the JSON Schema of a report's data; list_reports lists templates and the formats each renders.

Render notification text

render_string renders a template string synchronously, without a PDF, for email bodies, subjects, SMS and push messages. The Handlebars engine runs #if, #each and your helpers:

from typefold import RenderStringRequest, StringEngine, StringFormat

body = client.render_string(
    RenderStringRequest(
        template="{{#if (isEqual Status 'paid')}}Paid{{else}}Due{{/if}}: {{Booking.Code}}",
        data=model,
        engine=StringEngine.HANDLEBARS,
        helpers="function isEqual(a, b) { return a === b; }",
        format=StringFormat.HTML,
    )
)

StringFormat.HTML escapes values; StringFormat.TEXT (the default) writes them as they are. Without engine, only {{path}} placeholders are replaced.

Render HTML or a template version

render creates a render, waits for it and downloads the PDF:

from typefold import DirectRenderRequest, HtmlDocument, PageSize, RenderSettings, TemplateVersionRenderRequest

pdf = client.render(
    DirectRenderRequest(
        document=HtmlDocument(html="<h1>Invoice 42</h1>"),
        settings=RenderSettings(page_size=PageSize.A4),
    )
)

ticket = client.render(
    TemplateVersionRenderRequest(template_version_id=version_id, data={"ticket": {"number": "TK-1"}})
)

It gives up after two minutes unless you pass a timeout in seconds. To render in the background, use create_render, then wait_for_render, download or cancel_render, or the API's render webhooks.

Render a batch

create_batch creates up to 100 renders in one request; if any item is invalid, none is created. Items are DirectRenderRequest or TemplateVersionRenderRequest; their wait_ms is not sent. wait_for_batch polls until no item is queued or processing (10 minutes unless you pass a timeout), and download_batch returns one PDF of every item in order:

batch = client.create_batch(
    [TemplateVersionRenderRequest(template_version_id=version_id, data=ticket) for ticket in tickets],
    idempotency_key="tickets-2026-10-12",
)
batch = client.wait_for_batch(batch)

if batch.counts.succeeded == batch.total:
    pdf = client.download_batch(batch.id)

A failed item does not stop the others; batch.items gives each render's id and state, to read, download or cancel it as a single render. get_batch reads a batch again.

Async

AsyncTypefoldClient has the same methods as coroutines:

from typefold import AsyncTypefoldClient

async with AsyncTypefoldClient("https://pdf.example.com", api_key) as client:
    pdf = await client.render_report(request)

Errors

A request the API refuses with 429 RATE_LIMITED, such as a burst of parallel renders, is sent again after its Retry-After up to max_rate_limit_retries times (3 by default; pass 0 to turn this off). Other failures raise TypefoldError. error.code is the API's code, such as NOT_FOUND, RATE_LIMITED or RENDER_TIMEOUT, and error.retryable says whether a retry can help. When a render was created but failed or timed out, error.error.render_id names it.

Versions

Use a client no newer than your server: older clients keep working because /v1 only grows, but a newer client's fields or endpoints answer 400 or 404 on an older server.

httpx client

To share an httpx.Client (or httpx.AsyncClient), pass it as http_client with a timeout above 95 seconds, the longest a report PDF can take. The API key goes on each request, not the client, and never to download URLs. A client you pass in is not closed with the TypefoldClient.

Development

cd clients/python
python -m venv .venv && .venv/bin/pip install -e '.[test]'
.venv/bin/pytest && .venv/bin/mypy src tests && .venv/bin/ruff check . && .venv/bin/ruff format --check .