Browse documentation

Choose an API

Pick an API based on how you need to consume data, not just the input format.

The decision table

Need	Use	Result
Random access, editing, or repeated passes	`lokit.parse.<format>()`	`BaseStructure`
One bounded pass with document metadata	`lokit.stream.<format>()`	`StreamingStructure`
Async consumption without blocking the event loop	`lokit.parse.async_.<format>()`	Async iterator of `(unit_id, Data)`
Querying, navigation, and in-memory matching	`lokit.Lokit.parse()`	`Lokit` facade
Direct path-to-path conversion	`lokit.convert.<source>_to_<target>()`	Written output file
Flat interchange rows	`parse.to_dict` or `stream.to_dict`	List or lazy row iterator

Async parsers are iterated, not awaited

Async parse and stream functions return async iterators. Iterate them directly. The async writer functions are coroutines and are awaited.

async_pipeline.py

import lokit
async def copy_tmx() -> None:    async for unit_id, unit in lokit.parse.async_.tmx("messages.tmx"):        print(unit_id, unit.source)
    document = lokit.parse.tmx("messages.tmx")    await lokit.export.async_.xliff(document, "messages.xliff")

Streaming availability is intentionally narrower

Synchronous document streams are available for .lokit, TMX, XLIFF, DOCX, and PPTX. CSV, XLSX, PO, JSON i18n, HTML, and IDML are available as materialized parsers and async unit iterators. stream.to_dict() and stream.write_jsonl() are projection/output helpers rather than document parsers.