examples

examples — runnable scripts

Each script is self-contained and runs against the bundled sample/acme_invoice.pdf + sample/acme_invoice.png fixtures. The base URL defaults to http://localhost:8000 (override with TURBO_OCR_BASE_URL).

export TURBO_OCR_BASE_URL=http://localhost:8000  # optional, this is the default
python examples/00_quickstart.py
python examples/02_pdf_to_markdown.py

#	Example	What it shows	README section
00	00_quickstart.py	The smallest useful script — sync image OCR.	Quickstart
01	01_image_ocr_with_layout.py	Image OCR with layout, reading_order, include_blocks; dump blocks as JSON.	Image OCR
02	02_pdf_to_markdown.py	PDF -> Markdown via render_to_markdown.	PDF -> Markdown
03	03_searchable_pdf.py	PDF -> searchable PDF, then verified with pypdf.	Searchable PDF
04	04_async_client.py	AsyncClient + asyncio.gather for concurrent OCR.	Async
05	05_batch.py	recognize_batch over multiple images.	Batch
06	06_grpc.py	GrpcClient — same surface as Client, gRPC transport.	gRPC
07	07_retry_and_timeout.py	Custom RetryPolicy + per-request timeout=.	Retry policy
08	08_custom_httpx_client.py	Pass your own httpx.Client for TLS / limits.	Custom httpx.Client
09	09_markdown_style.py	Register a custom layout label + renderer on MarkdownStyle.	Custom Markdown labels
10	10_tables_and_formulas.py	Iterate response.tables / response.formulas — see caveat below.	Tables and formulas
11	11_folder_pipeline.py	AsyncClient + asyncio.Semaphore for a bounded-concurrency folder pipeline.	—
12	12_hooks_and_logging.py	httpx event hooks + the SDK's stdlib logger.	Logging

Tables and formulas — partial support today. As of server v2.2.3, the server detects table and formula regions (you get a bounding_box and row-major OCR'd text) but does not emit cell structure or LaTeX source. Table.html, Table.cells, and Formula.latex are always None in the current responses. The SDK is forward-compatible: when the server ships table-structure-recognition and LaTeX OCR, those fields will populate without any SDK code changes.

Sample fixtures

sample/acme_invoice.pdf and sample/acme_invoice.png are a fictional two-page ACME Corp invoice (line items, totals, terms). They're committed so the examples run out of the box; regenerate with:

python examples/sample/generate.py

Requires reportlab and pypdfium2 (both are dev-time only — not runtime dependencies of the SDK itself).