POMA AI ⬣ Docs

Appearance

python
from poma import PomaArchive

PomaArchive wraps a raw .poma archive. A .poma archive is a zip file that contains the structured SDK output.

In practice, the core archive files are chunks.json and chunksets.json, with optional helpers such as image_sources.json and files inside assets/. Some archives also include additional processing artifacts like final_input.md, metadata_archive.json, metadata_content.json, tables/, and input_prepared/.

Archive contents ​

A .poma file is a ZIP archive with the following structure. Click an item to see its format.

Core SDK files

These are the files the SDK relies on when unpacking into `PomaResult`.

Individual chunk records

Contains the extracted chunk objects. These hold the original text, hierarchy depth, page reference, file identity, and optional links back to tables or images.

[
  {
    "content": "## Results",
    "depth": 1,
    "poma_page": 8,
    "chunk_index": 42,
    "file_id": "report-20260317",
    "parent_chunk_index": 41,
    "table": "table_00003",
    "image_name": "image_00002.jpeg",
    "to_embed": "Results"
  }
]

Extended processing artifacts

These can be present in richer archive exports, such as the ones shown in the POMA app.

Constructor ​

python
PomaArchive(
    data: bytes | None = None,
    path = None,
)

Pass either:

  • data for in-memory archive bytes
  • path for an archive on disk

Methods ​

unpack ​

python
unpack() -> PomaResult

Unpack the archive and return typed PomaResult data.

unpack() focuses on typed chunks, chunksets, and images, even when the archive also contains extra processing artifacts.

Example ​

python
from poma import PomaArchive

archive = PomaArchive(path="store/example.poma")
result = archive.unpack()