PDF Data Extraction Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="pdf-data")

# 1. A bucket for the invoices, and a collection that splits every page into
#    classified blocks (paragraph, table, form, list, header, figure) with OCR text
bucket = client.buckets.create(
    bucket_name="invoices",
    bucket_schema={"properties": {"invoice": {"type": "pdf"}}},
)
collection = client.collections.create(
    collection_name="invoices",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]]},
    feature_extractor={
        "feature_extractor_name": "document_graph_extractor",
        "version": "v1",
        "parameters": {
            "layout_detector": "docling",
        },
    },
)

# 2. Upload and process
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "invoice", "type": "pdf", "data": "s3://your-bucket/invoices/inv-1042.pdf"}],
)
client.collections.trigger(collection["collection_id"])

# 3. Pull the table blocks, with page and bounding box
tables = client.documents.list(
    collection["collection_id"],
    filters={"AND": [{"field": "object_type", "operator": "eq", "value": "table"}]},
)
for block in tables["results"]:
    print(block["page_number"], block["bbox"], block.get("text_corrected") or block["text_raw"])