blazegraph-io 0.2.0

Python SDK for Blazegraph — parse PDFs into typed semantic document graphs

blazegraph-io

Python SDK for Blazegraph — parse PDFs into typed semantic document graphs.

Install

pip install blazegraph-io

Python 3.9+. Only dependency: httpx.

Quick Start

import blazegraphio as bg

# Local mode — no account needed, runs on your machine
graph = bg.parse_pdf("document.pdf")

print(f"{len(graph.nodes)} nodes, {len(graph.sections)} sections")

for section in graph.sections:
    print(section.content.text)
    print(section.location.semantic.breadcrumbs)
    print(f"  Page {section.location.physical.page}")

On first run, the SDK downloads the blazegraph-cli binary and a JRE automatically. Subsequent runs are instant.

API Mode

Switch to the hosted API with one line:

bg.configure(api_key="blaze_prod_...")
graph = bg.parse_pdf("document.pdf")  # same code, cloud processing

Async is also supported:

graph = await bg.parse_pdf_async("document.pdf")

What You Get

Every node has typed fields with IDE autocomplete:

node = graph.sections[0]

node.content.text                        # "Introduction"
node.token_count                         # 206
node.location.semantic.path              # "2.3"
node.location.semantic.breadcrumbs       # ["paper.pdf", "Introduction"]
node.location.physical.page              # 1
node.location.physical.bounding_box      # BoundingBox(x=91.9, y=585.9, ...)

Navigate the tree:

parent = node.get_parent(graph)
children = node.get_children(graph)

Render text:

print(section.render(graph, breadcrumbs=True))
# [paper.pdf > Introduction]
# The fundamental problem of communication...

Documentation

• Python SDK Guide — Full usage, configuration, error handling
• Schema Reference — Complete graph.json field documentation
• Configuration Reference — Tuning for your document type

License

MIT