schemasnap 0.1.1

Structured data extraction from any URL — tables, forms, lists as JSON. No API key needed.

SchemaSnap

Structured data extraction from any URL — tables, forms, lists as JSON.

Born from shadow-web's SchemaSnap module. Focused, standalone, and designed to be the API that AI agents (and humans) call when they need structured web data.

from schemasnap import extract_from_url

result = extract_from_url("https://en.wikipedia.org/wiki/List_of_countries_by_GDP")
# result.tables[0] → {columns: ["Country", "GDP", ...], types: ["string", "currency", ...], rows: [[...], ...]}
# result.forms   → [{action: "/search", method: "GET", fields: [...]}]
# result.lists   → [{type: "unordered", items: ["Item 1", "Item 2"], total: 2}]

What it does

Input	Output	Browser needed?
URL	Tables + Forms + Lists as JSON	Yes (Playwright)
Raw HTML	Tables + Forms + Lists as JSON	No (lxml only)

Quick install

pip install schemasnap

# For URL extraction (requires Playwright)
pip install "schemasnap[browser]"
playwright install chromium

# For API server
pip install "schemasnap[server]"

Usage

Python SDK

from schemasnap import extract_from_url, extract_from_html, parse_page

# From URL (needs Playwright)
result = extract_from_url("https://example.com/data")
print(result.tables)      # [{columns, types, rows, total_rows, ...}]
print(result.forms)       # [{action, method, fields: [{name, type, label, required}]}]
print(result.lists)       # [{type, items, total}]
print(result.token_budget) # {raw_bytes, compressed_bytes, compression_ratio}

# From HTML string (no browser)
html = open("page.html").read()
result = extract_from_html(html)
print(result.tables)

API Server

pip install "schemasnap[server]"
uvicorn schemasnap.server:app --reload --port 8000

# Extract from URL
curl -X POST http://localhost:8000/v1/extract \
  -H "Content-Type: application/json" \
  -d '{"url": "https://en.wikipedia.org/wiki/List_of_countries_by_GDP"}'

# Compress HTML
curl -X POST http://localhost:8000/v1/compress \
  -H "Content-Type: application/json" \
  -d '{"html": "<table>...", "url": "https://example.com"}'

# Health check
curl http://localhost:8000/v1/health

Features

• Tables → columns, types (string/integer/number/currency/percentage/date/email/url), rows, total_rows
• Forms → action, method, fields with name, type, label, required, placeholder, options
• Lists → type (unordered/ordered/select), items, total
• HTML compression → 83-97% size reduction before extraction
• Semantic groups → "Login Form", "Navigation", "Checkout" auto-detected

Relationship to shadow-web

SchemaSnap started as schema_snap.py inside shadow-web — a broader agent-facing web interaction suite. This standalone version:

• Removes MCP server, security scan, self-healing, form fill, browser-use integration
• Inlines semantic grouping (no external dependency)
• Fixes the tree mutation bug during tag removal
• Adds FastAPI server with rate limiting and input validation
• Focuses entirely on one thing: URL → structured data

License

MIT