nomad-sdk 0.1.0

Nomad headless browser engine — Python SDK

Nomad SDK — Python

Nomad is a zero-overhead headless browser engine. 5 MB binary, 6 MB RSS, 550 req/s. This package connects to a running Nomad daemon over a Unix Domain Socket.

pip install nomad-sdk

Quickstart

Make sure the Nomad daemon is running:

MALLOC_ARENA_MAX=1 nomad-daemon &

Then scrape a page in 2 lines:

from nomad_sdk import NomadClient

with NomadClient() as client:
    page = client.navigate("https://news.ycombinator.com")
    print(f"Title: {page.title}")
    print(f"Links: {len(page.find_all(tag='a'))}")

API

NomadClient(socket_path="/tmp/nomad-daemon.sock")

Method	Description
navigate(url, fmt=3)	Fetch and parse a page
snapshot(tab_id)	Get current page state without re-fetching
click(entity_id)	Click an element by entity ID
input_text(entity_id, value)	Type text into an input field
find(query)	Find elements by role, tag, intent, text
evaluate(js)	Execute arbitrary JavaScript
scroll(dx, dy)	Scroll the viewport
resize(w, h)	Resize the viewport
health()	Daemon health info
get_action_log(tab_id)	Tab action history

Output formats (fmt)

ID	Name	Use
0	RawBincode	Machine IPC
1	JSON	Debug
2	Markdown	Human read
3	AgentJson	LLM (full entities)
4	Summary	LLM (~150 bytes)
5	TokenMinimal	LLM (shortest)
6--10	Section/Accessibility/Reader/Form	Specialized

Page methods

page.find_one(role="button", text="Sign In")  # first match
page.find_all(tag="a", intent="navigation")    # all matches

Example — Click a button

from nomad_sdk import NomadClient

with NomadClient() as client:
    page = client.navigate("https://example.com/login")
    btn = page.find_one(role="button", text="Submit")
    if btn:
        client.click(btn.id)

Requirements

• Python ≥ 3.10
• Nomad daemon running (UDS socket at /tmp/nomad-daemon.sock)