seekit

seekit wraps multiple search providers behind a common SerpItem model and can search live using curl_cffi request templates derived from captured HAR files.

Supported providers:

• Bilibili
• Brave
• DuckDuckGo
• Reddit
• So.com
• Sogou
• Threads
• Toutiao
• Weibo
• YouTube

Install

pip install seekit

Usage

Search live:

import seekit

results = seekit.search("latest OpenAI reasoning model", provider="google")
for item in results[:3]:
    print(item.provider, item.title, item.url)

CLI:

seekit OpenAI
seekit latest OpenAI reasoning model --engine google --limit 5
seekit OpenAI --engine tiktok --format json
seekit OpenAI --engine google --format csv
seekit OpenAI --engine google --debug

Each parsed item uses the shared pydantic model:

class SerpItem(BaseModel):
    provider: str
    title: str | None
    excerpt: str | None
    url: str | None
    author: str | None
    cover_url: str | None
    time: str | None

Fixture parsing is still available for tests and parser debugging:

from pathlib import Path
from har import parse_har

results = parse_har("google", Path("data/pages/google.har"))

Development

uv run pytest

Tests read the HAR fixtures under data/pages, verify the normalized item shape, and validate that live request templates substitute the search keyword correctly.

Server

uv run uvicorn seekit.server:app --reload

Then open http://127.0.0.1:8000/ or call:

curl 'http://127.0.0.1:8000/api/search?q=OpenAI&provider=google'

Docs

Read the Docs config lives in .readthedocs.yaml and Sphinx sources live in docs/.

Release

make publish-testpypi
make publish-pypi

These targets build with uv build and publish with uvx uv-publish.