How to Set Up the Browser Environment for Patent-Disclosure-Skill

The patent-disclosure-skill uses Playwright to drive a Chromium-based browser, with automatic fallback from system Chrome → Edge → bundled Chromium, controlled via environment variables PATENT_BROWSER_CHANNEL and PLAYWRIGHT_HEADED.

Setting up the browser environment for patent-disclosure-skill enables automated web-based patent searches through the CNIPA and other sources. This guide walks through installation, configuration, and verification based on the actual implementation in handsomestWei/patent-disclosure-skill.

Prerequisites

Before configuring the browser, ensure your system meets these requirements:

• Python 3.9+ — required by the project's dependencies
• Playwright — listed in requirements.txt and installed via pip
• System Chrome or Microsoft Edge (optional but recommended) — the skill prefers these over downloading bundled Chromium

Install Dependencies

Start with the core Python packages.

pip install -r requirements.txt

This installs playwright along with other required libraries.

If you do not have system Chrome or Edge installed, download the bundled Chromium binary:

python -m playwright install chromium

Configure Browser Channel Selection

The skill automatically detects available browsers, but you can override this behavior with environment variables.

Automatic Detection (Default)

When PATENT_BROWSER_CHANNEL is unset, browser.py iterates through _AUTO_CHANNELS in this order:

1. chrome — system Google Chrome
2. msedge — system Microsoft Edge
3. chromium — Playwright's bundled Chromium

The launch_chromium() function in skills/patent-search/tools/browser.py attempts each channel until one succeeds.

Force a Specific Channel

Set PATENT_BROWSER_CHANNEL to skip auto-detection:

export PATENT_BROWSER_CHANNEL=chrome   # Use system Chrome

export PATENT_BROWSER_CHANNEL=msedge   # Use system Edge

export PATENT_BROWSER_CHANNEL=chromium # Use bundled Chromium

The _forced_channel() helper at lines 52–60 in browser.py reads this variable and validates it against allowed values.

Control Headless vs. Headed Mode

By default, the browser runs headlessly (no visible window). To enable the GUI:

export PLAYWRIGHT_HEADED=1

The headed() function (lines 32–38) checks for 1, true, or yes in this variable. When headless=None is passed to launch_chromium(), this environment variable controls the actual behavior.

Verify Your Setup

Run the built-in probe to confirm everything works:

python -m skills.patent-search.tools.browser --probe

Expected output:

{"playwright": true, "channel": "chrome", "ok": true, "error": null, "hint": null}

The probe_launch() function (lines 21–28) performs a temporary headless launch and reports status. If ok is false, the hint field contains troubleshooting guidance in Chinese: "请安装 Google Chrome 或 Microsoft Edge 后重试；若均无，再执行：python -m playwright install chromium".

Programmatic Browser Launch

Import the helper in your own scripts:

from playwright.sync_api import sync_playwright
from skills.patent_search.tools.browser import launch_chromium, headed

with sync_playwright() as p:
    browser, label = launch_chromium(p, headless=None)
    print(f"Successfully launched: {label}")
    
    page = browser.new_page()
    page.goto("https://patents.google.com")
    print(page.title())
    browser.close()

The launch_chromium() function (lines 67–100) applies these default arguments:

• --disable-blink-features=AutomationControlled — reduces bot detection
• --no-sandbox — required for containerized environments

Troubleshooting Common Issues

Playwright not found

The playwright_installed() check (lines 40–45) raises a clear error suggesting pip install playwright.

All browser channels fail

If Chrome, Edge, and Chromium all fail to launch, launch_chromium() raises RuntimeError with a hint to install system browsers or run python -m playwright install chromium.

Permission errors in Docker

Ensure --no-sandbox is passed (automatic) and the container has sufficient /dev/shm size for Chromium.

Summary

• Install dependencies: pip install -r requirements.txt and optionally python -m playwright install chromium
• Control browser selection via PATENT_BROWSER_CHANNEL (chrome/msedge/chromium)
• Toggle visibility with PLAYWRIGHT_HEADED=1
• Verify with python -m skills.patent-search.tools.browser --probe
• Core logic lives in skills/patent-search/tools/browser.py with launch_chromium(), headed(), and playwright_installed()

Frequently Asked Questions

What browsers does patent-disclosure-skill support?

The skill supports Google Chrome, Microsoft Edge, and Playwright's bundled Chromium. The browser.py module prioritizes system-installed Chrome or Edge for faster startup, falling back to downloaded Chromium only when necessary.

Do I need to install Chrome or Edge if I have Playwright?

No. The bundled Chromium works standalone. However, system browsers launch faster and avoid the ~100MB download. If neither Chrome nor Edge is installed, the skill automatically uses chromium after running python -m playwright install chromium.

How do I run the browser in visible mode for debugging?

Set PLAYWRIGHT_HEADED=1 in your environment before running the skill. This forces launch_chromium() to launch with headless=False, showing the browser window for interactive debugging or visual confirmation.

What does the --probe flag do?

The probe mode starts a temporary browser instance and outputs JSON status to stdout. It is implemented in probe_launch() and called from main() when --probe is passed. Use it in CI pipelines or agent setups to verify browser availability before invoking full patent searches.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}