dsh-plugin-chrome

A DeepSeek Harness browser visualization plugin: opens a real, visible Chrome window per session, lets the agent drive the browser through the chrome_* tool suite, and streams the live view into a Chrome tab in the Web GUI — take over manually at any time.

中文文档

Features

• A visible window per session: every DSH session gets its own Chrome window (real window, not headless). Watch every agent action as it happens; the window uses an isolated user-data-dir, so it never mixes with your daily browser.
• Live view: the Chrome tab in the Web GUI streams the window through Chrome screencast (smooth while pages are active). A screenshot heartbeat keeps idle pages from freezing (about one frame every 3 seconds).
• Complete agent tool suite (16 tools): chrome_open / chrome_status / chrome_close / chrome_navigate / chrome_tabs / chrome_snapshot / chrome_screenshot / chrome_click / chrome_click_at / chrome_fill / chrome_type / chrome_press_key / chrome_hover / chrome_scroll / chrome_evaluate / chrome_wait.
• Accessibility-tree snapshots: chrome_snapshot returns a compact a11y tree with stable element uids; clicks and fills target uids directly — far lighter than DOM dumps and robust against fragile selectors.
• Dual-channel screenshots: chrome_screenshot sends the image into the model context (as an image block) AND saves it to the session's screenshot history shown in the panel.
• Security-minded: CDP never exposes a fixed port; the Web API enforces same-origin checks and a sessionId whitelist; browser data is isolated per session.
• Resource governance: idle windows auto-close (default 10 min, configurable), chrome_close closes explicitly, and plugin unload / host shutdown closes every window it opened.

Install

Prerequisites: DeepSeek Harness (DSH) installed, and Chrome or Edge on the machine.

# Option 1: install from GitHub (recommended)
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:jiaererw/dsh-plugin-chrome

# Option 2: local path (development)
npx -p @deepseek-ai/dsh dsh plugin --profile web add D:/harness/dsh-plugin-chrome

Restart DSH after installing — a Chrome tab appears at the top of every conversation.

Usage

For the agent (tools)

The agent gets the chrome_* suite automatically. Just ask it:

Open Chrome, go to https://example.com, take a screenshot, then click the "Login" button and fill in the username.

The agent will: chrome_open → chrome_navigate → chrome_screenshot (sees the image) → chrome_snapshot (gets uids) → chrome_click / chrome_fill.

For you (visualization)

1. Open the Chrome tab at the top of the conversation:
   • Live view: continuously shows the window. Native screencast frames flow while the page changes; a heartbeat fallback force-captures idle pages so the picture never freezes.
   • Tab management: switch or close tabs from the side list, in sync with the real window.
   • Manual takeover: click around in the Chrome window yourself at any time — the agent sees your changes on its next tool call.
2. Screenshot history: every chrome_screenshot is stored in the panel; click a thumbnail to enlarge.

Window lifecycle & resilience

…