Foreword¶

DSH’s ecosystem philosophy is “everything is a plugin.” dsh-all-in is a recreation-oriented DSH plugin designed to fill the waiting time during agent thinking: it adds an All In action to the Harness sidebar, opening a full-screen local Texas Hold’em table without occupying the running agent.

What This Is¶

The repository for dsh-all-in is leeclouddragon/dsh-all-in, maintained by leeclouddragon, licensed under MIT. Its positioning can be summarized in one sentence: a local-only Texas Hold’em table that lets you play cards while waiting for the DeepSeek Harness agent.

It opens the shell.overlay slot supported by Harness, rather than a standalone page. The All In action in the sidebar is responsible for entering this full-screen table; once opened, the current agent continues running in the background.

Core Features¶

The following are capabilities that can be directly confirmed from verified materials:

• Six-player table: 1 human player and 5 local bots.
• Bot decisions use Monte Carlo bot equity, combined with pot-odds, opponent-count, position, stack-to-pot, and personality-aware factors.
• Opponent levels include Casual, Competitive, Shark-table, and experimental solver-style GTO, adjustable from the table header.
• Different bot personalities feature occasional action-aware table talk.
• Game flow covers Pre-flop, flop, turn, river, and showdown.
• Player actions include Fold, check, call, pot-sized raise, and all-in.
• Custom raise-to sizing provides Min, 1/2-pot, 3/4-pot, Pot, Max, slider, and direct amount entry.
• Starting stack is 5M Tokens, with blinds at 25K/50K.
• Supports seven-card hand evaluation, split pots, and side-pot settlement under unequal all-in contributions.
• Token stacks persist locally, surviving page reloads.
• UI displays live Agent thinking / Agent idle status.
• UI supports English and Simplified Chinese.
• Layout adapts to desktop and narrow-window configurations.

Bot equity estimation only uses its own hole cards and the public board; simulations do not inspect opponents’ hidden cards or the real deck order.

Installation and Enablement¶

The materials provide three installation paths: prebuilt GitHub Release, npm, and local checkout. Whether npm is currently published has not been confirmed in verified facts, so focus on GitHub Release and local checkout for now.

Installing from Prebuilt GitHub Release¶

First, run the installation command to add the v0.1.1 package to the web profile; then start dsh --profile web:

dsh plugin --profile web add https://github.com/leeclouddragon/dsh-all-in/releases/download/v0.1.1/dsh-all-in-0.1.1.tgz
dsh --profile web

Installing from Local Checkout¶

First, clone the repository and install dependencies, then run pnpm check for local verification; finally, add the current directory as a local plugin:

git clone https://github.com/leeclouddragon/dsh-all-in.git
cd dsh-all-in
pnpm install
pnpm check
dsh plugin --profile web add "$PWD"
dsh --profile web

When installing from GitHub source, the package includes a prepare script for building lib/; pnpm 10+ may require users to explicitly allow this build script.

npm Alternative¶

The materials provide the following command, but it is marked as “after the first npm release.” If the package is not yet published on npm, this command may not be available:

dsh plugin --profile web add dsh-all-in
dsh --profile web

Typical Usage¶

After installation and startup, use it as follows:

1. Run dsh --profile web to enter Harness.
2. Use the All In action in the sidebar to open the full-screen table.
3. Select the opponent level in the table header: Casual, Competitive, Shark-table, or experimental solver-style GTO.
4. Starting from Pre-flop, choose Fold, check, call, pot-sized raise, or all-in based on the situation.
5. When custom bets are needed, use Min, 1/2-pot, 3/4-pot, Pot, Max, the slider, or enter the amount directly.
6. Observe the Agent thinking / Agent idle status on the table to confirm the agent is still running in the background.

Applicable Scenarios and Notes¶

• Suitable for DSH users who want a light pastime while waiting for an agent locally.
• It is for entertainment only. Tokens have no monetary value, and there are no deposits, withdrawals, purchases, accounts, or multiplayer wagering.
• The game is local-only: it does not write to the conversation, does not call models, and does not use network connections.
• The plugin runs with the permissions of the current dsh process; review the source code and MIT license before installation to confirm it fits your environment.
• DeepSeek Harness is currently in developer preview, and the materials also mention that the contribution guide indicates external pull requests are not currently accepted.
• Whether the npm installation command has a currently available package has not been confirmed in verified facts.

Links¶

• GitHub: https://github.com/leeclouddragon/dsh-all-in
• Directory page: No confirmable directory URL is provided in the current verified facts; if you are using the DSH plugin directory, you can search for dsh-all-in.