dsh-token-meter-mood: DeepSeek Harness Web Plugin
The DeepSeek Harness (DSH) plugin `dsh-token-meter-mood` is designed to visualize token consumption when debugging agents. It displays detailed statistics below each assistant reply, including input, cache-hit, and output volumes, using Chinese units and grouping by four digits. It shows tiered mood emojis based on the number of output tokens (e.g., from 🥳 to 😡) and supports viewing artifacts and reasoning information. The installation command is `pnpm dsh plugin --profile web add "github:firekula/dsh-token-meter-mood"`, helping developers precisely optimize costs.
Read Moredsh-zhipu: Zhipu BigModel All-in-One Plugin
`dsh-zhipu` is the Zhipu BigModel plugin for DeepSeek Harness (DSH), providing unified web search, page reading, visual understanding, and image recognition capabilities via `ZAI_API_KEY`. The plugin relies on the Zhipu Coding Plan endpoint and allows custom tool toggles through a configuration file. Data is sent to `open.bigmodel.cn`, and search results should be cited with caution. After installation, add the package name to the profile bundle list and configure the LLM model to support image input to enable image recognition.
Read MoreDeepSeek Harness Plugin: dsh-evolving-memory
`dsh-evolving-memory` is a memory plugin for DeepSeek Harness, designed to address the problems of complex structure or poor integration capability in traditional memory modules. The plugin returns memory to editable Markdown documents, providing features such as searching, reading, saving, and cross-session consolidation (Dream Mode). The system adopts a “human-machine co-governance” model, giving users ultimate decision-making authority over memory files. It does not rely on schemas or state machines, and instead uses the file system to enable persistence and evolution of agent memory.
Read Moredsh-conversation-nav: Right-Side Question Navigation Plugin for Conversations
`dsh-conversation-nav` is a DSH Web UI plugin designed to address the challenge of locating user questions in long conversations. It renders a numbered vertical navigation bar on the right side of the conversation. Hovering over an item previews its content, while clicking smoothly jumps to and highlights the target message. The plugin supports automatic scroll-position highlighting, virtual scrolling, scroll-to-bottom behavior, and maid theme support, and is compatible with dynamic rendering. After installation, simply restart to start using it.
Read Moredsh-glm-vision: Integrating Zhipu GLM Vision Models into DeepSeek Harness
The DeepSeek Harness plugin `dsh-glm-vision` aims to address the lack of visual capabilities in local text models. By integrating Zhipu's GLM vision model, it provides image-viewing functionality for models such as DeepSeek. Users must install the plugin, configure a Zhipu API key, and restart DSH. It offers two usage modes—routing and tool-based—and supports local image paths and attachment IDs. By default, it takes over DeepSeek's official routing and automatically invokes GLM to analyze images when they are uploaded. Note that the plugin depends on `@deepseek-ai/schemastery` (>=3.18.0).
Read Moredsh-llm-longcat: A Long-Context and Thinking Mode Adapter for DeepSeek Harness
dsh-llm-longcat is a DeepSeek Harness (DSH) plugin designed to integrate the LongCat-2.0 model. It primarily resolves issues in the Harness ecosystem related to long-context (1M token) processing, structured chain-of-thought display, and native tool calling. The plugin supports streaming, multi-turn conversation reasoning review, and API key management via environment variables. Installation requires running a local script, and configuration should note that image input, stop sequences, and binary limits on reasoning strength are not supported.
Read Moredsh-local-diagnostics: A Privacy-First Local Diagnostics Plugin for DeepSeek Harness
`dsh-local-diagnostics` is a local memory diagnostics plugin for DeepSeek Harness (DSH). It leverages Cordis extension points to focus on collecting performance metrics (such as dispatch latency) and structured error codes without modifying the Agent Loop. It strictly follows a privacy-first principle and does not collect prompts, sensitive information, or make network requests. The plugin provides sanitized aggregate reports through the `diagnostics_report` interface. Installation requires Node.js and other environment prerequisites, and, after configuration, it can quickly troubleshoot tool-calling issues locally. The current version does not include a UI, and statistics are limited to the plugin process lifecycle.
Read Moredsh-learning-plugin: An Example Tool for Learning the Harness Plugin Mechanism
`dsh-learning-plugin` is an open-source plugin maintained by fengyufengzi, designed to demonstrate the plugin development mechanism of the DeepSeek Harness ecosystem. It registers a `learning_greet` tool that illustrates four core extension points: declaring service dependencies, validating deployment configurations, registering automatic revocation capabilities in the lifecycle, and deriving parameter and output types from schemas. The project includes complete type checking, testing, and release workflows. After installation, the tool can be invoked in a conversation with the `Use learning_greet to greet Ada` command. Note that Harness is in developer preview, so its API may undergo breaking changes; this plugin is intended only for learning and demonstration, and should not be used in production environments.
Read Moredsh-file-bridge: DeepSeek Harness Bidirectional File Exchange Bridge
The DSH (DeepSeek Harness) File Bridge plugin provides file transfer capabilities in conversational scenarios. It supports bidirectional file exchange through three methods: the attachment button, tool-based delivery, and path navigation. Features include automatic image compression, native preview, static file serving, and protection against path traversal. Installation requires running the specified command and restarting the service, followed by configuring the public URL and storage directory in the configuration file. During use, users can click the attachment button to upload files, while the Agent can call a tool to deliver files and reference source file paths wrapped in backticks within replies to invoke the sidebar manager. When deploying, take note of Nginx request body size limits and file selection compatibility issues on the app client.
Read Moredsh-session-icons: DSH Session Title Icons Plugin
`dsh-session-icons` is a helper plugin for DeepSeek Harness (DSH) designed to reduce clutter in session list titles. The plugin leverages AI to automatically generate 24x24 monochrome SVG icons for each session and inject them into the sidebar to support visual identification. Core features include icon generation and injection, a disk-based caching mechanism (with support for sharing under the same name), and output safety constraints. The plugin depends on the `dsh web` environment, works with default configuration, and supports adjusting the cache count and timeout via the configuration file.
Read Moredsh-thirteen-bg: Dynamic Background Plugin for DeepSeek Harness Web GUI
dsh-thirteen-bg is a Live Wallpaper plugin for DeepSeek Harness, designed to provide dynamic background effects to the Web GUI. It supports GIFs, videos, and various static image formats, utilizing HTML5 video tags and CSS rendering, along with a client-side local storage mechanism, and includes a built-in dimming slider to ensure text readability. After installation, the service must be restarted; the background can be configured by uploading a local file or pasting a URL. It does not occupy pseudo-elements and has zero conflicts with skin plugins.
Read MoreDSH Plugin: dsh-skill-viewer (Desktop Skill Management Tool)
`dsh-skill-viewer` is a Windows desktop skill management tool designed for the DSH Web UI. Developed by Failing-coachman563, it aims to simplify the process of viewing, adding, deleting, modifying, and controlling skill states. The app supports group management by workspace, offers list views, one-click toggling, and batch operations, and includes a CLI tool. Users must manually download, extract, and install it; currently only Windows is supported. Since deletions are irreversible, it is recommended to keep a record before deleting.
Read MoreDeepSeek Harness Plugin dsh-diffstream: Show Code Changes Directly in the Conversation Flow
dsh-diffstream is a Web plugin for DeepSeek Harness (DSH) designed to address the difficulty of tracking code changes. It displays line-level diffs in real time within the conversation flow, supports streaming animations and end-of-turn change summaries, and is compatible with Git or pure parameter-based comparisons. By taking over the rendering logic, the plugin achieves zero-latency display, offers simple installation and configuration, and greatly enhances the visual experience of debugging agent-generated code.
Read Moredsh-authorize-app: The "Connected Apps" Center in the DeepSeek Harness Plugin
`dsh-authorize-app` is a DeepSeek Harness (DSH) plugin designed to address the cluttered interface that occurs when plugins display connection status and feature entry points in the settings panel. It provides a standard unified “Connected Apps” block, allowing other plugins to register cards through a slot mechanism. The plugin adds a “Connected Apps” entry to the settings page and supports bilingual empty states and card ordering. Developers simply need to register card components by injecting slots to appear in the unified list, eliminating the need to manage a registry manually. During installation, note the required Node.js version and `next` version tag.
Read MorePolaris: General Capability Aggregation and Selective Evolution Layer
Polaris is a general-purpose capability aggregation layer within the DeepSeek Harness (DSH) ecosystem, designed to address the high costs associated with managing plugins and tools. Through heterogeneous probing, dual-track arena validation, SLSA signing, and immutable indexing, it provides a complete closed loop from discovery and validation to invocation. Polaris supports automatic registration of model tools and atomic CLI primitives, enabling developers to transparently dispatch requests to the optimal implementation in a stateless, low-cost manner, and is particularly well suited for science, engineering, and code optimization scenarios.
Read MoreDeepSeek Harness Plugin Health Check Tool: dsh-perfscope
As the DSH plugin ecosystem expands, plugin management becomes increasingly difficult. `dsh-perfscope` is designed to provide unified health checks and scoring. This open-source tool evaluates plugin health (0–100 score) through local scans, supports offline analysis, generates Markdown/HTML reports, and offers one-click remediation. Users can install it via the official command and use the CLI to scan, disable faulty plugins, or revert fixes. The project requires Node.js 22.19.0+ and DSH 0.1.0-rc.8, with all data processed locally.
Read Moredsh-llmwiki: A Local Markdown Wiki for DeepSeek Harness
dsh-llmwiki is a local-first Markdown Wiki plugin for DeepSeek Harness (DSH), designed to provide a deterministic storage and retrieval solution. The plugin leverages the local file system, SHA-256 hashing, and the BM25 algorithm to ensure that data is immutable, privacy-safe, and free of external dependencies. Users must explicitly activate the plugin via a patch and configure optional parameters to manage the Wiki root directory and file size limits. Once installed, the plugin can be used by agents as services, tools, and commands, making it suitable for development scenarios that require offline operation.
Read MoreDeepSeek Harness Peak and Off-Peak Billing Plugin
This plugin is specifically designed for DeepSeek Harness (DSH) to monitor and calculate token usage and costs in real time under DeepSeek's official peak/off-peak pricing policy. It supports viewing whale ASCII-style bills via the `/billing` command in the TUI status bar or during conversations, and includes built-in agent tools (`show_billing` and `check_pricing_period`). Billing strictly follows official rules, applies only to the `deepseek-v4` series, uses time-based pricing across peak and off-peak periods (50% discount), and supports session inheritance and deduplication. Installation is simple, ready to use out of the box, and requires no additional configuration.
Read Moredsh-official-document-mode: DeepSeek Harness Official Document Mode Plugin
`dsh-official-document-mode` is a DeepSeek Harness (DSH) plugin designed to provide automatic formatted rendering and sign-off feedback for official document agents. It injects style constraints on the host side and performs client-side DOM rendering, transforming assistant replies into a standard official document layout that includes a document header, a document number, a red separator line, and a signature block. The plugin also supports a 36px drawing canvas on the left side, using a geometric heuristic algorithm to recognize sign-off strokes and automatically write approval or disapproval feedback. Users can adjust parameters such as the document number and signature block through a configuration file and switch modes with a toggle. This plugin resolves technical issues such as segmented rendering, title extraction, and hot reloading that required restarting the desktop client, offering a closed-loop solution for official document processing from generation to sign-off.
Read MoreIntroducing the DeepSeek Harness “Xiao Ji” Plugin
The DSH Harness plugin `dsh-notes` is designed to address the pain points of switching between conversational flows and external tools. It provides a persistent sidebar that integrates persistent session cards, context notes with draggable partitions, a multi-state todo section, and a freeform note area. The plugin supports both global and workspace scopes, automatically persists data, adapts to light and dark themes, and uses an agent isolation mechanism to ensure that the content is invisible to the model. Installation requires running the specified command and restarting, with data stored in `~/.dsh/storages/notes.json`.
Read Moredsh-code-pipeline: Injecting Stage Sub-Agent Tools into PTC Code Mode Pipelines
`dsh-code-pipeline` is a dynamic pipeline management tool provided as a DSH bundle plugin, designed to address the rigid pipeline model configuration in PTC Code Mode. The plugin dynamically injects sub-agent tools by listening to events and enables runtime model configuration via the settings page without restarting the session. Core features include dynamic model configuration, background execution mode, multi-round review reuse, and on-disk message persistence. It also provides the `pipeline_followup` tool for handling requirement changes during multi-round reviews and supports manual gate approval at the Plan stage. The plugin requires Node.js >= 22.19.0 and dsh >= 0.1.6-alpha.1.
Read Moredsh-system-prompt: DeepSeek Harness Real-Time Session Prompt Inspection Plugin
The `dsh-system-prompt` plugin for DeepSeek Harness (DSH) is designed to help developers debug agents through the "everything is a plugin" philosophy. It is a read-only tool that provides a browser interface and an API for statically and dynamically inspecting the current session's effective prompt, context, variables, and tool configuration. It includes two main tabs: the "Conversation Inspection" tab for displaying the components of the prompt, and the "Sections Details" tab implemented via DOM bridging. The plugin exposes a JSON projection of session data through the `/dsh-system-prompt` endpoint, ensuring that live objects are not serialized. Installation is simple; local development requires running a build and aims to reduce the cost of obtaining debugging information.
Read Moredsh-digital-life: Adding Configurable Digital Life Agents to DeepSeek Harness
The DSH plugin `dsh-digital-life` provides a solution for managing conversational digital life agents, supporting the persistent storage, import, and management of agent identities. Users can create agents through the settings interface and invoke them using tools in separate sessions or the main conversation. Core features include single consultation, concurrent consultation by category, and read-only sessions without a Workspace. When using the plugin, note the build dependencies and category limitations.
Read MoreDSH Image-to-Text Input Plugin: Intercept Image Input for Text-Only LLMs
The dsh-plugin-image-input plugin converts images into structured textual descriptions by locally invoking a vision API, enabling pure text large models (such as DeepSeek R1) to handle image input. Users only need to paste or drag and drop images; the plugin can automatically or manually convert them into textual context sent to the model, with support for multiple images. After installation, an OpenAI-compatible vision API must be configured. The plugin features multimodal compatibility, fault tolerance, and safety restrictions, effectively resolving request failures caused by the model's inability to recognize images.
Read MoreDeepSeek Harness (DSH) Plugin: dsh-blue-sea-player (Whale Music Player)
`dsh-blue-sea-player` is a resident music player plugin for DeepSeek Harness (DSH), providing a non-intrusive, draggable playback interface. Key features include a rotating CD animation, rhythmic line effects, full playback controls, and playlist management. The plugin supports switching between 12 skins, offers a minimize-and-hide function, and provides asset management with batch operations. After installation, restart DSH Web to use it; data is stored in a local directory.
Read More