Reasoning-effort tier manager for custom API models plus cross-session preference memory: inline low/medium/high/max toggles inside Settings → Models, with atomic settings writes.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-model-memory
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:Mutx163/dsh-model-memory
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
dsh-model-memory is a plugin for DeepSeek Harness (DSH) that provides reasoning effort management, channel-level retry policy control, and cross-session preference memory for custom API models.
🌟 Key Features
1. 🧠 Custom Model Reasoning Effort Configuration
- Seamless UI Integration: Automatically mounts inline reasoning effort controls directly under custom model items inside DSH Settings -> Models.
- Effort Tiers Selection: Enables
supportsReasoningEffortand allows selecting supported tiers (low/medium/high/max). - Instant Persistence: Saves atomically into
~/.dsh/settings.yamlwith immediate effect.
2. ⚡ Cross-Session Preference Memory
- Per-Channel Memory: Accurately remembers the last chosen model and reasoning effort (e.g.
max) per custom provider channel. - Auto-Fill on Session Start: Automatically restores your preferred model and thinking tier when creating new sessions or switching channels.
3. 🔁 Channel-Level Request Retry Policy
- Three Modes: Channel default (up to 5 retries after a failure), unlimited retries (keeps retrying until success or cancellation), or a custom count (0 disables retries).
- Applies per Channel: DSH's
retryPolicyis a channel-level setting shared by all models inside the channel; updates preserve the channel's existing backoff parameters and reuse the same write path as reasoning effort (host settings service first, atomic direct file write as fallback).
4. 🛡️ Lightweight & Upgrade-Immune
- Zero UI Clutter: Clean integration within settings pages without adding stray buttons to the main view.
- Upgrade-Safe: Retains your configuration and local plugin across global DSH CLI upgrades.
🚀 Installation
Option 1: Install from NPM (Recommended)
dsh plugin --profile web add dsh-model-memory
Option 2: Local Development with Link
# 1. Clone repository
git clone https://github.com/Mutx163/dsh-model-memory.git
cd dsh-model-memory
# 2. Install dependencies & verify
pnpm install
pnpm run verify
# 3. Mount to DSH Web Profile
dsh plugin --profile web add link:/path/to/dsh-model-memory
📂 Project Structure
dsh-model-memory/
├── src/
│ ├── index.ts # Plugin lifecycle and Agent default model wrapper
│ ├── memory.ts # Model & reasoning effort memory algorithm engine
│ ├── rpc.ts # /model-memory bidirectional RPC channel
│ ├── settings.ts # settings.yaml atomic configuration manager
│ ├── store.ts # Disk persistence engine (~/.dsh/model-memory.json)
│ ├── types.ts # TypeScript definitions
│ └── client/
│ └── index.ts # Web UI client components
├── test/ # Vitest unit test suites
├── cordis.patch.yml # Cordis patch layer
├── LICENSE # MIT License
├── README.md # Chinese Documentation (Default)
└── README_EN.md # English Documentation
🧪 Development & Testing
# TypeScript typecheck
pnpm run typecheck
# Run unit tests
pnpm run test
# Full verification & build
pnpm run verify
📄 License
Licensed under the MIT License.
Links
More in this category
V1ki/dsh-plugin-subscriptions★ 321
Use ChatGPT (Codex), Claude, and Grok subscriptions as DeepSeek Harness LLM providers, with Settings login, model catalogs, usage, plus image_generate, video_generate, and x_search tools.
Mars-Sea/dsh-commandcode-provider★ 159
Unofficial Command Code LLM provider: registers a `commandcode` route with a live model catalog and reasoning-effort support.
volcengine/ark-cli#ark-plan-api★ 118
Registers Ark Agent Plan, Coding Plan and postpaid model routes in the native DSH model picker.
Stormycry-cryp/dsh-AuthInOne★ 105
Adds account login, API and custom Provider setup, model switching, image fallback for text-only models, and token/cost attribution to DeepSeek Harness 47f.
franksong2702/dsh-codex-connect★ 82
Connect ChatGPT OAuth and OpenAI Codex models to DeepSeek Harness, with opt-in search and image tools.
WSL043/dsh-codex-subscription★ 42
ChatGPT OAuth provider for Codex models with subscription web search, quota and safe reset, image tools, Fast mode, and model-aware context; no API key or Codex CLI.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.