dsh-plugin-local-model
Local GGUF model manager for DeepSeek Harness with on-demand llama.cpp loading and auto-unload.
Install
$ dsh plugin --profile desktop add github:DZQJOKER/dsh-plugin-local-modelPlugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Dedicated 'Local Model' settings panel within the DeepSeek Harness settings sidebar
- On-demand llama-server process launch on first request and automatic idle unload after 5 minutes
- Parameter presets management allowing one-click switching of model and inference configuration profiles
- Support for multi-token prediction (MTP) speculative decoding and multimodal vision projections (mmproj)
- Automatic command-line argument probing and graceful fallback for upstream llama.cpp and kvmem branches
- Request rewriting for thinking chain support (chat_template_kwargs, reasoning effort mapping, and history think stripping)
- Protection against context overflow and automatic output token headroom adjustments
Useful For
- Running offline local GGUF models seamlessly inside DeepSeek Harness without constant VRAM occupation
- Switching between different model tuning profiles and vision/reasoning presets with a single click
- Executing vision-capable local LLMs or MTP-accelerated local models through the DSH interface
Who It Fits
- DeepSeek Harness users running local GGUF models on GPUs or CPUs
- Developers seeking granular control over llama.cpp inference parameters and VRAM lifecycle
- Users running custom llama.cpp forks such as kvmem
Documented Limitations
- Requires users to provide their own GGUF model files and llama.cpp/kvmem runtime binaries
- Restricted to single active local model instance at a time
- Settings API bridge strictly listens to loopback (localhost) and rejects non-local/remote web requests
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 0 runtime, 3 development, 3 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- DZQJOKER/dsh-plugin-local-model
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- bee0024 · 2026-10-27
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 4GitHub stars
- Forks
- 1
- Open issues
- 1
- Last commit
- 2026-09-29
- Last release
- 2026-09-29
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Multi-account failover model pool, quota aggregator, and automation suite for DeepSeek Harness.
View plugin →Free frontier LLM provider plugin for DeepSeek Harness with upstream probing and local OpenAI-compatible gateway.
View plugin →Command Code LLM provider bundle for DeepSeek Harness with model catalog, quota tracking, and multi-account rotation.
View plugin →Use ChatGPT, Claude, Grok, GitHub Copilot, and Antigravity subscriptions as DeepSeek Harness LLM providers via OAuth.
View plugin →