dsh-local-llm-controller
Control local llama.cpp instances and integrate local LLMs directly into DSH sessions with dual-slot management and custom compaction.
Install
$ dsh plugin --profile web add <地址>Plugin 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
- One-click start, stop, and process lifecycle management for local llama-server directly from Settings
- Dual-slot model management (A/B) with automatic scanning for GGUF model files and mmproj vision adapters
- Configurable preset launch parameter groups across slot, modality (text/vision), and context length (fast/long)
- Automatic registration of running local models into DSH model providers for session selection
- Proportional-threshold local compaction engine submodule optimized for 32k to 150k context windows
Useful For
- Running offline or private local GGUF models as conversational backends inside DeepSeek Harness
- Switching seamlessly between text and vision models hosted on a local machine
- Managing llama-server execution parameters and context budgets without leaving the DSH web UI
- Preventing aggressive compaction triggers on smaller local context windows using the specialized compaction engine
Who It Fits
- DSH users looking for private, offline local LLM inference via llama.cpp
- Developers testing and benchmarking quantized open-weight models within DSH
- Local AI practitioners running hardware-constrained or specialized GGUF models
Documented Limitations
- Does not include the llama.cpp server binaries or model weights; both must be provided locally
- Replacing model files requires manual removal of old provider entries in DSH model settings
- Older llama-server builds lack WebP image support (requires conversion to PNG/JPEG)
- Custom compaction engine requires manual integration into custom agent mode presets
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, 0 development, 10 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
- Lbunc/dsh-local-llm-controller
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- edd6bf9 · 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
- 10GitHub stars
- Forks
- 0
- Open issues
- 0
- Last commit
- 2026-10-23
- Last release
- 2026-10-23
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Use ChatGPT, Claude, Grok, GitHub Copilot, and Antigravity subscriptions as DeepSeek Harness LLM providers via OAuth.
View plugin →Generative UI plugin for DeepSeek Harness that renders interactive cards, charts, forms, and panels inline from dsh-ui code fences.
View plugin →Infinite canvas UI plugin for DeepSeek Harness that turns branching LLM conversations into an editable thought graph.
View plugin →Command Code LLM provider bundle for DeepSeek Harness with model catalog, quota tracking, and multi-account rotation.
View plugin →