dsh-llm-mimo
Run the following command in DeepSeek Harness:
dsh plugin install dfhxxc666/dsh-llm-mimo
Paste the following prompt into your AI chat to install this plugin:
Install the dsh-llm-mimo plugin in DeepSeek Harness from https://github.com/dfhxxc666/dsh-llm-mimo by running dsh plugin install dfhxxc666/dsh-llm-mimo.
About this plugin
dsh-llm-mimo plugs Xiaomi MiMo v2.5 into the DeepSeek Harness LLM seam so you can call MiMo-V2.5 and MiMo-V2.5-Pro from within your existing Harness setup. Compared to the upstream repository, this fork resolves several defects that caused MiMo to return HTTP 400 INVALID_REQUEST: truncated JSON arguments left in conversation history after interrupted tool calls, a maxTokens default that exceeded MiMo 131072-token hard limit, and a missing prepareCall hook required by dsh-llm 0.1.1-rc.2. Together these fixes ensure multi-turn tool-calling conversations replay cleanly without triggering invalid-request errors.
On the capability side, the adapter fully recognises MiMo reasoning_content thinking chain and translates it into the Harness ReasoningBlock format, preserves assistant-side reasoning context on tool-calling turns as MiMo requires, serialises tool results into OpenAI-compatible role: "tool" messages, and supports complete SSE streaming with proper block assembly. It also passes through sampling parameters (top_p, frequency_penalty, presence_penalty) and ships engineering optimisations such as a 60-second API-key cache, fetch keepalive for connection reuse, and a dynamic stream-idle timeout capped at 450 seconds when thinking mode is active, balancing latency against reliability.
Best suited for developers and teams who want to add Xiaomi MiMo v2.5 to their DeepSeek Harness model lineup, especially those whose workflows depend on thinking-chain visualisation, multi-turn tool calling, and streaming inference. A single MiMo API Key is all you need; after installation, simply select the model from the GUI picker and start working.
Use Cases
- Invoking Xiaomi MiMo v2.5 in DeepSeek Harness for complex reasoning tasks
- Maintaining thinking-chain and context integrity in multi-turn tool-calling workflows
- Real-time streaming inference requiring low latency and complete block assembly
Best For
- Developers using DeepSeek Harness who want to integrate MiMo models
- AI application teams relying on thinking-chain visualisation and tool calling
- Tech leads in production environments needing low-latency streaming inference
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.