dsh-steering
Run the following command in DeepSeek Harness:
dsh plugin install ManoloRemiddi/dsh-steering
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install ManoloRemiddi/dsh-steering inside DeepSeek Harness to install; source code is at https://github.com/ManoloRemiddi/dsh-steering
About this plugin
In a DSH multi-agent workflow, Steer corrections used to be stuck in the queue until the model finished its entire response before they could take effect. dsh-steering changes that: your correction now cuts directly into a generation-in-progress instead of waiting for the turn to close.
Under the hood, the same queued message ID is preserved. The stale request is canceled while pending work stays alive, and the correction wakes the driver ahead of ordinary follow-ups. If a tool is mid-execution, the correction waits for its safe step boundary before landing; already-completed actions are never undone or re-triggered. Internally this produces an aborted-turn boundary rather than spawning a new conversation.
The package ships behavior only, with no extra queue UI, no API keys, and no additional model calls. It mounts through host-provided lifecycle hooks onto the agents in your chosen DSH profile, making it a natural fit for developers who run local models, care about latency, and refuse to sit through a full monologue before their correction is heard.
Use Cases
- Correct a drifting model response mid-generation instead of waiting for the full turn
- Run local models with high inference latency and need immediate steering without idle wait
- Fine-tune Steer timing in multi-agent flows so corrections never queue behind a complete answer
Best For
- Developers deep in DSH multi-agent workflows who are latency-sensitive to Steer response
- Local-model users who want to cut the dead wait for a full turn to end
- Automation engineers who need precise control over correction timing in agent pipelines
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.