dsh-thinking-levels
Dynamic thinking-level and reasoning effort controller for DeepSeek Harness with automated round scheduling.
Install
$ dsh plugin --profile web add dsh-thinking-levels -wPlugin 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
- Injects an Auto reasoning effort mask into the DSH model selector to dynamically scale thinking levels based on tool execution complexity
- Provides manual thinking level control across off, on, minimal, low, medium, high, xhigh, and max settings
- Applies safety guards to strip unsupported reasoning_effort fields from non-reasoning or custom gateway models to prevent API 400 errors
- Offers a segmented effort slider UI with support for animated custom runners and immediate release snapping
- Enables quick context-window limit configuration (presets from 64K to 1M or custom limits) synced directly with llm-pi-ai and llm-deepseek
- Includes a built-in token pricing and billing catalog across multiple Chinese and regional model offerings
Useful For
- Minimizing wall-clock latency and token consumption during simple, multi-step agent tool loops without manual intervention
- Escalating thinking depth automatically when heavy tool invocations or complex data transformations occur
- Fixing compatibility issues when integrating third-party OpenAI-compatible providers or local vLLM gateways with DSH
- Adjusting context-window pressure thresholds and compacting limits directly from the composer interface
Who It Fits
- DeepSeek Harness developers building automated multi-turn agent workflows
- Engineers running local or self-hosted models (vLLM, SGLang, LM Studio) alongside DeepSeek Harness
- DSH users seeking finer control over reasoning token usage and API latency
Documented Limitations
- Pinned tightly to specific DSH host versions (rc-level peer dependencies between 0.1.0-rc.8 and 0.2.1-rc.2)
- Japanese and Korean localizations require host-level locale registration in DSH to display without errors
- Auto reasoning mask only schedules between low, high, and max; it never automatically selects off
- Model UI takeover features default to off and require explicit opt-in via configuration
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 2 runtime, 18 development, 8 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
- drscrewdriver/dsh-thinking-levels
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin
- Source checked
- fa95437 · 2026-10-19
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
- 12GitHub stars
- Forks
- 1
- Open issues
- 4
- Last commit
- 2026-10-19
- Last release
- No release detected
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
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 →Free frontier LLM provider plugin for DeepSeek Harness with upstream probing and local OpenAI-compatible gateway.
View plugin →Interactive reasoning-effort slider and runner visual integration for DeepSeek Harness model selectors.
View plugin →