@neptune810/dsh-model-router
仅限 Flash 的 DeepSeek Harness 推理强度路由：根据每个步骤的形态进行分类，并相应设置 DeepSeek flash 模型的推理强度。模型不会改变；最高强度需主动启用。
插件会安装到这里；不确定时保持 web。
npx -y @deepseek-ai/dsh plugin --profile web add github:Neptune810/dsh-model-router#32bd6fae66ed51c75b9c317dd0efe7fdabf5e4c5Model Router 以 @neptune810/dsh-model-router 发布，当前版本为 0.3.0。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。
点击下方 Badge 复制 Markdown，粘贴到 README 即可。
验证 package.json 声明的 GitHub 仓库，即可管理这个公开页面。认领后，Hub 会优先安排当前版本的安全扫描，并在通过后公开展示结果。
认领这个 Plugin →English | 中文
A DeepSeek Harness plugin that sets the reasoning effort of the DeepSeek flash model per step. The model never changes — this plugin only decides how hard a step should think.
Host-only: no browser UI, no client bundle. It runs silently in the background.
| Step class | Decided by | Reasoning effort | Internal point |
|---|---|---|---|
trivial | a clearly cheap intent (translate, rename, reformat) in a short step | off | thinking disabled |
standard | a plain short request with no engineering cue | low | ~50 |
engineering | engineering cues, code/diff/XML structure, or an agent tool loop | high | ~75 |
hard | a dense engineering brief, or failures earned inside the task | high (max when allowMax) | ~75 / 100 |
high is the default ceiling. V4.1-Flash drives effort from an internal 1–100 scalar and exposes
three preset names; high corresponds to roughly 75, which is where the published effort curve is
still steep. Pushing past it costs about 1.6–1.8x the output tokens for a marginal gain.
scoring.turnPerPoint: 0) and tool
calls are counted per task, so a long agent run does not drift toward the most expensive effort.
A long run is not a harder task.escalateOnErrors failing tool results,
or the same tool call retried with identical arguments escalateOnRepeats times, step the class
up — at most maxEscalations times. Nothing else moves it.max is opt-in. Automatic routing never emits max unless allowMax is set. The ceiling is
enforced at the effort level, so even a hand-written route table asking for max is clamped.max is demoted too (demoteManualMax), but only for models this plugin
manages. Other models are left alone. Enabling allowMax turns both clamps off.Task boundaries come from agent/inbox/claimed, which is what actually opens a new piece of work.
When a task ends on an unresolved failure, the next one inherits a single hesitant step up — and only
if it is engineering or hard work. A one-line "translate this" never inherits a crash.
The plugin reads its config from its row in the profile's cordis.patch.yml:
- id: model-router
config:
mode: auto # auto | off
model: deepseek-flash
allowMax: false # true enables max for auto routing and manual selection alike
maxFallback: high # where max collapses when allowMax is false
escalateOnErrors: 2 # failed tool results needed to step up
escalateOnRepeats: 3 # identical retries needed to step up
scoring:
turnPerPoint: 0 # raise this to let long sessions weigh more (not recommended)
routes:
trivial: { effort: off }
standard: { effort: low }
engineering: { effort: high }
hard: { effort: max }
DSH_MODEL_ROUTER=off|auto overrides mode at boot.
| Key | Default | Meaning |
|---|---|---|
mode | auto | off passes every request through, except the manual-max clamp |
provider | deepseek-official | only this provider is ever touched |
model | deepseek-flash | the one model this plugin drives |
familyPattern | ^deepseek-(flash|v4) | conversation models it may take over; a pro session is pulled back to flash |
allowMax | false | whether max is reachable at all |
maxFallback | high | what max collapses to |
demoteManualMax | true | also demote a manually selected max |
leaveImageSteps | true | steps carrying images keep the caller's model |
imagePolicy | keep | set to flash to route image steps too (the flash model has native vision) |
escalateOnErrors | 2 | failure-evidence threshold |
escalateOnRepeats | 3 | repeated-call threshold |
maxEscalations | 2 | most classes a single task may climb |
carryUnresolved | true | carry one step up from an unresolved failure |
dsh plugin is a pnpm forwarder, so any pnpm specifier works. From the npm registry:
dsh plugin --profile web add @neptune810/dsh-model-router
or straight from the repository:
dsh plugin --profile web add github:Neptune810/dsh-model-router
Restart dsh web afterwards. Because the router registers listeners at boot, a reload is not
enough.
@deepseek-ai/dsh 0.1.5-rc.2. The plugin uses agent/request,
agent/inbox/claimed, and session.deriveMessages(). No engines.dsh range is declared, so the
plugin market keeps this entry visible rather than guessing it incompatible.@deepseek-ai/* packages, which a plugin installed beside the profile cannot resolve.deepseek-official provider and models matching familyPattern.node --test
36 tests. test/policy.test.js (26) covers classification, the absence of a ratchet, the
unreachable max, evidence escalation, effort clamping, and tool-result error parsing;
test/plugin.test.js (10) drives the host wiring with ctx/agent doubles — registering listeners,
claiming messages, routing each step, pulling a pro conversation back to flash, and demoting a
manually selected max.
MIT