model-routing 22

Local-model (Ollama) integration for DeepSeek Harness: discover, pull, remove, and inspect local models, route requests to them by task type or keyword with automatic fallback to the cloud, and get a one-shot status overview via /ollama.

Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens

kimi-tide（月汐）— preset-and-rule model router for DeepSeek Harness（预设+规则驱动的逐步模型路由）：named presets with ordered image/keyword-group rules route each agent step across ANY mounted provider/model (candidate pool = full host catalog enumeration, no whitelist), w

Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution — prefercmd/memory as knowledge allow/deny lists, so repeated tasks start from proven commands and save tokens

AI Services: shared model routing, bounded auxiliary calls, and usage receipts for other plugins.

为 DeepSeek Harness 子代理配置默认模型：多模型顺序/随机分配、跨供应商连接失败故障转移，并在轨迹/对话视图显示当前供应商与模型，同时以一行真实提示词把当前路由注入子代理上下文。

Two-level model circuit breaker with failover for DeepSeek Harness: trip a model or a whole provider after repeated request failures and route the next request to a configured fallback

Two-tier model router for DeepSeek Harness: an LLM Judge sends each turn to the cheap Fast chain or the Smart chain by expected cost, with multi-model fallback chains, exponential-backoff failover, cache-aware switching and task-level orchestration. A DSH

Unified ModelID routing for DeepSeek Harness: one logical model id over multiple providers, first-token failover with cooldown, health-aware ranking, fallback reasoning efforts, purpose-based simple/complex task split, per-tier reasoning effort, and a set

DeepSee: DeepSeek Harness vision, multi-model routing, and Gemini visual self-checks before delivery

Semantic model routing plugin for DeepSeek Harness

Model-routed subagent delegation for DeepSeek Harness: a `subagent_model` tool with per-call provider/model/max_tokens overrides and a built-in `model: "auto"` routing policy (anchored to the parent's own model by default; task-tier upgrades, failure esca

Per-turn model and reasoning-effort routing for DeepSeek Harness -- one Jev choice question per turn

DeepSeek Harness plugin that asks which provider and model should run each delegation before the child starts, so a subagent can go to a cheap or local route instead of the parent's.

Role-based model manager for DeepSeek Harness: assign models to planning, execution, vision and main roles, route subagent work by role, and report which model every subagent used.

Task-aware automatic model router (auto route) for the DeepSeek Harness: picks a backend model per request (vision / long-context / default) and falls back across configured models on clean pre-content failures.

Automatic tier-based model routing for DeepSeek Harness: a virtual `smart` model classifies each request by difficulty (hard / normal / easy) and by vision need, then delegates it to the models you already configured. 三级难度 + 视觉自动路由，虚拟 smart 模型零配置接入。

Model Center: configure model tiers, provider connections, and AI call limits.

DeepSeek Harness plugin: role-based model routing — both roles default to deepseek-flash (V4.1 Flash, native multimodal) until V4.1-Pro launches.

Model authorization gate: subagent model/provider pulls must be in the user-controlled allowlist. 模型授权闸：子代理拉模型必须 ∈ 用户 allowlist

Shows which backend model actually served each request when using the auto route.

DeepSeek Harness plugin: learning agentic router — rule + UCB + LinUCB + k-NN experts under an EXP3 meta-selector, with a durable quality-reward flywheel, REAL model switching via the session request-header precedence chain, and STEP-LEVEL routing (v1.5.0