pi-cache-optimizer 2.5.3 → 2.5.4

package/README.md

@@ -8,8 +8,6 @@
 
 Pi extension for improving provider-side KV / prompt cache hit rates. It keeps stable prompt content near the front, adds a conservative OpenAI-compatible `prompt_cache_key` fallback, warns about common proxy cache-routing gaps, and shows read-only footer cache stats.
 
-**GitHub About:** Improve Pi prompt/KV cache hit rates with stable prompts, OpenAI-compatible cache keys, proxy compat warnings, and footer cache stats.
-
 > Renamed from `pi-deepseek-cache-optimizer`. Existing footer counters migrate automatically. This package never creates, edits, backs up, or deletes your `~/.pi/agent/models.json`.
 
 ## Contents
@@ -100,6 +98,7 @@ Notes:
 - `supportsLongCacheRetention: true` is optional. Add it only when the endpoint explicitly supports OpenAI long prompt cache retention.
 - If you see `400 Unsupported parameter: prompt_cache_retention`, remove/avoid `supportsLongCacheRetention` for that channel. Keep `sendSessionAffinityHeaders` if supported.
 - Use `/cache-optimizer compat` or `/cache-optimizer doctor` to see model-specific advice.
+- For DeepSeek models, the Pi Mono guidance expects `compat.requiresReasoningContentOnAssistantMessages: true` and `compat.thinkingFormat: "deepseek"` alongside cache/session-affinity flags when the endpoint supports them.
 - This extension only advises; it does not edit `models.json`.
 
 ## Footer stats
@@ -114,7 +113,7 @@ OpenAI cache 3/10 · 0.002M/0.005M tok (40%) ⚠️ compat
 
 Format: `<label> <hit requests>/<total requests> · <cached input tokens>/<total input tokens> tok (<token hit rate>)`. Some adapters may also append `· write <tokens> tok`, and runtime diagnostics may append `⚠️ compat` or `⚠️ integrity`.
 
-Supported footer labels include: DS, Claude, OpenAI, Gemini, Kimi, Qwen, GLM, MiniMax, Hunyuan, Mistral, Grok, Llama, Nemotron, Cohere, Yi, Doubao, ERNIE, Baichuan, StepFun, Spark, InternLM, Gemma, Phi, Jamba, Solar, Sonar, Nova, Reka, Falcon, DBRX, MPT, StableLM, Aquila, EXAONE, HyperCLOVA, Luminous, Hermes, Granite, Arctic, Pangu, SenseNova, Zhinao, MiniCPM, XVERSE, Orion, OpenChat, Vicuna, Wizard, Zephyr, Dolphin, OpenOrca, Starling, BLOOM, RWKV, and Aya.
+Supported footer labels include: DS, Claude, OpenAI, Gemini, Kimi, Qwen, GLM, MiniMax, Mimo, Hunyuan, Mistral, Grok, Llama, Nemotron, Cohere, Yi, Doubao, ERNIE, Baichuan, StepFun, Spark, InternLM, Gemma, Phi, Jamba, Solar, Sonar, Nova, Reka, Falcon, DBRX, MPT, StableLM, Aquila, EXAONE, HyperCLOVA, Luminous, Hermes, Granite, Arctic, Pangu, SenseNova, Zhinao, MiniCPM, XVERSE, Orion, OpenChat, Vicuna, Wizard, Zephyr, Dolphin, OpenOrca, Starling, BLOOM, RWKV, and Aya.
 
 Adapter selection uses only model id/name (plus assistant message model/name on message end). Generic OpenAI-shaped APIs are not treated as OpenAI-family unless the model id/name matches a supported family.
 

package/README.zh-CN.md

@@ -8,8 +8,6 @@
 
 用于提升 Pi 中 provider 侧 KV Cache / Prompt Cache 命中率的扩展：把稳定 prompt 内容前置，给 OpenAI-compatible 请求补保守的 `prompt_cache_key`，提示代理渠道常见缓存路由兼容问题，并在底部显示只读缓存统计。
 
-**GitHub About：** Improve Pi prompt/KV cache hit rates with stable prompts, OpenAI-compatible cache keys, proxy compat warnings, and footer cache stats.
-
 > 本包已从 `pi-deepseek-cache-optimizer` 改名。已有底部统计会自动迁移。本扩展绝不会创建、修改、备份或删除你的 `~/.pi/agent/models.json`。
 
 ## 目录
@@ -100,6 +98,7 @@ LiteLLM / OneAPI / NewAPI / 类 OpenRouter 渠道等第三方 `openai-completion
 - `supportsLongCacheRetention: true` 是可选项。只有 endpoint 明确支持 OpenAI long prompt cache retention 时才添加。
 - 如果出现 `400 Unsupported parameter: prompt_cache_retention`，请为该渠道移除 / 避免 `supportsLongCacheRetention`；如支持，可保留 `sendSessionAffinityHeaders`。
 - 使用 `/cache-optimizer compat` 或 `/cache-optimizer doctor` 查看当前模型的具体建议。
+- 对 DeepSeek 模型，Pi Mono 指南期望在支持时同时设置 `compat.requiresReasoningContentOnAssistantMessages: true` 和 `compat.thinkingFormat: "deepseek"`，再配合缓存 / session-affinity 相关 compat。
 - 本扩展只给建议，不会修改 `models.json`。
 
 ## Footer 统计
@@ -114,7 +113,7 @@ OpenAI cache 3/10 · 0.002M/0.005M tok (40%) ⚠️ compat
 
 格式：`<label> <命中请求数>/<总请求数> · <cached input tokens>/<total input tokens> tok (<token 命中率>)`。部分 adapter 还可能追加 `· write <tokens> tok`，运行时诊断可能追加 `⚠️ compat` 或 `⚠️ integrity`。
 
-支持的 footer label 包括：DS、Claude、OpenAI、Gemini、Kimi、Qwen、GLM、MiniMax、Hunyuan、Mistral、Grok、Llama、Nemotron、Cohere、Yi、Doubao、ERNIE、Baichuan、StepFun、Spark、InternLM、Gemma、Phi、Jamba、Solar、Sonar、Nova、Reka、Falcon、DBRX、MPT、StableLM、Aquila、EXAONE、HyperCLOVA、Luminous、Hermes、Granite、Arctic、Pangu、SenseNova、Zhinao、MiniCPM、XVERSE、Orion、OpenChat、Vicuna、Wizard、Zephyr、Dolphin、OpenOrca、Starling、BLOOM、RWKV、Aya。
+支持的 footer label 包括：DS、Claude、OpenAI、Gemini、Kimi、Qwen、GLM、MiniMax、Mimo、Hunyuan、Mistral、Grok、Llama、Nemotron、Cohere、Yi、Doubao、ERNIE、Baichuan、StepFun、Spark、InternLM、Gemma、Phi、Jamba、Solar、Sonar、Nova、Reka、Falcon、DBRX、MPT、StableLM、Aquila、EXAONE、HyperCLOVA、Luminous、Hermes、Granite、Arctic、Pangu、SenseNova、Zhinao、MiniCPM、XVERSE、Orion、OpenChat、Vicuna、Wizard、Zephyr、Dolphin、OpenOrca、Starling、BLOOM、RWKV、Aya。
 
 Adapter 选择只看模型 id/name（以及 message_end 时 assistant message 的 model/name）。仅使用 OpenAI-shaped API 不会被当作 OpenAI-family，除非模型 id/name 匹配受支持的家族。
 

package/index.ts

@@ -126,6 +126,7 @@ const MIN_STABLE_CANDIDATE_LENGTH = 8;
 const ASSISTANT_MESSAGE_MODEL_TOKEN_KEYS = ["model", "name"];
 const OPENAI_REASONING_MODEL_PATTERN = /(^|[/\s:_-])o[1345]($|[-_.:/\s])/;
 const XAI_MODEL_PATTERN = /(^|[/\s:_-])xai($|[-_.:/\s])/;
+const MIMO_MODEL_PATTERN = /(^|[/\s:_-])mi-?mo($|[-_.:/\s])/i;
 const PPLX_MODEL_PATTERN = /(^|[/\s:_-])pplx($|[-_.:/\s])/i;
 const NOVA_MODEL_PATTERN = /(^|[/\s:_-])nova($|[-_.:/\s])/i;
 const MPT_MODEL_PATTERN = /(^|[/\s:_-])mpt($|[-_.:/\s])/i;
@@ -141,6 +142,7 @@ type CacheCompat = {
   sendSessionIdHeader?: boolean;
   supportsLongCacheRetention?: boolean;
   thinkingFormat?: string;
+  requiresReasoningContentOnAssistantMessages?: boolean;
   cacheControlFormat?: string;
 };
 
@@ -831,6 +833,18 @@ function isMiniMaxLikeAssistantMessage(message: unknown, model: PiModel | undefi
   return modelOrAssistantMessageHas(message, model, ["minimax"]);
 }
 
+function isMimoLikeModel(model: PiModel | undefined): boolean {
+  const tokens = getModelIdNameTokenValues(model);
+  return hasAnyTokenContaining(tokens, ["xiaomimimo"]) || tokens.some((t) => MIMO_MODEL_PATTERN.test(t));
+}
+function isMimoLikeAssistantMessage(message: unknown, model: PiModel | undefined): boolean {
+  const allTokens = [
+    ...getModelIdNameTokenValues(model),
+    ...getAssistantMessageModelTokenValues(message),
+  ];
+  return hasAnyTokenContaining(allTokens, ["xiaomimimo"]) || allTokens.some((t) => MIMO_MODEL_PATTERN.test(t));
+}
+
 function isHunyuanLikeModel(model: PiModel | undefined): boolean {
   return hasAnyTokenContaining(getModelIdNameTokenValues(model), ["hunyuan"]);
 }
@@ -1492,7 +1506,7 @@ function describeMissingOpenAIFamilyProxyCompat(model: PiModel): string[] {
 /**
  * Like describeMissingOpenAIFamilyProxyCompat but without the isOpenAIFamilyModel
  * gate. Warns for ANY model using openai-completions through a non-official base
- * URL — covers GPT, Kimi, Qwen, GLM, MiniMax, Hunyuan, and any other
+ * URL — covers GPT, Kimi, Qwen, GLM, MiniMax, Mimo, Hunyuan, and any other
  * OpenAI-compatible proxy.
  */
 function describeMissingOpenAICompatibleProxyCompat(model: PiModel): string[] {
@@ -1590,10 +1604,88 @@ function describeMissingDeepSeekCompat(model: PiModel): string[] {
   } else if (compat.sendSessionAffinityHeaders !== true) {
     missing.push("sendSessionAffinityHeaders");
   }
+  if (compat.requiresReasoningContentOnAssistantMessages !== true) {
+    missing.push("requiresReasoningContentOnAssistantMessages");
+  }
+  if (compat.thinkingFormat !== "deepseek") {
+    missing.push("thinkingFormat");
+  }
 
   return missing;
 }
 
+function isDeepSeekCompatCheckApplicable(model: PiModel): boolean {
+  return isDeepSeekLikeModel(model) && isOpenAICompatibleApi(model.api);
+}
+
+function describeMissingCacheCompatForModel(model: PiModel): string[] {
+  if (isDeepSeekCompatCheckApplicable(model)) {
+    return describeMissingDeepSeekCompat(model);
+  }
+  return describeMissingOpenAICompatibleProxyCompat(model);
+}
+
+function buildDeepSeekCompatSuggestion(missing: string[]): Record<string, unknown> {
+  const suggestion: Record<string, unknown> = {};
+
+  if (missing.includes("supportsLongCacheRetention")) {
+    suggestion.supportsLongCacheRetention = true;
+  }
+  if (missing.includes("sendSessionIdHeader")) {
+    suggestion.sendSessionIdHeader = true;
+  }
+  if (missing.includes("sendSessionAffinityHeaders")) {
+    suggestion.sendSessionAffinityHeaders = true;
+  }
+  if (missing.includes("requiresReasoningContentOnAssistantMessages")) {
+    suggestion.requiresReasoningContentOnAssistantMessages = true;
+  }
+  if (missing.includes("thinkingFormat")) {
+    suggestion.thinkingFormat = "deepseek";
+  }
+
+  return suggestion;
+}
+
+function appendDeepSeekCompatAdviceLines(lines: string[], missing: string[]): void {
+  const suggestion = buildDeepSeekCompatSuggestion(missing);
+  if (Object.keys(suggestion).length > 0) {
+    lines.push("Recommended DeepSeek compat snippet:");
+    lines.push(JSON.stringify(suggestion, null, 2));
+  }
+
+  if (missing.includes("requiresReasoningContentOnAssistantMessages")) {
+    lines.push('- requiresReasoningContentOnAssistantMessages: true keeps replayed assistant turns compatible with DeepSeek reasoning_content requirements.');
+  }
+  if (missing.includes("thinkingFormat")) {
+    lines.push('- thinkingFormat: "deepseek" tells Pi to use DeepSeek reasoning/thinking parameter format.');
+  }
+  if (missing.includes("sendSessionAffinityHeaders")) {
+    lines.push("- sendSessionAffinityHeaders: recommended for OpenAI-compatible DeepSeek proxies when supported; it helps keep one Pi session on the same upstream/backend.");
+  }
+  if (missing.includes("sendSessionIdHeader")) {
+    lines.push("- sendSessionIdHeader: recommended for OpenAI Responses-compatible DeepSeek proxies when supported.");
+  }
+  if (missing.includes("supportsLongCacheRetention")) {
+    lines.push("- supportsLongCacheRetention: enable for DeepSeek-compatible endpoints that support long cache retention.");
+  }
+}
+
+function buildDeepSeekCompatWarningText(key: string, missing: string[]): string {
+  const slashIdx = key.indexOf("/");
+  const providerLabel = slashIdx > 0 ? key.slice(0, slashIdx) : key;
+  const modelsJsonPath = getModelsJsonDisplayPath();
+  const lines: string[] = [
+    `💡 pi-cache-optimizer: ${key} is DeepSeek-like but merged compat lacks ${missing.join(" and ")}.`,
+    `Proxies may reduce or hide cache hits. Edit ${modelsJsonPath} -> providers["${providerLabel}"] -> compat (at the same level as baseUrl/api/apiKey/models).`,
+    "",
+  ];
+
+  appendDeepSeekCompatAdviceLines(lines, missing);
+
+  return lines.join("\n");
+}
+
 const CACHE_PROVIDER_ADAPTERS: CacheProviderAdapter[] = [
   {
     id: "deepseek",
@@ -1613,13 +1705,7 @@ const CACHE_PROVIDER_ADAPTERS: CacheProviderAdapter[] = [
       if (missing.length === 0) return undefined;
 
       const key = modelKey(model);
-      const slashIdx = key.indexOf("/");
-      const providerLabel = slashIdx > 0 ? key.slice(0, slashIdx) : key;
-      const modelsJsonPath = getModelsJsonDisplayPath();
-      return (
-        `💡 pi-cache-optimizer: ${key} is DeepSeek-like but merged compat lacks ${missing.join(" and ")}. ` +
-        `Proxies may reduce or hide cache hits. Edit ${modelsJsonPath} -> providers["${providerLabel}"] -> compat (at the same level as baseUrl/api/apiKey/models).`
-      );
+      return buildDeepSeekCompatWarningText(key, missing);
     },
   },
   {
@@ -1742,6 +1828,23 @@ const CACHE_PROVIDER_ADAPTERS: CacheProviderAdapter[] = [
       return buildOpenAIProxyCompatWarningText(modelKey(model), missing);
     },
   },
+  {
+    id: "openai" as CacheProviderId,
+    label: "Mimo cache",
+    matchesModel: isMimoLikeModel,
+    matchesAssistantMessage(message, model) {
+      if (!isAssistantMessage(message)) return false;
+      return isMimoLikeAssistantMessage(message, model);
+    },
+    normalizeUsage(message) {
+      return normalizeWithFallback(message, getOpenAIRawUsage);
+    },
+    warningText(model) {
+      const missing = describeMissingOpenAICompatibleProxyCompat(model);
+      if (missing.length === 0) return undefined;
+      return buildOpenAIProxyCompatWarningText(modelKey(model), missing);
+    },
+  },
   {
     id: "openai" as CacheProviderId,
     label: "Hunyuan cache",
@@ -3028,6 +3131,12 @@ function isCompatCheckApplicable(model: PiModel): boolean {
   return lower(model.api) === "openai-completions" && !isOfficialOpenAIBaseUrl(model);
 }
 
+function isPromptCacheRetention400Applicable(model: PiModel): boolean {
+  return isOpenAICompatibleApi(model.api) &&
+    !isOfficialOpenAIBaseUrl(model) &&
+    getCompat(model).supportsLongCacheRetention === true;
+}
+
 /**
  * Detect router / channel profiles from a PiModel and return diagnostic notes.
  *
@@ -3171,7 +3280,7 @@ function describeRouterChannelDiagnostics(model: PiModel): string[] {
 
   // ── 4. Generic third-party OpenAI-compatible proxy ─────────────────
   if (api === "openai-completions" && baseUrl) {
-    const missing = describeMissingOpenAICompatibleProxyCompat(model);
+    const missing = describeMissingCacheCompatForModel(model);
     notes.push(
       "🔀 Router/channel: Third-party OpenAI-compatible proxy. If cache hit rates are low:",
     );
@@ -3207,7 +3316,8 @@ function buildDoctorDiagnosis(model: PiModel, options: { promptCacheRetention400
   const compat = getCompat(model);
   lines.push(`Compat:   ${JSON.stringify(compat)}`);
 
-  const missing = describeMissingOpenAICompatibleProxyCompat(model);
+  const deepSeekCompatApplicable = isDeepSeekCompatCheckApplicable(model);
+  const missing = describeMissingCacheCompatForModel(model);
   if (missing.length > 0) {
     lines.push(`⚠️  Missing compat flags: ${missing.join(", ")}`);
     const key = modelKey(model);
@@ -3215,14 +3325,18 @@ function buildDoctorDiagnosis(model: PiModel, options: { promptCacheRetention400
     const providerLabel = slashIdx > 0 ? key.slice(0, slashIdx) : key;
     const modelsJsonPath = getModelsJsonDisplayPath();
     lines.push(`Edit ${modelsJsonPath} -> providers["${providerLabel}"] -> compat (same level as baseUrl/api/apiKey/models).`);
-    appendOpenAIProxyCompatAdviceLines(lines, missing);
-  } else if (isCompatCheckApplicable(model)) {
+    if (deepSeekCompatApplicable) {
+      appendDeepSeekCompatAdviceLines(lines, missing);
+    } else {
+      appendOpenAIProxyCompatAdviceLines(lines, missing);
+    }
+  } else if (deepSeekCompatApplicable || isCompatCheckApplicable(model)) {
     lines.push("✅ Compat fully configured.");
   } else {
     lines.push("ℹ️ Compat check not applicable for this model.");
   }
 
-  if (isCompatCheckApplicable(model) && compat.supportsLongCacheRetention === true) {
+  if (isPromptCacheRetention400Applicable(model)) {
     lines.push("");
     if (options.promptCacheRetention400) {
       lines.push("⚠️  A 400 response was observed while supportsLongCacheRetention is enabled.");
@@ -3274,8 +3388,8 @@ function buildLowHitDiagnosis(
 ): string[] {
   const lines: string[] = [];
 
-  // 1. Missing compat flags (reuse existing check)
-  const missingCompat = describeMissingOpenAICompatibleProxyCompat(model);
+  // 1. Missing compat flags (adapter-aware: DeepSeek has extra reasoning compat)
+  const missingCompat = describeMissingCacheCompatForModel(model);
 
   // 2. Router/channel risk (reuse existing check)
   const routerNotes = describeRouterChannelDiagnostics(model);
@@ -3371,7 +3485,8 @@ function buildLowHitDiagnosis(
 }
 
 function buildCompatDiagnosis(model: PiModel): string | undefined {
-  const missing = describeMissingOpenAICompatibleProxyCompat(model);
+  const missing = describeMissingCacheCompatForModel(model);
+  const deepSeekCompatApplicable = isDeepSeekCompatCheckApplicable(model);
   const routerNotes = describeRouterChannelDiagnostics(model);
 
   if (missing.length === 0 && routerNotes.length === 0) return undefined;
@@ -3388,14 +3503,18 @@ function buildCompatDiagnosis(model: PiModel): string | undefined {
     lines.push("");
     lines.push(`Edit ${modelsJsonPath} -> providers["${providerLabel}"] -> compat`);
     lines.push(`(at the same level as baseUrl/api/apiKey/models).`);
-    appendOpenAIProxyCompatAdviceLines(lines, missing);
+    if (deepSeekCompatApplicable) {
+      appendDeepSeekCompatAdviceLines(lines, missing);
+    } else {
+      appendOpenAIProxyCompatAdviceLines(lines, missing);
+    }
   }
 
   // When compat is fully configured but router notes exist, prefix the status.
   if (routerNotes.length > 0 && missing.length === 0) {
-    if (isCompatCheckApplicable(model)) {
+    if (deepSeekCompatApplicable || isCompatCheckApplicable(model)) {
       lines.push("✅ Compat fully configured.");
-      if (getCompat(model).supportsLongCacheRetention === true) {
+      if (isPromptCacheRetention400Applicable(model)) {
         lines.push(getPromptCacheRetentionUnsupportedHint());
       }
     } else {
@@ -3441,9 +3560,16 @@ export const __internals_for_tests = {
   isOpenAIFamilyToken,
   describeMissingOpenAIFamilyProxyCompat,
   describeMissingOpenAICompatibleProxyCompat,
+  describeMissingDeepSeekCompat,
+  isDeepSeekCompatCheckApplicable,
+  describeMissingCacheCompatForModel,
+  buildDeepSeekCompatSuggestion,
+  buildDeepSeekCompatWarningText,
   buildSafeOpenAIProxyCompatSuggestion,
   getPromptCacheRetentionUnsupportedHint,
   isOfficialOpenAIBaseUrl,
+  isCompatCheckApplicable,
+  isPromptCacheRetention400Applicable,
   // Non-GPT OpenAI-compatible model detection
   isKimiLikeModel,
   isKimiLikeAssistantMessage,
@@ -3453,6 +3579,8 @@ export const __internals_for_tests = {
   isGLMLikeAssistantMessage,
   isMiniMaxLikeModel,
   isMiniMaxLikeAssistantMessage,
+  isMimoLikeModel,
+  isMimoLikeAssistantMessage,
   isHunyuanLikeModel,
   isHunyuanLikeAssistantMessage,
   // Additional OpenAI-compatible model detection
@@ -3551,6 +3679,8 @@ export const __internals_for_tests = {
   isRwkvLikeAssistantMessage,
   isAyaLikeModel,
   isAyaLikeAssistantMessage,
+  selectAdapterForModel,
+  selectAdapterForAssistantMessage,
   buildOpenAIProxyCompatWarningText,
   getModelIdNameTokenValues,
   getAssistantMessageModelTokenValues,
@@ -3899,15 +4029,15 @@ export default function (pi: ExtensionAPI) {
       }
     }
 
-    // ⚠️ compat footer marker: if the active model is a non-official
-    // openai-completions model with missing supportsLongCacheRetention
-    // or sendSessionAffinityHeaders, append the marker to indicate that
-    // compat configuration is incomplete. Re-evaluated on every status
-    // update so the marker persists through stats changes and day
-    // rollovers. Redundant setStatus calls are blocked by the
+    // ⚠️ compat footer marker: if the active model has adapter-specific
+    // missing compat (DeepSeek reasoning/cache compat, or a non-official
+    // openai-completions model missing cache/session-affinity flags), append
+    // the marker to indicate that compat configuration is incomplete.
+    // Re-evaluated on every status update so the marker persists through stats
+    // changes and day rollovers. Redundant setStatus calls are blocked by the
     // `lastStatusText` early return above.
     if (runtimeOptimizerEnabled && statusText !== undefined && model) {
-      const compatMissing = describeMissingOpenAICompatibleProxyCompat(model);
+      const compatMissing = describeMissingCacheCompatForModel(model);
       if (compatMissing.length > 0) {
         statusText = statusText + " ⚠️ compat";
       }
@@ -4027,8 +4157,7 @@ export default function (pi: ExtensionAPI) {
     const model = ctx.model;
     if (!runtimeOptimizerEnabled || !model) return;
     if (event.status !== 400) return;
-    if (!isCompatCheckApplicable(model)) return;
-    if (getCompat(model).supportsLongCacheRetention !== true) return;
+    if (!isPromptCacheRetention400Applicable(model)) return;
 
     const key = modelKey(model);
     promptCacheRetention400Models.add(key);
@@ -4140,7 +4269,7 @@ export default function (pi: ExtensionAPI) {
           cmdCtx.ui.notify(compatResult, "warning");
         } else {
           cmdCtx.ui.notify(
-            isCompatCheckApplicable(model)
+            isDeepSeekCompatCheckApplicable(model) || isCompatCheckApplicable(model)
               ? "✅ Compat fully configured."
               : "ℹ️ Compat check not applicable for this model.",
             "info",
@@ -4238,7 +4367,7 @@ export default function (pi: ExtensionAPI) {
                 cmdCtx.ui.notify(compatResult, "warning");
               } else {
                 cmdCtx.ui.notify(
-                  isCompatCheckApplicable(model)
+                  isDeepSeekCompatCheckApplicable(model) || isCompatCheckApplicable(model)
                     ? "✅ Compat fully configured."
                     : "ℹ️ Compat check not applicable for this model.",
                   "info",
@@ -4285,11 +4414,11 @@ export default function (pi: ExtensionAPI) {
         diagnosis.push("");
         if (model) {
           const displayKey = modelKey(model);
-          const missing = describeMissingOpenAICompatibleProxyCompat(model);
+          const missing = describeMissingCacheCompatForModel(model);
           if (missing.length > 0) {
             diagnosis.push(`⚠️  Active model "${displayKey}" missing compat: ${missing.join(", ")}`);
             diagnosis.push('Run "/cache-optimizer compat" for edit instructions.');
-          } else if (isCompatCheckApplicable(model)) {
+          } else if (isDeepSeekCompatCheckApplicable(model) || isCompatCheckApplicable(model)) {
             diagnosis.push(`✅ Active model "${displayKey}": compat fully configured.`);
           } else {
             diagnosis.push(`ℹ️ Active model "${displayKey}": compat check not applicable.`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-cache-optimizer",
-  "version": "2.5.3",
+  "version": "2.5.4",
   "description": "Improve Pi prompt/KV cache hit rates with stable prompts, OpenAI-compatible cache keys, proxy compat warnings, and footer cache stats.",
   "keywords": [
     "pi-package",