dsh-llm-workbuddy 0.1.9 → 0.1.11

package/README.md

@@ -280,8 +280,10 @@ DSH 的 `dsh.client` 机制只要求 `package.json` 里：
 
 模型列表默认取插件内置目录；代理可达时改为实时拉取 `/v1/models`（支持
 `{"models": [...]}` / `{"data": [...]}` 两种返回），未列出的模型 id 仍可原样
-传递。每个模型的推理等级（`reasoningEffort`）默认来自内置目录，代理可达时优先
-沿用目录值（代理 `/v1/models` 当前只上报 `High`，不区分模型）。
+传递。**内置目录充当白名单**：代理返回的已下架/不可用模型（如 glm-4.6v、
+glm-5.0、minimax-m2.5 等，上游返回 `service info not found`）会被过滤，不显示
+在 UI 中。每个模型的推理等级（`reasoningEffort`）默认来自内置目录，代理可达时
+优先沿用目录值（代理 `/v1/models` 当前只上报 `High`，不区分模型）。
 
 ---
 

package/lib/index.js

@@ -75,19 +75,15 @@ const DISCOVERY_TIMEOUT_MS = 2_000;
 const DEFAULT_MODELS = [
   { id: "deepseek-v4-pro", name: "Deepseek-V4-Pro", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
   { id: "deepseek-v4-flash", name: "Deepseek-V4-Flash", contextWindow: 1_000_000, maxTokens: 50_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
+  { id: "deepseek-v3-2-volc", name: "DeepSeek-V3.2", contextWindow: 96_000, maxTokens: 32_000, reasoningEffort: "medium" },
   { id: "glm-5.2", name: "GLM-5.2", contextWindow: 1_000_000, maxTokens: 48_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
   { id: "glm-5.1", name: "GLM-5.1", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "medium" },
-  { id: "glm-5.0", name: "GLM-5.0", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "high" },
-  { id: "glm-4.7", name: "GLM-4.7", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "high" },
-  { id: "glm-4.6", name: "GLM-4.6", contextWindow: 168_000, maxTokens: 32_000, reasoningEffort: "high" },
   { id: "glm-5v-turbo", name: "GLM-5v-Turbo", contextWindow: 200_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
   { id: "kimi-k3-1", name: "Kimi-K3", contextWindow: 1_000_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
   { id: "kimi-k2.7", name: "Kimi-K2.7-Code", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
   { id: "kimi-k2.6", name: "Kimi-K2.6", contextWindow: 256_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
   { id: "kimi-k2.5", name: "Kimi-K2.5", contextWindow: 164_000, maxTokens: 32_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "kimi-k2-thinking", name: "Kimi-K2-Thinking", contextWindow: 164_000, maxTokens: 32_000, reasoningEffort: "medium" },
   { id: "minimax-m3", name: "MiniMax-M3", contextWindow: 512_000, maxTokens: 128_000, inputModalities: ["text", "image"], reasoningEffort: "medium" },
-  { id: "minimax-m2.5", name: "MiniMax-M2.5", contextWindow: 200_000, maxTokens: 48_000, reasoningEffort: "medium" },
   { id: "hy3", name: "Hy3", contextWindow: 192_000, maxTokens: 64_000, inputModalities: ["text", "image"], reasoningEffort: "high" },
   { id: "hunyuan-2.0-thinking", name: "Hunyuan-2.0-Thinking", contextWindow: 128_000, maxTokens: 24_000, reasoningEffort: "medium" },
   { id: "hunyuan-chat", name: "Hunyuan-Turbos", contextWindow: 200_000, maxTokens: 8_192 },
@@ -183,7 +179,10 @@ async function serializeMessages(messages, attachments, signal) {
     if (attachments === undefined) {
       throw new LlmError("WorkBuddy image input requires the durable attachment service", "UNSUPPORTED_CONTENT");
     }
-    const policy = { maxPixels: 4_000_000, maxBytes: 20_000_000 };
+    // Keep images small enough for the upstream platform: the proxy relays the
+    // base64 body verbatim, and oversized payloads are dropped/ignored upstream
+    // (models reply "no image attached"). Mirror the official adapter budgets.
+    const policy = { maxPixels: 2048 * 2048, maxBytes: 1024 * 1024 };
     const ordered = [...refs.values()];
     const prepared = await Promise.all(ordered.map((ref) => attachments.readImageRequest(ref, policy, signal)));
     for (let index = 0; index < ordered.length; index += 1) {
@@ -553,6 +552,11 @@ export class WorkBuddyAdapter extends LlmAdapter {
         if (seen.has(entry.id)) continue;
         seen.add(entry.id);
         const catalog = byId.get(entry.id);
+        // Only keep models the static catalog declares. The proxy announces
+        // retired models (glm-4.6v, hunyuan-image-v3.0, ...) that upstream
+        // rejects with `service info not found`; the catalog doubles as the
+        // whitelist so those never reach the UI.
+        if (catalog === undefined) continue;
         const effort = liveReasoningEffort(entry, catalog?.reasoningEffort);
         const reasoning = modelReasoningInfo(effort);
         merged.push({

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-llm-workbuddy",
-  "version": "0.1.9",
+  "version": "0.1.11",
   "description": "WorkBuddy (via the local workbuddy2api proxy) LLM provider adapter for DeepSeek Harness, with a Web login-status widget",
   "type": "module",
   "main": "lib/index.js",