feat(codex): auto-populate OpenCode Go model catalog on fetch

For OpenCode Go providers, 'Fetch Models' now enriches the upstream /v1/models
result into the full Codex catalog: each model gets its per-model context window,
messages-only models (e.g. qwen3.7-max) are excluded, and newly added upstream
models are picked up automatically with a single click.

Adds src/config/opencodeGoModelMeta.ts (base-url detection, per-family context
windows, messages-only set, enrichment) and wires it into the Codex form's fetch
success handler. No backend change: reuses the existing fetch_models_for_config
command. Other providers keep the existing fetched-models dropdown behavior.

Author: cc-switch contributor

src/components/providers/forms/CodexFormFields.tsx

@@ -25,6 +25,10 @@ import {
   showFetchModelsError,
   type FetchedModel,
 } from "@/lib/api/model-fetch";
+import {
+  enrichOpencodeGoModels,
+  isOpencodeGoBaseUrl,
+} from "@/config/opencodeGoModelMeta";
 import { CustomUserAgentField } from "./CustomUserAgentField";
 import { cn } from "@/lib/utils";
 import type {
@@ -245,18 +249,36 @@ export function CodexFormFields({
         setFetchedModels(models);
         if (models.length === 0) {
           toast.info(t("providerForm.fetchModelsEmpty"));
-        } else {
+          return;
+        }
+        // OpenCode Go: directly enrich the fetched models into the full catalog
+        // (with per-model context windows, excluding messages-only models), so a
+        // single "Fetch Models" click auto-syncs any newly added upstream models.
+        if (isOpencodeGoBaseUrl(codexBaseUrl) && onCatalogModelsChange) {
+          const enriched = enrichOpencodeGoModels(models);
+          setCatalogRows(enriched.map((m) => createCatalogRow(m)));
           toast.success(
-            t("providerForm.fetchModelsSuccess", { count: models.length }),
+            t("providerForm.fetchModelsSuccess", { count: enriched.length }),
           );
+          return;
         }
+        toast.success(
+          t("providerForm.fetchModelsSuccess", { count: models.length }),
+        );
       })
       .catch((err) => {
         console.warn("[ModelFetch] Failed:", err);
         showFetchModelsError(err, t);
       })
       .finally(() => setIsFetchingModels(false));
-  }, [codexBaseUrl, codexApiKey, isFullUrl, customUserAgent, t]);
+  }, [
+    codexBaseUrl,
+    codexApiKey,
+    isFullUrl,
+    customUserAgent,
+    onCatalogModelsChange,
+    t,
+  ]);
 
   const handleAddCatalogRow = useCallback(() => {
     if (!onCatalogModelsChange) return;

src/config/opencodeGoModelMeta.ts

@@ -0,0 +1,101 @@
+/**
+ * OpenCode Go 模型元数据辅助：用于在「获取模型」时把上游 /v1/models 返回的模型
+ * 富化为带正确上下文窗口的 Codex 模型目录条目，并排除仅 /messages 通路的模型。
+ *
+ * 这样当 OpenCode Go 上新增模型时，用户点一次「获取模型」即可自动拉取并套用上下文窗口，
+ * 无需手动逐个填写。
+ */
+import type { CodexCatalogModel } from "../types";
+
+/** OpenCode Go 的标准 base_url。 */
+export const OPENCODE_GO_BASE_URL = "https://opencode.ai/zen/go/v1";
+
+/** 判断某个 base_url 是否指向 OpenCode Go（容忍有无 /v1 与大小写）。 */
+export function isOpencodeGoBaseUrl(baseUrl?: string | null): boolean {
+  return !!baseUrl && baseUrl.toLowerCase().includes("opencode.ai/zen/go");
+}
+
+/**
+ * 仅在 Anthropic /messages 通路可用、无法走 Codex 的 openai_chat 通路的模型。
+ * 这些模型会从 chat 目录中排除（参见 docs：Go 的部分模型仅 /messages）。
+ */
+const MESSAGES_ONLY_MODELS = new Set<string>(["qwen3.7-max"]);
+
+/**
+ * 上下文窗口按模型 id 的有序前缀规则匹配；新加入的同系模型会自动继承同族窗口。
+ * 数据依据 models.dev 的 opencode-go provider 与实测。
+ */
+const CONTEXT_RULES: Array<[RegExp, number]> = [
+  [/^deepseek-v4/, 1_000_000],
+  [/^glm-5\.2/, 1_000_000],
+  [/^glm-5(\.1)?$/, 200_000],
+  [/^glm-/, 200_000],
+  [/^kimi-/, 262_144],
+  [/^mimo-v2\.5-pro/, 1_048_576],
+  [/^mimo-v2-pro/, 1_048_576],
+  [/^mimo-v2\.5/, 1_000_000],
+  [/^mimo-/, 262_144],
+  [/^qwen3\.[67]-plus/, 1_000_000],
+  [/^qwen3\.5-plus/, 262_144],
+  [/^qwen/, 262_144],
+  [/^minimax-m3/, 512_000],
+  [/^minimax-/, 204_800],
+];
+
+/** 未知模型的兜底上下文窗口（保守值）。 */
+const DEFAULT_CONTEXT_WINDOW = 200_000;
+
+function contextWindowFor(id: string): number {
+  for (const [pattern, ctx] of CONTEXT_RULES) {
+    if (pattern.test(id)) return ctx;
+  }
+  return DEFAULT_CONTEXT_WINDOW;
+}
+
+/** 已知模型的友好显示名；未知模型回退到「按 - 分词 + 首字母大写」。 */
+const DISPLAY_NAME_OVERRIDES: Record<string, string> = {
+  "deepseek-v4-flash": "DeepSeek V4 Flash",
+  "deepseek-v4-pro": "DeepSeek V4 Pro",
+  "glm-5.2": "GLM-5.2",
+  "glm-5.1": "GLM-5.1",
+  "glm-5": "GLM-5",
+  "kimi-k2.7-code": "Kimi K2.7 Code",
+  "kimi-k2.6": "Kimi K2.6",
+  "kimi-k2.5": "Kimi K2.5",
+  "mimo-v2.5-pro": "MiMo V2.5 Pro",
+  "mimo-v2.5": "MiMo V2.5",
+  "mimo-v2-pro": "MiMo V2 Pro",
+  "mimo-v2-omni": "MiMo V2 Omni",
+  "qwen3.7-plus": "Qwen3.7 Plus",
+  "qwen3.6-plus": "Qwen3.6 Plus",
+  "qwen3.5-plus": "Qwen3.5 Plus",
+  "minimax-m3": "MiniMax M3",
+  "minimax-m2.7": "MiniMax M2.7",
+  "minimax-m2.5": "MiniMax M2.5",
+};
+
+function displayNameFor(id: string): string {
+  const override = DISPLAY_NAME_OVERRIDES[id];
+  if (override) return override;
+  return id
+    .split("-")
+    .map((part) => (part ? part[0].toUpperCase() + part.slice(1) : part))
+    .join(" ");
+}
+
+/**
+ * 把 /v1/models 返回的模型（结构上只需 `{ id }`）富化为 Codex 目录条目，
+ * 排除仅 /messages 的模型，并按族套用上下文窗口。
+ */
+export function enrichOpencodeGoModels(
+  models: ReadonlyArray<{ id: string }>,
+): CodexCatalogModel[] {
+  return models
+    .map((m) => m.id)
+    .filter((id) => !!id && !MESSAGES_ONLY_MODELS.has(id))
+    .map((id) => ({
+      model: id,
+      displayName: displayNameFor(id),
+      contextWindow: contextWindowFor(id),
+    }));
+}