@khanglvm/llm-router 2.3.0 → 2.3.2

package/src/node/web-console-ui/provider-presets.js

@@ -0,0 +1,211 @@
+import { JSON_HEADERS } from "./constants.js";
+import { CODEX_SUBSCRIPTION_MODELS, CLAUDE_CODE_SUBSCRIPTION_MODELS } from "../../runtime/subscription-constants.js";
+
+/** Factory for OpenAI-compatible model discovery via /models endpoint. */
+function createOpenAICompatDiscover(endpoint, { requiresAuth = true } = {}) {
+  return Object.freeze({
+    requiresAuth,
+    fetchModels: async ({ apiKey, apiKeyEnv } = {}) => {
+      const body = { endpoints: [endpoint] };
+      if (apiKey) body.apiKey = apiKey;
+      if (apiKeyEnv) body.apiKeyEnv = apiKeyEnv;
+      const res = await fetchJson("/api/config/discover-provider-models", {
+        method: "POST",
+        headers: JSON_HEADERS,
+        body: JSON.stringify(body)
+      });
+      return (res.result?.models || []).map((id) => String(id || "").trim()).filter(Boolean);
+    }
+  });
+}
+
+/** Unified provider preset registry — single source of truth for all presets. */
+const PROVIDER_PRESETS = Object.freeze([
+  // ── API presets ──
+  Object.freeze({
+    key: "custom",
+    category: "api",
+    label: "Custom",
+    description: "Generic OpenAI-compatible API provider.",
+    providerName: "My Provider",
+    providerId: "my-provider",
+    endpoint: "",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze({ openai: Object.freeze(["gpt-4o-mini", "gpt-4.1-mini"]), claude: Object.freeze(["claude-3-5-sonnet", "claude-3-5-haiku"]) }),
+    rateLimitDefaults: Object.freeze({ limit: 60, windowValue: 1, windowUnit: "minute" })
+  }),
+  Object.freeze({
+    key: "groq",
+    category: "api",
+    label: "Groq",
+    description: "Groq cloud inference with Llama, Qwen, and GPT-OSS models.",
+    providerName: "Groq",
+    providerId: "groq",
+    endpoint: "https://api.groq.com/openai/v1",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze(["llama-3.3-70b-versatile", "llama-3.1-8b-instant"]),
+    rateLimitDefaults: Object.freeze({ limit: 30, windowValue: 1, windowUnit: "minute" }),
+    freeTierRpm: Object.freeze({
+      host: "api.groq.com",
+      models: Object.freeze({
+        "llama-3.1-8b-instant": 30,
+        "llama-3.3-70b-versatile": 30,
+        "openai/gpt-oss-20b": 30,
+        "openai/gpt-oss-120b": 15,
+        "qwen/qwen3-32b": 30,
+        "meta-llama/llama-4-scout-17b-16e-instruct": 15,
+        "moonshotai/kimi-k2-instruct": 15,
+        "_default": 30
+      })
+    }),
+    discover: createOpenAICompatDiscover("https://api.groq.com/openai/v1")
+  }),
+  Object.freeze({
+    key: "gemini",
+    category: "api",
+    label: "Google Gemini",
+    description: "Google Gemini models via OpenAI-compatible endpoint.",
+    providerName: "Google Gemini",
+    providerId: "gemini",
+    endpoint: "https://generativelanguage.googleapis.com/v1beta/openai",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze(["gemini-3-flash-preview", "gemini-3.1-flash-lite-preview"]),
+    rateLimitDefaults: Object.freeze({ limit: 10, windowValue: 1, windowUnit: "minute" }),
+    freeTierRpm: Object.freeze({
+      host: "generativelanguage.googleapis.com",
+      models: Object.freeze({
+        "gemini-3-flash-preview": 15,
+        "gemini-3.1-flash-lite-preview": 15,
+        "gemini-3.1-pro-preview": 5,
+        "gemini-2.5-flash": 15,
+        "gemini-2.5-flash-lite": 15,
+        "gemini-2.5-pro": 5,
+        "_default": 10
+      })
+    }),
+    discover: createOpenAICompatDiscover("https://generativelanguage.googleapis.com/v1beta/openai")
+  }),
+  Object.freeze({
+    key: "zai-global",
+    category: "api",
+    label: "Z.AI Coding (Global)",
+    description: "Z.AI coding models via the global coding endpoint.",
+    providerName: "Z.AI Coding",
+    providerId: "zai-coding",
+    endpoint: "https://api.z.ai/api/coding/paas/v4",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze(["glm-5.1", "glm-5V-turbo", "glm-4.7", "glm-4.7-flash"]),
+    rateLimitDefaults: Object.freeze({ limit: 60, windowValue: 1, windowUnit: "minute" }),
+    discover: createOpenAICompatDiscover("https://api.z.ai/api/coding/paas/v4")
+  }),
+  Object.freeze({
+    key: "zai-china",
+    category: "api",
+    label: "Z.AI Coding (China)",
+    description: "Z.AI coding models via the China mainland coding endpoint.",
+    providerName: "Z.AI Coding CN",
+    providerId: "zai-coding-cn",
+    endpoint: "https://open.bigmodel.cn/api/coding/paas/v4",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze(["glm-5.1", "glm-5V-turbo", "glm-4.7", "glm-4.7-flash"]),
+    rateLimitDefaults: Object.freeze({ limit: 60, windowValue: 1, windowUnit: "minute" }),
+    discover: createOpenAICompatDiscover("https://open.bigmodel.cn/api/coding/paas/v4")
+  }),
+  Object.freeze({
+    key: "openrouter",
+    category: "api",
+    label: "OpenRouter",
+    description: "300+ models from multiple providers, including free tier models.",
+    providerName: "OpenRouter",
+    providerId: "openrouter",
+    endpoint: "https://openrouter.ai/api/v1",
+    apiKeyEnv: "",
+    defaultModels: Object.freeze(["qwen/qwen3.6-plus:free", "google/gemma-4-26b-a4b-it"]),
+    rateLimitDefaults: Object.freeze({ limit: 200, windowValue: 1, windowUnit: "minute" }),
+    discover: createOpenAICompatDiscover("https://openrouter.ai/api/v1", { requiresAuth: false })
+  }),
+
+  // ── Subscription (OAuth) presets ──
+  Object.freeze({
+    key: "oauth-gpt",
+    category: "subscription",
+    label: "ChatGPT",
+    description: "Use ChatGPT subscription login with GPT models.",
+    providerName: "ChatGPT Subscription",
+    providerId: "chatgpt-sub",
+    subscriptionType: "chatgpt-codex",
+    format: "openai",
+    defaultModels: CODEX_SUBSCRIPTION_MODELS,
+    rateLimitDefaults: Object.freeze({ limit: 999999, windowValue: 1, windowUnit: "month" }),
+    warning: "chatgpt-tos"
+  }),
+  Object.freeze({
+    key: "oauth-claude",
+    category: "subscription",
+    label: "Claude",
+    description: "Use Claude Code subscription login with Claude models.",
+    providerName: "Claude Subscription",
+    providerId: "claude-sub",
+    subscriptionType: "claude-code",
+    format: "claude",
+    defaultModels: CLAUDE_CODE_SUBSCRIPTION_MODELS,
+    rateLimitDefaults: Object.freeze({ limit: 999999, windowValue: 1, windowUnit: "month" }),
+    warning: "claude-extra-usage"
+  })
+]);
+
+/** Index helpers for PROVIDER_PRESETS registry. */
+const PROVIDER_PRESET_BY_KEY = Object.freeze(Object.fromEntries(PROVIDER_PRESETS.map((p) => [p.key, p])));
+
+function findPresetByKey(key) {
+  return PROVIDER_PRESET_BY_KEY[key] || PROVIDER_PRESET_BY_KEY.custom;
+}
+
+function findPresetByHost(hostname) {
+  return PROVIDER_PRESETS.find((p) => p.freeTierRpm?.host === hostname) || null;
+}
+
+function getPresetOptionsByCategory(category) {
+  return PROVIDER_PRESETS.filter((p) => p.category === category);
+}
+
+/** Build the free-tier RPM lookup map keyed by hostname (used by detectPresetHostFromEndpoints). */
+const PROVIDER_PRESET_FREE_TIER_RPM_BY_HOST = Object.freeze(
+  Object.fromEntries(
+    PROVIDER_PRESETS
+      .filter((p) => p.freeTierRpm?.host && p.freeTierRpm?.models)
+      .map((p) => [p.freeTierRpm.host, p.freeTierRpm.models])
+  )
+);
+
+/** Module-level cache for preset model discovery — survives React re-renders. */
+const presetModelCache = new Map();
+let _presetInitPromise = null;
+
+/** Non-blocking background init: fetches models for presets that don't require auth. */
+function initPresetModels() {
+  if (_presetInitPromise) return _presetInitPromise;
+  _presetInitPromise = Promise.allSettled(
+    PROVIDER_PRESETS
+      .filter((p) => p.discover && !p.discover.requiresAuth)
+      .map(async (preset) => {
+        try {
+          const models = await preset.discover.fetchModels();
+          if (models.length) presetModelCache.set(preset.key, models);
+        } catch { /* background — swallow errors */ }
+      })
+  );
+  return _presetInitPromise;
+}
+
+export {
+  createOpenAICompatDiscover,
+  PROVIDER_PRESETS,
+  PROVIDER_PRESET_BY_KEY,
+  findPresetByKey,
+  findPresetByHost,
+  getPresetOptionsByCategory,
+  PROVIDER_PRESET_FREE_TIER_RPM_BY_HOST,
+  presetModelCache,
+  initPresetModels
+};