codna 0.1.0__py3-none-any.whl

codna/agent-core/vendor/cline/sdk/packages/llms/src/providers/builtins.ts

@@ -0,0 +1,1084 @@
+import {
+	type GatewayModelCapability,
+	type GatewayModelDefinition,
+	type GatewayProviderManifest,
+	type GatewayProviderMetadata,
+	type GatewayProviderSettings,
+	getClineEnvironmentConfig,
+	type JsonValue,
+	type ProviderCapability,
+	type ProviderConfigField,
+} from "@cline/shared";
+import { getGeneratedModelsForProvider } from "../catalog/catalog.generated-access";
+import type {
+	ModelCollection,
+	ModelInfo,
+	ProviderClient,
+	ProviderProtocol,
+} from "../catalog/types";
+import { filterOpenAICodexModels } from "./openai-codex-models";
+import {
+	ANTHROPIC_AND_QWEN_CACHE_ROUTING_METADATA,
+	ANTHROPIC_ROUTING_METADATA,
+	QWEN_CACHE_ROUTING_METADATA,
+} from "./routing/anthropic-compatible";
+import { GLM_THINKING_ROUTING_METADATA } from "./routing/glm-thinking";
+import { MINIMAX_THINKING_ROUTING_METADATA } from "./routing/minimax-thinking";
+
+export const DEFAULT_INTERNAL_OCA_BASE_URL =
+	"https://code-internal.aiservice.us-chicago-1.oci.oraclecloud.com/20250206/app/litellm";
+export const DEFAULT_EXTERNAL_OCA_BASE_URL =
+	"https://code.aiservice.us-chicago-1.oci.oraclecloud.com/20250206/app/litellm";
+const OPENAI_CODEX_DEFAULT_MODEL_ID = "gpt-5.4";
+
+export type ProviderFamily =
+	| "openai"
+	| "openai-compatible"
+	| "anthropic"
+	| "google"
+	| "vertex"
+	| "bedrock"
+	| "mistral"
+	| "claude-code"
+	| "openai-codex"
+	| "opencode"
+	| "dify"
+	| "sap-ai-core";
+
+export interface BuiltinSpec {
+	id: string;
+	name: string;
+	description: string;
+	family: ProviderFamily;
+	protocol?: ProviderProtocol;
+	client?: ProviderClient;
+	capabilities?: ProviderCapability[];
+	popular?: number;
+	modelsProviderId?: string;
+	defaultModelId?: string;
+	modelsFactory?: () => Record<string, ModelInfo>;
+	env?: readonly ("browser" | "node")[];
+	apiKeyEnv?: readonly string[];
+	modelsSourceUrl?: string;
+	docsUrl?: string;
+	defaults?: GatewayProviderSettings;
+	configFields?: readonly ProviderConfigField[];
+	metadata?: GatewayProviderMetadata;
+}
+
+const API_KEY_FIELD: ProviderConfigField = {
+	path: "apiKey",
+	label: "API Key",
+	type: "password",
+	placeholder: "Enter API key...",
+	description: "API key issued by the provider.",
+	secret: true,
+};
+
+const BASE_URL_FIELD: ProviderConfigField = {
+	path: "baseUrl",
+	label: "Base URL",
+	type: "url",
+	placeholder: "https://...",
+	description: "Base endpoint used for provider requests.",
+};
+
+const VERTEX_CONFIG_FIELDS: readonly ProviderConfigField[] = [
+	{
+		path: "gcp.projectId",
+		label: "Google Cloud Project ID",
+		type: "text",
+		placeholder: "my-gcp-project",
+		description: "Google Cloud project that owns the Vertex AI resources.",
+		required: true,
+	},
+	{
+		path: "gcp.region",
+		label: "Vertex Region",
+		type: "text",
+		placeholder: "us-central1",
+		description: "Vertex AI location to run models in.",
+		defaultValue: "us-central1",
+	},
+	{
+		...API_KEY_FIELD,
+		label: "API Key",
+		description:
+			"Optional Google API key for Gemini models. Vertex Anthropic models use Google Cloud credentials.",
+	},
+];
+
+const BEDROCK_CONFIG_FIELDS: readonly ProviderConfigField[] = [
+	{
+		path: "aws.authentication",
+		label: "Authentication",
+		type: "select",
+		description: "Credential source for Amazon Bedrock requests.",
+		options: [
+			{ label: "AWS SDK / IAM", value: "iam" },
+			{ label: "AWS Profile", value: "profile" },
+			{ label: "API Key", value: "api-key" },
+		],
+		defaultValue: "iam",
+	},
+	{
+		path: "aws.region",
+		label: "AWS Region",
+		type: "text",
+		placeholder: "us-east-1",
+		description: "AWS region for Bedrock runtime requests.",
+	},
+	{
+		path: "aws.profile",
+		label: "AWS Profile",
+		type: "text",
+		placeholder: "default",
+		description: "Named AWS profile when using profile authentication.",
+	},
+	{
+		path: "aws.accessKey",
+		label: "Access Key ID",
+		type: "password",
+		placeholder: "AKIA...",
+		secret: true,
+	},
+	{
+		path: "aws.secretKey",
+		label: "Secret Access Key",
+		type: "password",
+		secret: true,
+	},
+	{
+		path: "aws.sessionToken",
+		label: "Session Token",
+		type: "password",
+		secret: true,
+	},
+	{
+		path: "apiKey",
+		label: "Bedrock API Key",
+		type: "password",
+		description: "Optional Bedrock bearer token for API key authentication.",
+		secret: true,
+	},
+	{
+		path: "aws.endpoint",
+		label: "Endpoint URL",
+		type: "url",
+		placeholder: "https://bedrock-runtime.us-east-1.amazonaws.com",
+		description: "Optional custom Bedrock runtime endpoint.",
+	},
+	{
+		path: "aws.useCrossRegionInference",
+		label: "Cross-Region Inference",
+		type: "boolean",
+	},
+	{
+		path: "aws.useGlobalInference",
+		label: "Global Inference",
+		type: "boolean",
+	},
+	{
+		path: "aws.usePromptCache",
+		label: "Prompt Cache",
+		type: "boolean",
+	},
+];
+
+const OCA_CONFIG_FIELDS: readonly ProviderConfigField[] = [
+	{
+		path: "oca.mode",
+		label: "OCA Mode",
+		type: "select",
+		options: [
+			{ label: "External", value: "external" },
+			{ label: "Internal", value: "internal" },
+		],
+		defaultValue: "external",
+	},
+	API_KEY_FIELD,
+	{
+		path: "oca.usePromptCache",
+		label: "Prompt Cache",
+		type: "boolean",
+	},
+];
+
+const QWEN_CONFIG_FIELDS: readonly ProviderConfigField[] = [
+	API_KEY_FIELD,
+	BASE_URL_FIELD,
+	{
+		path: "apiLine",
+		label: "API Line",
+		type: "select",
+		description: "Regional API line for Qwen routing.",
+		options: [
+			{ label: "International", value: "international" },
+			{ label: "China", value: "china" },
+		],
+	},
+];
+
+function defaultConfigFieldsForSpec(
+	spec: BuiltinSpec,
+): readonly ProviderConfigField[] {
+	const fields: ProviderConfigField[] = [];
+	if (spec.apiKeyEnv?.length) {
+		fields.push(API_KEY_FIELD);
+	}
+	if (spec.defaults?.baseUrl?.trim()) {
+		fields.push(BASE_URL_FIELD);
+	}
+	return fields;
+}
+
+function cloneModels(
+	models: Record<string, ModelInfo>,
+): Record<string, ModelInfo> {
+	return Object.fromEntries(
+		Object.entries(models).map(([id, info]) => [id, { ...info }]),
+	);
+}
+
+function uniqueCapabilities(
+	capabilities: readonly ProviderCapability[] | undefined,
+): ProviderCapability[] | undefined {
+	if (!capabilities?.length) return undefined;
+	return [...new Set(capabilities)];
+}
+
+function getProviderCapabilities(
+	spec: BuiltinSpec,
+): ProviderCapability[] | undefined {
+	return uniqueCapabilities([
+		...(spec.capabilities ?? []),
+		...(spec.popular !== undefined ? (["popular"] as const) : []),
+	]);
+}
+
+function getProviderMetadata(
+	spec: BuiltinSpec,
+): GatewayProviderMetadata | undefined {
+	const configFields = spec.configFields ?? defaultConfigFieldsForSpec(spec);
+	const metadata: GatewayProviderMetadata = {
+		...spec.metadata,
+		configFields,
+	};
+	if (spec.popular !== undefined) {
+		metadata.popularRank = spec.popular;
+	}
+	return metadata;
+}
+
+function generatedModels(providerId: string): Record<string, ModelInfo> {
+	return cloneModels(getGeneratedModelsForProvider(providerId));
+}
+
+function pickAnthropicModel(match: (id: string) => boolean): ModelInfo {
+	const entry = Object.entries(generatedModels("anthropic")).find(([id]) =>
+		match(id),
+	);
+	if (entry) {
+		return entry[1];
+	}
+	return {
+		id: "sonnet",
+		name: "Claude Sonnet",
+		capabilities: ["streaming", "reasoning"],
+	};
+}
+
+function buildClaudeCodeModels(): Record<string, ModelInfo> {
+	function toClaudeCodeModel(id: "opus" | "sonnet" | "haiku"): ModelInfo {
+		const source =
+			id === "opus"
+				? pickAnthropicModel((modelId) => modelId.includes("opus"))
+				: id === "haiku"
+					? pickAnthropicModel((modelId) => modelId.includes("haiku"))
+					: pickAnthropicModel((modelId) => modelId.includes("sonnet"));
+		return {
+			...source,
+			id,
+			name: `Claude ${id.charAt(0).toUpperCase()}${id.slice(1)}`,
+		};
+	}
+
+	return {
+		opus: toClaudeCodeModel("opus"),
+		sonnet: toClaudeCodeModel("sonnet"),
+		haiku: toClaudeCodeModel("haiku"),
+	};
+}
+
+function buildOpenAICodexModels(): Record<string, ModelInfo> {
+	return filterOpenAICodexModels(generatedModels("openai-native"));
+}
+
+function fallbackModelInfo(id: string, spec?: BuiltinSpec): ModelInfo {
+	const info: ModelInfo = {
+		id,
+		name: id,
+	};
+	if (spec?.family === "openai-compatible") {
+		info.contextWindow = 128_000;
+		info.maxInputTokens = 128_000;
+		info.capabilities = ["streaming", "tools", "images"];
+	}
+	if (spec?.id === "qwen" || spec?.id === "qwen-code") {
+		info.family = "qwen";
+		info.capabilities = ["prompt-cache"];
+	}
+	return info;
+}
+
+function modelInfoToGateway(
+	providerId: string,
+	info: ModelInfo,
+): GatewayModelDefinition {
+	const capabilities = new Set<GatewayModelCapability>(["text"]);
+	for (const cap of info.capabilities ?? []) {
+		switch (cap) {
+			case "tools":
+				capabilities.add("tools");
+				break;
+			case "reasoning":
+				capabilities.add("reasoning");
+				break;
+			case "prompt-cache":
+				capabilities.add("prompt-cache");
+				break;
+			case "images":
+				capabilities.add("images");
+				break;
+			case "structured_output":
+				capabilities.add("structured-output");
+				break;
+		}
+	}
+	const metadata: Record<string, JsonValue | undefined> = {};
+	if (info.family) {
+		metadata.family = info.family;
+	}
+	if (info.pricing) {
+		metadata.pricing = info.pricing;
+	}
+	if (info.status) {
+		metadata.status = info.status;
+	}
+	if (info.releaseDate) {
+		metadata.releaseDate = info.releaseDate;
+	}
+	if (typeof info.metadata?.reasoningDefaultOn === "boolean") {
+		metadata.reasoningDefaultOn = info.metadata.reasoningDefaultOn;
+	}
+	return {
+		id: info.id,
+		name: info.name ?? info.id,
+		providerId,
+		description: info.description,
+		contextWindow: info.contextWindow,
+		maxInputTokens: info.maxInputTokens,
+		maxOutputTokens: info.maxTokens,
+		capabilities: [...capabilities],
+		metadata: Object.keys(metadata).length > 0 ? metadata : undefined,
+	};
+}
+
+function inferProtocol(spec: BuiltinSpec): ProviderProtocol {
+	if (spec.client === "openai") {
+		return "openai-responses";
+	}
+	switch (spec.family) {
+		case "openai":
+			return "openai-responses";
+		case "anthropic":
+		case "bedrock":
+			return "anthropic";
+		case "google":
+		case "vertex":
+			return "gemini";
+		default:
+			return "openai-chat";
+	}
+}
+
+function inferClient(spec: BuiltinSpec): ProviderClient {
+	if (spec.protocol === "openai-responses") {
+		return "openai";
+	}
+	switch (spec.family) {
+		case "openai":
+			return "openai";
+		case "anthropic":
+			return "anthropic";
+		case "google":
+			return "gemini";
+		case "vertex":
+			return "vertex";
+		case "bedrock":
+			return "bedrock";
+		case "mistral":
+		case "claude-code":
+		case "openai-codex":
+		case "opencode":
+		case "dify":
+		case "sap-ai-core":
+			return "ai-sdk-community";
+		default:
+			return "openai-compatible";
+	}
+}
+
+const OPENAI_COMPATIBLE_SPECS: BuiltinSpec[] = [
+	{
+		id: "openai-compatible",
+		name: "OpenAI Compatible",
+		description: "OpenAI-compatible chat completions endpoint",
+		family: "openai-compatible",
+		popular: 7,
+		capabilities: ["tools"],
+		defaultModelId: "gpt-4o",
+		apiKeyEnv: ["OPENAI_API_KEY"],
+		defaults: { baseUrl: "https://api.openai.com/v1" },
+	},
+	{
+		id: "cline",
+		name: "Cline",
+		description: "Cline API endpoint",
+		family: "openai-compatible",
+		popular: 1,
+		capabilities: ["reasoning", "prompt-cache", "tools", "oauth"],
+		modelsProviderId: "openrouter",
+		defaultModelId: "anthropic/claude-sonnet-4.6",
+		apiKeyEnv: ["CLINE_API_KEY"],
+		defaults: {
+			get baseUrl(): string {
+				return `${getClineEnvironmentConfig().apiBaseUrl}/api/v1`;
+			},
+		},
+		metadata: ANTHROPIC_AND_QWEN_CACHE_ROUTING_METADATA,
+	},
+	{
+		id: "deepseek",
+		name: "DeepSeek",
+		description: "Advanced AI models with reasoning capabilities",
+		family: "openai-compatible",
+		popular: 3,
+		capabilities: ["reasoning", "prompt-cache"],
+		defaultModelId: "deepseek-v4-flash",
+		apiKeyEnv: ["DEEPSEEK_API_KEY"],
+		defaults: { baseUrl: "https://api.deepseek.com/v1" },
+	},
+	{
+		id: "xai",
+		name: "xAI",
+		description: "Creator of Grok AI assistant",
+		family: "openai-compatible",
+		capabilities: ["reasoning"],
+		defaultModelId: "grok-4.20-0309-non-reasoning",
+		apiKeyEnv: ["XAI_API_KEY"],
+		defaults: { baseUrl: "https://api.x.ai/v1" },
+	},
+	{
+		id: "together",
+		name: "Together AI",
+		description: "Fast inference for open-source models",
+		family: "openai-compatible",
+		capabilities: ["reasoning"],
+		defaultModelId: "Qwen/Qwen3.5-397B-A17B",
+		apiKeyEnv: ["TOGETHER_API_KEY"],
+		defaults: { baseUrl: "https://api.together.xyz/v1" },
+	},
+	{
+		id: "fireworks",
+		name: "Fireworks AI",
+		description: "High-performance inference platform",
+		family: "openai-compatible",
+		defaultModelId: "accounts/fireworks/models/kimi-k2p6",
+		apiKeyEnv: ["FIREWORKS_API_KEY"],
+		defaults: { baseUrl: "https://api.fireworks.ai/inference/v1" },
+	},
+	{
+		id: "groq",
+		name: "Groq",
+		description: "Ultra-fast LPU inference",
+		family: "openai-compatible",
+		defaultModelId: "moonshotai/kimi-k2-instruct-0905",
+		apiKeyEnv: ["GROQ_API_KEY"],
+		defaults: { baseUrl: "https://api.groq.com/openai/v1" },
+	},
+	{
+		id: "poolside",
+		name: "Poolside",
+		description: "OpenAI-compatible code intelligence models",
+		family: "openai-compatible",
+		capabilities: ["tools", "reasoning"],
+		defaultModelId: "poolside/laguna-m.1",
+		apiKeyEnv: ["POOLSIDE_API_KEY"],
+		defaults: { baseUrl: "https://inference.poolside.ai/v1" },
+	},
+	{
+		id: "cerebras",
+		name: "Cerebras",
+		description: "Fast inference on Cerebras wafer-scale chips",
+		family: "openai-compatible",
+		defaultModelId: "zai-glm-4.7",
+		apiKeyEnv: ["CEREBRAS_API_KEY"],
+		defaults: { baseUrl: "https://api.cerebras.ai/v1" },
+	},
+	{
+		id: "sambanova",
+		name: "SambaNova",
+		description: "High-performance AI inference",
+		family: "openai-compatible",
+		apiKeyEnv: ["SAMBANOVA_API_KEY"],
+		modelsProviderId: "sambanova",
+		defaults: { baseUrl: "https://api.sambanova.ai/v1" },
+	},
+	{
+		id: "nebius",
+		name: "Nebius",
+		description: "European cloud AI infrastructure",
+		family: "openai-compatible",
+		defaultModelId: "nvidia/nemotron-3-super-120b-a12b",
+		apiKeyEnv: ["NEBIUS_API_KEY"],
+		defaults: { baseUrl: "https://api.studio.nebius.ai/v1" },
+	},
+	{
+		id: "baseten",
+		name: "Baseten",
+		description: "ML inference platform",
+		family: "openai-compatible",
+		apiKeyEnv: ["BASETEN_API_KEY"],
+		modelsProviderId: "baseten",
+		defaults: { baseUrl: "https://model-api.baseten.co/v1" },
+	},
+	{
+		id: "requesty",
+		name: "Requesty",
+		description: "AI router with multiple provider support",
+		family: "openai-compatible",
+		capabilities: ["reasoning"],
+		defaultModelId: "openai/gpt-5.4",
+		apiKeyEnv: ["REQUESTY_API_KEY"],
+		modelsProviderId: "requesty",
+		defaults: { baseUrl: "https://router.requesty.ai/v1" },
+	},
+	{
+		id: "litellm",
+		name: "LiteLLM",
+		description: "Self-hosted LLM proxy",
+		family: "openai-compatible",
+		protocol: "openai-responses",
+		popular: 8,
+		capabilities: ["prompt-cache"],
+		defaultModelId: "gpt-5.4",
+		apiKeyEnv: ["LITELLM_API_KEY"],
+		defaults: { baseUrl: "http://localhost:4000/v1" },
+	},
+	{
+		id: "huggingface",
+		name: "Hugging Face",
+		description: "Hugging Face inference API",
+		family: "openai-compatible",
+		defaultModelId: "MiniMaxAI/MiniMax-M2.5",
+		apiKeyEnv: ["HF_TOKEN"],
+		modelsProviderId: "huggingface",
+		defaults: { baseUrl: "https://api-inference.huggingface.co/v1" },
+	},
+	{
+		id: "vercel-ai-gateway",
+		name: "Vercel AI Gateway",
+		description: "Vercel's AI gateway service",
+		family: "openai-compatible",
+		capabilities: ["reasoning"],
+		defaultModelId: "alibaba/qwen3.6-plus",
+		apiKeyEnv: ["AI_GATEWAY_API_KEY"],
+		modelsProviderId: "vercel-ai-gateway",
+		defaults: { baseUrl: "https://ai-gateway.vercel.sh/v1" },
+		metadata: ANTHROPIC_AND_QWEN_CACHE_ROUTING_METADATA,
+	},
+	{
+		id: "v0",
+		name: "Vercel V0",
+		description:
+			"The Vercel provider gives you access to the v0 API, designed for building modern web applications.",
+		family: "openai-compatible",
+		protocol: "openai-responses",
+		capabilities: ["reasoning", "tools"],
+		defaultModelId: "v0-1.5-md",
+		apiKeyEnv: ["V0_API_KEY"],
+		modelsProviderId: "v0",
+		defaults: { baseUrl: "https://api.v0.dev/v1" },
+	},
+	{
+		id: "aihubmix",
+		name: "AI Hub Mix",
+		description: "AI model aggregator",
+		family: "openai-compatible",
+		defaultModelId: "gpt-4o",
+		apiKeyEnv: ["AIHUBMIX_API_KEY"],
+		modelsProviderId: "aihubmix",
+		defaults: { baseUrl: "https://api.aihubmix.com/v1" },
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	{
+		id: "hicap",
+		name: "HiCap",
+		description: "HiCap AI platform",
+		family: "openai-compatible",
+		defaultModelId: "hicap-pro",
+		apiKeyEnv: ["HICAP_API_KEY"],
+		defaults: { baseUrl: "https://api.hicap.ai/v1" },
+	},
+	{
+		id: "nousResearch",
+		name: "Nous Research",
+		description: "Open-source AI research lab",
+		family: "openai-compatible",
+		defaultModelId: "DeepHermes-3-Llama-3-3-70B-Preview",
+		apiKeyEnv: ["NOUS_RESEARCH_API_KEY", "NOUSRESEARCH_API_KEY"],
+		modelsProviderId: "nousResearch",
+		defaults: { baseUrl: "https://inference-api.nousresearch.com/v1" },
+	},
+	{
+		id: "huawei-cloud-maas",
+		name: "Huawei Cloud MaaS",
+		description: "Huawei's model-as-a-service platform",
+		family: "openai-compatible",
+		defaultModelId: "DeepSeek-R1",
+		apiKeyEnv: ["HUAWEI_CLOUD_MAAS_API_KEY"],
+		defaults: {
+			baseUrl: "https://infer-modelarts.cn-southwest-2.myhuaweicloud.com/v1",
+		},
+	},
+	{
+		id: "qwen",
+		name: "Alibaba Qwen",
+		description: "Alibaba Qwen platform models",
+		family: "openai-compatible",
+		capabilities: ["tools", "reasoning"],
+		defaultModelId: "qwen-plus-latest",
+		apiKeyEnv: ["QWEN_API_KEY"],
+		modelsProviderId: "qwen",
+		defaults: { baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1" },
+		configFields: QWEN_CONFIG_FIELDS,
+		metadata: QWEN_CACHE_ROUTING_METADATA,
+	},
+	{
+		id: "qwen-code",
+		name: "Alibaba Qwen Code",
+		description: "Qwen OAuth coding models",
+		family: "openai-compatible",
+		capabilities: ["tools", "reasoning"],
+		defaultModelId: "qwen3-coder-plus",
+		modelsProviderId: "qwen-code",
+		defaults: { baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1" },
+		configFields: QWEN_CONFIG_FIELDS,
+		metadata: QWEN_CACHE_ROUTING_METADATA,
+	},
+	{
+		id: "doubao",
+		name: "Doubao",
+		description: "Volcengine Ark platform models",
+		family: "openai-compatible",
+		capabilities: ["tools"],
+		defaultModelId: "doubao-1-5-pro-256k-250115",
+		apiKeyEnv: ["DOUBAO_API_KEY"],
+		modelsProviderId: "doubao",
+		defaults: { baseUrl: "https://ark.cn-beijing.volces.com/api/v3" },
+	},
+	{
+		id: "zai",
+		name: "Z.AI",
+		description: "Z.AI's family of LLMs",
+		family: "openai-compatible",
+		capabilities: ["reasoning"],
+		defaultModelId: "glm-5v-turbo",
+		apiKeyEnv: ["ZHIPU_API_KEY"],
+		modelsProviderId: "zai",
+		defaults: { baseUrl: "https://api.z.ai/api/paas/v4" },
+		metadata: GLM_THINKING_ROUTING_METADATA,
+	},
+	{
+		id: "zai-coding-plan",
+		name: "Z.AI Coding Plan",
+		description: "Z.AI's coding-focused models",
+		family: "openai-compatible",
+		capabilities: ["reasoning", "tools"],
+		defaultModelId: "glm-5v-turbo",
+		apiKeyEnv: ["ZHIPU_API_KEY"],
+		modelsProviderId: "zai-coding-plan",
+		defaults: { baseUrl: "https://api.z.ai/api/coding/paas/v4" },
+		metadata: GLM_THINKING_ROUTING_METADATA,
+	},
+	{
+		id: "moonshot",
+		name: "Moonshot",
+		description: "Moonshot AI Studio models",
+		family: "openai-compatible",
+		capabilities: ["tools", "reasoning"],
+		defaultModelId: "kimi-k2-0905-preview",
+		apiKeyEnv: ["MOONSHOT_API_KEY"],
+		modelsProviderId: "moonshot",
+		defaults: { baseUrl: "https://api.moonshot.ai/v1" },
+	},
+	{
+		id: "wandb",
+		name: "W&B by CoreWeave",
+		description: "Weights & Biases",
+		family: "openai-compatible",
+		capabilities: ["reasoning", "prompt-cache", "tools"],
+		defaultModelId: "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
+		apiKeyEnv: ["WANDB_API_KEY"],
+		modelsProviderId: "wandb",
+		defaults: { baseUrl: "https://api.inference.wandb.ai/v1" },
+	},
+	{
+		id: "xiaomi",
+		name: "Xiaomi",
+		description: "Xiaomi",
+		family: "openai-compatible",
+		protocol: "openai-responses",
+		capabilities: ["prompt-cache", "tools", "reasoning"],
+		defaultModelId: "mimo-v2-omni",
+		apiKeyEnv: ["XIAOMI_API_KEY"],
+		modelsProviderId: "xiaomi",
+		defaults: { baseUrl: "https://api.xiaomimimo.com/v1" },
+	},
+	{
+		id: "kilo",
+		name: "Kilo Gateway",
+		description: "Kilo Gateway",
+		family: "openai-compatible",
+		protocol: "openai-responses",
+		capabilities: ["prompt-cache", "reasoning", "tools"],
+		defaultModelId: "gpt-4o",
+		apiKeyEnv: ["KILO_GATEWAY_API_KEY"],
+		modelsProviderId: "kilo",
+		defaults: { baseUrl: "https://api.kilo.ai/api/gateway" },
+	},
+	{
+		id: "openrouter",
+		name: "OpenRouter",
+		description: "OpenRouter AI platform",
+		family: "openai-compatible",
+		popular: 5,
+		capabilities: ["reasoning", "prompt-cache"],
+		defaultModelId: "anthropic/claude-sonnet-4.6",
+		apiKeyEnv: ["OPENROUTER_API_KEY"],
+		modelsProviderId: "openrouter",
+		docsUrl: "https://openrouter.ai/models",
+		defaults: { baseUrl: "https://openrouter.ai/api/v1" },
+		metadata: ANTHROPIC_AND_QWEN_CACHE_ROUTING_METADATA,
+	},
+	{
+		id: "ollama",
+		name: "Ollama",
+		description: "Ollama Cloud and local LLM hosting",
+		family: "openai-compatible",
+		popular: 6,
+		defaultModelId: "",
+		apiKeyEnv: ["OLLAMA_API_KEY"],
+		defaults: { baseUrl: "http://localhost:11434/v1" },
+		modelsSourceUrl: "http://localhost:11434/api/tags",
+	},
+	{
+		id: "lmstudio",
+		name: "LM Studio",
+		description: "Local model inference with LM Studio",
+		family: "openai-compatible",
+		defaultModelId: "",
+		apiKeyEnv: ["LMSTUDIO_API_KEY"],
+		modelsProviderId: "lmstudio",
+		defaults: { baseUrl: "http://localhost:1234/v1" },
+		modelsSourceUrl: "http://localhost:1234/v1/models",
+	},
+	{
+		id: "oca",
+		name: "Oracle Code Assist",
+		description: "Oracle Code Assist (OCA) LiteLLM gateway",
+		family: "openai-compatible",
+		capabilities: ["reasoning", "prompt-cache", "tools"],
+		defaultModelId: "anthropic/claude-3-7-sonnet-20250219",
+		apiKeyEnv: ["OCA_API_KEY"],
+		modelsProviderId: "oca",
+		defaults: { baseUrl: DEFAULT_EXTERNAL_OCA_BASE_URL },
+		configFields: OCA_CONFIG_FIELDS,
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	{
+		id: "asksage",
+		name: "AskSage",
+		description: "AskSage platform",
+		family: "openai-compatible",
+		client: "fetch",
+		capabilities: ["tools"],
+		defaultModelId: "gpt-4o",
+		apiKeyEnv: ["ASKSAGE_API_KEY"],
+		modelsFactory: () => ({}),
+		defaults: { baseUrl: "https://api.asksage.ai/server" },
+	},
+];
+
+export const BUILTIN_SPECS: BuiltinSpec[] = [
+	{
+		id: "openai-native",
+		name: "OpenAI",
+		description: "Creator of GPT and ChatGPT",
+		family: "openai",
+		capabilities: ["reasoning"],
+		modelsProviderId: "openai-native",
+		defaultModelId: "gpt-5.4",
+		apiKeyEnv: ["OPENAI_API_KEY"],
+		defaults: { baseUrl: "https://api.openai.com/v1" },
+	},
+	{
+		id: "openai-codex",
+		name: "OpenAI ChatGPT Subscription",
+		description:
+			"OpenAI ChatGPT subscription access uses an OAuth device code flow.",
+		family: "openai",
+		popular: 2,
+		capabilities: ["reasoning", "oauth"],
+		defaultModelId: OPENAI_CODEX_DEFAULT_MODEL_ID,
+		modelsFactory: buildOpenAICodexModels,
+		defaults: { baseUrl: "https://chatgpt.com/backend-api/codex" },
+		configFields: [],
+		metadata: { usageCostDisplay: "hide" },
+	},
+	{
+		id: "openai-codex-cli",
+		name: "OpenAI Codex CLI",
+		description: "OpenAI Codex via the local Codex CLI provider",
+		family: "openai-codex",
+		capabilities: ["reasoning", "provider-tools", "local-auth"],
+		defaultModelId: "gpt-5.3-codex",
+		modelsProviderId: "openai",
+		defaults: { baseUrl: "https://chatgpt.com/backend-api/codex" },
+		configFields: [],
+		metadata: { usageCostDisplay: "hide" },
+	},
+	{
+		id: "anthropic",
+		name: "Anthropic",
+		description: "Creator of Claude, the AI assistant",
+		family: "anthropic",
+		popular: 4,
+		capabilities: ["reasoning", "prompt-cache"],
+		defaultModelId: "claude-sonnet-4-6",
+		apiKeyEnv: ["ANTHROPIC_API_KEY"],
+		modelsProviderId: "anthropic",
+		defaults: { baseUrl: "https://api.anthropic.com/v1" },
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	{
+		id: "claude-code",
+		name: "Claude Code",
+		description: "Use Claude Code SDK with Claude Pro/Max subscription",
+		family: "claude-code",
+		capabilities: ["reasoning"],
+		defaultModelId: "sonnet",
+		modelsFactory: buildClaudeCodeModels,
+		defaults: { baseUrl: "" },
+		configFields: [],
+	},
+	{
+		id: "gemini",
+		name: "Google Gemini",
+		description: "Google Gemini API",
+		family: "google",
+		popular: 9,
+		capabilities: ["reasoning", "prompt-cache"],
+		defaultModelId: "gemma-4-26b",
+		apiKeyEnv: ["GOOGLE_GENERATIVE_AI_API_KEY", "GEMINI_API_KEY"],
+		modelsProviderId: "gemini",
+		defaults: { baseUrl: "https://generativelanguage.googleapis.com/v1beta" },
+	},
+	{
+		id: "vertex",
+		name: "Google Vertex AI",
+		description: "Google Cloud Vertex AI",
+		family: "vertex",
+		capabilities: ["reasoning", "prompt-cache"],
+		apiKeyEnv: [
+			"GCP_PROJECT_ID",
+			"GOOGLE_CLOUD_PROJECT",
+			"GOOGLE_APPLICATION_CREDENTIALS",
+			"GEMINI_API_KEY",
+			"GOOGLE_API_KEY",
+			"GOOGLE_VERTEX_PROJECT",
+			"GOOGLE_VERTEX_LOCATION",
+		],
+		modelsProviderId: "vertex",
+		configFields: VERTEX_CONFIG_FIELDS,
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	{
+		id: "bedrock",
+		name: "AWS Bedrock",
+		description: "Amazon Bedrock managed foundation models",
+		family: "bedrock",
+		popular: 7,
+		capabilities: ["reasoning", "prompt-cache"],
+		defaultModelId: "minimax.minimax-m2.5",
+		apiKeyEnv: [
+			"AWS_BEARER_TOKEN_BEDROCK",
+			"AWS_REGION",
+			"AWS_ACCESS_KEY_ID",
+			"AWS_SECRET_ACCESS_KEY",
+			"AWS_SESSION_TOKEN",
+		],
+		modelsProviderId: "bedrock",
+		configFields: BEDROCK_CONFIG_FIELDS,
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	{
+		id: "mistral",
+		name: "Mistral",
+		description: "Mistral AI models via AI SDK provider",
+		family: "mistral",
+		capabilities: ["reasoning"],
+		defaultModelId: "mistral-medium-latest",
+		apiKeyEnv: ["MISTRAL_API_KEY"],
+		modelsFactory: () => ({}),
+		defaults: { baseUrl: "https://api.mistral.ai/v1" },
+	},
+	{
+		id: "minimax",
+		name: "MiniMax",
+		description: "MiniMax models via Anthropic-compatible API",
+		family: "anthropic",
+		capabilities: ["tools", "reasoning", "prompt-cache"],
+		defaultModelId: "MiniMax-M2.5",
+		apiKeyEnv: ["MINIMAX_API_KEY"],
+		modelsProviderId: "minimax",
+		defaults: { baseUrl: "https://api.minimax.io/anthropic/v1" },
+		metadata: MINIMAX_THINKING_ROUTING_METADATA,
+	},
+	{
+		id: "opencode",
+		name: "OpenCode",
+		description: "OpenCode SDK multi-provider runtime",
+		family: "opencode",
+		capabilities: ["reasoning", "oauth"],
+		defaultModelId: "openai/gpt-5.4",
+		modelsProviderId: "opencode",
+		defaults: { baseUrl: "" },
+		configFields: [],
+	},
+	{
+		id: "dify",
+		name: "Dify",
+		description: "Dify workflow/application provider via AI SDK",
+		family: "dify",
+		defaultModelId: "default",
+		apiKeyEnv: ["DIFY_API_KEY"],
+		modelsFactory: () => ({}),
+	},
+	{
+		id: "sapaicore",
+		name: "SAP AI Core",
+		description: "SAP AI Core inference and orchestration platform",
+		family: "sap-ai-core",
+		client: "ai-sdk-community",
+		capabilities: ["tools", "reasoning", "prompt-cache"],
+		defaultModelId: "anthropic--claude-3.5-sonnet",
+		apiKeyEnv: ["AICORE_SERVICE_KEY", "VCAP_SERVICES"],
+		modelsProviderId: "sapaicore",
+		metadata: ANTHROPIC_ROUTING_METADATA,
+	},
+	...OPENAI_COMPATIBLE_SPECS,
+];
+
+function getModels(spec: BuiltinSpec): Record<string, ModelInfo> {
+	if (spec.modelsFactory) {
+		return spec.modelsFactory();
+	}
+	if (spec.modelsProviderId) {
+		return generatedModels(spec.modelsProviderId);
+	}
+	return {};
+}
+
+function toModelCollection(spec: BuiltinSpec): ModelCollection {
+	const sourceModels = getModels(spec);
+	const capabilities = getProviderCapabilities(spec);
+	const metadata = getProviderMetadata(spec);
+	const models =
+		Object.keys(sourceModels).length > 0
+			? sourceModels
+			: spec.defaultModelId
+				? {
+						[spec.defaultModelId]: fallbackModelInfo(spec.defaultModelId, spec),
+					}
+				: {};
+	const modelIds = Object.keys(models);
+	const defaultModelId = spec.defaultModelId || modelIds[0] || "default";
+
+	return {
+		provider: {
+			id: spec.id,
+			name: spec.name,
+			description: spec.description,
+			protocol: spec.protocol ?? inferProtocol(spec),
+			baseUrl: spec.defaults?.baseUrl,
+			modelsSourceUrl: spec.modelsSourceUrl,
+			defaultModelId,
+			capabilities,
+			env: spec.apiKeyEnv ? [...spec.apiKeyEnv] : undefined,
+			client: spec.client ?? inferClient(spec),
+			source: "system",
+			metadata,
+		},
+		models,
+	};
+}
+
+export function toManifest(spec: BuiltinSpec): GatewayProviderManifest {
+	const collection = toModelCollection(spec);
+	const capabilities = getProviderCapabilities(spec);
+	const metadata = getProviderMetadata(spec);
+	const models = Object.values(collection.models).map((info) =>
+		modelInfoToGateway(spec.id, info),
+	);
+	const resolvedModels =
+		models.length > 0
+			? models
+			: [
+					{
+						id: collection.provider.defaultModelId || "default",
+						name: collection.provider.defaultModelId || "Default",
+						providerId: spec.id,
+						capabilities: ["text"] as GatewayModelCapability[],
+					},
+				];
+
+	return {
+		id: spec.id,
+		name: spec.name,
+		description: spec.description,
+		defaultModelId:
+			collection.provider.defaultModelId || resolvedModels[0]?.id || "default",
+		models: resolvedModels,
+		capabilities,
+		env: spec.env ?? ["browser", "node"],
+		api: spec.defaults?.baseUrl,
+		apiKeyEnv: spec.apiKeyEnv,
+		docsUrl: spec.docsUrl,
+		metadata,
+	};
+}
+
+export const BUILTIN_PROVIDER_COLLECTION_LIST: ModelCollection[] =
+	BUILTIN_SPECS.map(toModelCollection);
+
+export const BUILTIN_PROVIDER_COLLECTIONS_BY_ID: Record<
+	string,
+	ModelCollection
+> = Object.fromEntries(
+	BUILTIN_PROVIDER_COLLECTION_LIST.map((collection) => [
+		collection.provider.id,
+		collection,
+	]),
+);