@hasna/switcher 0.1.0 → 0.1.1

package/dist/sdk.js

@@ -1,6 +1,6 @@
 // src/domain.ts
 import { z } from "zod";
-var harnessSchema = z.enum(["claude", "codex", "grok", "opencode2"]);
+var harnessSchema = z.enum(["claude", "codex", "grok", "opencode2", "pi"]);
 var protocolSchema = z.enum(["anthropic-messages", "openai-responses", "openai-chat"]);
 var idSchema = z.string().regex(/^[a-zA-Z0-9][a-zA-Z0-9._-]{0,79}$/);
 var label = z.string().min(1).max(200);
@@ -28,6 +28,7 @@ var modelSchema = z.object({
   id: z.string().min(1).max(300),
   name: label,
   description: z.string().max(8000).optional(),
+  available: z.boolean().optional(),
   contextWindow: z.number().int().positive().optional(),
   maxOutputTokens: z.number().int().positive().optional(),
   inputModalities: z.array(z.string().max(50)).max(20).optional(),
@@ -41,9 +42,32 @@ var providerInputSchema = z.object({
   protocol: protocolSchema,
   credentialEnv: envRef.optional(),
   authStyle: z.enum(["bearer", "x-api-key"]).default("bearer"),
+  catalogBaseUrl: urlSchema.optional(),
+  catalogFormat: z.enum(["openai", "ollama", "mistral", "together", "fireworks", "dashscope", "none"]).optional(),
+  catalogAuthStyle: z.enum(["bearer", "x-api-key", "none"]).optional(),
+  catalogCredentialEnv: envRef.optional(),
+  catalogAccountId: z.string().regex(/^[A-Za-z0-9][A-Za-z0-9._-]{0,127}$/).optional(),
   modelsPath: z.string().regex(/^[a-zA-Z0-9_/-]+$/).max(200).default("models"),
   manualModels: z.array(modelSchema).max(1e4).default([])
 }).strict().refine((p) => !p.modelsPath.split("/").includes("..") && !p.modelsPath.startsWith("/"), "modelsPath must be relative");
+var providerPresetSchema = z.object({
+  id: idSchema,
+  name: label,
+  credentialEnv: envRef.optional(),
+  credentialAliases: z.array(z.string().regex(/^[A-Z][A-Z0-9_]+$/)),
+  protocols: z.array(z.object({
+    protocol: protocolSchema,
+    baseUrl: urlSchema.optional(),
+    authStyle: z.enum(["bearer", "x-api-key"]),
+    catalogBaseUrl: urlSchema.optional(),
+    catalogFormat: z.enum(["openai", "ollama", "mistral", "together", "fireworks", "dashscope", "none"]),
+    catalogAuthStyle: z.enum(["bearer", "x-api-key", "none"]).optional(),
+    modelsPath: z.string(),
+    notes: z.array(z.string())
+  }).strict()).min(1),
+  sources: z.array(z.string().url()),
+  verification: z.literal("documented")
+}).strict();
 var profileInputSchema = z.object({
   id: idSchema,
   name: label,
@@ -71,6 +95,15 @@ class Fault extends Error {
     this.code = code;
   }
 }
+function parse(schema, value) {
+  const result = schema.safeParse(value);
+  if (!result.success)
+    throw new Fault(400, "invalid_request", result.error.issues.map((i) => `${i.path.join(".")}: ${i.message}`).join("; "));
+  return result.data;
+}
+function compatible(harness, protocol) {
+  return harness === "claude" ? protocol === "anthropic-messages" : harness === "codex" ? protocol === "openai-responses" : true;
+}
 
 // src/http.ts
 var MAX_BYTES = 16 * 1024 * 1024;
@@ -109,6 +142,118 @@ async function boundedJson(response, maxBytes = MAX_BYTES) {
 // src/sdk.ts
 import { resolveCredential } from "@hasna/contracts/client";
 
+// src/presets.ts
+var route = (protocol, baseUrl, options = {}) => ({
+  protocol,
+  baseUrl,
+  authStyle: "bearer",
+  catalogFormat: "openai",
+  modelsPath: "models",
+  notes: [],
+  ...options
+});
+var preset = (id, name, protocols, sources, alias) => parse(providerPresetSchema, {
+  id,
+  name,
+  protocols,
+  sources,
+  credentialAliases: alias ? [alias] : [],
+  credentialEnv: alias ? `SWITCHER_PROVIDER_${id.toUpperCase().replace(/-/g, "_")}` : undefined,
+  verification: "documented"
+});
+var providerPresets = [
+  preset("deepseek", "DeepSeek", [
+    route("openai-chat", "https://api.deepseek.com", { catalogBaseUrl: "https://api.deepseek.com" }),
+    route("anthropic-messages", "https://api.deepseek.com/anthropic/v1", { catalogBaseUrl: "https://api.deepseek.com" })
+  ], ["https://api-docs.deepseek.com/guides/anthropic_api", "https://api-docs.deepseek.com/api/list-models"], "DEEPSEEK_API_KEY"),
+  preset("openrouter", "OpenRouter", ["openai-chat", "openai-responses", "anthropic-messages"].map((protocol) => route(protocol, "https://openrouter.ai/api/v1", { catalogAuthStyle: "none" })), ["https://openrouter.ai/docs/api/api-reference/models/list-all-models-and-their-properties", "https://openrouter.ai/docs/guides/overview"], "OPENROUTER_API_KEY"),
+  preset("anthropic", "Anthropic", [route("anthropic-messages", "https://api.anthropic.com/v1", { authStyle: "x-api-key" })], ["https://platform.claude.com/docs/en/api/overview", "https://platform.claude.com/docs/en/api/models/list"], "ANTHROPIC_API_KEY"),
+  preset("openai", "OpenAI", [route("openai-responses", "https://api.openai.com/v1"), route("openai-chat", "https://api.openai.com/v1")], ["https://platform.openai.com/docs/api-reference/introduction", "https://platform.openai.com/docs/api-reference/models/list"], "OPENAI_API_KEY"),
+  preset("xai", "xAI", ["openai-chat", "openai-responses", "anthropic-messages"].map((protocol) => route(protocol, "https://api.x.ai/v1")), ["https://api.x.ai/docs/", "https://docs.x.ai/developers/model-capabilities/text/generate-text"], "XAI_API_KEY"),
+  preset("ollama", "Ollama", ["openai-chat", "openai-responses"].map((protocol) => route(protocol, "http://127.0.0.1:11434/v1", {
+    catalogBaseUrl: "http://127.0.0.1:11434",
+    modelsPath: "api/tags",
+    catalogFormat: "ollama",
+    catalogAuthStyle: "none",
+    notes: protocol === "openai-responses" ? ["Requires Ollama 0.13.3 or newer; only stateless Responses are supported."] : []
+  })), ["https://docs.ollama.com/api/openai-compatibility", "https://docs.ollama.com/api/tags"]),
+  preset("lmstudio", "LM Studio", ["openai-chat", "openai-responses", "anthropic-messages"].map((protocol) => route(protocol, "http://127.0.0.1:1234/v1", {
+    notes: ["Server authentication is optional. Use --credential-env when authentication is enabled."]
+  })), ["https://lmstudio.ai/docs/developer/rest"]),
+  preset("vllm", "vLLM (operator endpoint)", [
+    route("openai-chat", undefined, { notes: ["Pass the operator's OpenAI-compatible URL, normally ending in /v1. vLLM exposes Chat Completions at /chat/completions and GET /models under that prefix; configure --credential-env only when the server was started with --api-key or VLLM_API_KEY."] }),
+    route("openai-responses", undefined, { notes: ["Pass the operator's OpenAI-compatible URL, normally ending in /v1. Responses is available for supported text-generation models at /responses; configure --credential-env only when the server was started with --api-key or VLLM_API_KEY."] }),
+    route("anthropic-messages", undefined, { notes: ["Pass the operator's URL, normally ending in /v1. vLLM exposes the Anthropic Messages API at /messages for supported deployments. Chat templates and the configured tool parser determine whether streaming and tool calls work for a served model; configure --credential-env only when the server was started with --api-key or VLLM_API_KEY."] })
+  ], ["https://docs.vllm.ai/en/latest/serving/online_serving/openai_compatible_server/", "https://github.com/vllm-project/vllm/blob/main/docs/serving/online_serving/README.md"]),
+  preset("litellm", "LiteLLM Proxy (operator endpoint)", [
+    route("openai-chat", undefined, { notes: ["Pass the proxy's documented OpenAI-compatible base URL exactly; the official quick start uses the root server URL, while a deployment may add a prefix such as /v1. LiteLLM serves Chat Completions and GET /models relative to that URL; use --credential-env for the proxy's configured master key or other bearer token."] }),
+    route("openai-responses", undefined, { notes: ["Pass the proxy's documented OpenAI-compatible base URL exactly; LiteLLM documents the Responses API under the same proxy root or deployment prefix. Use --credential-env for the proxy's configured master key or other bearer token."] }),
+    route("anthropic-messages", undefined, { notes: ["Pass the complete inference prefix ending in /v1, including any deployment prefix. LiteLLM registers /v1/messages; Switcher appends /messages to the stored prefix and discovers /models there. This is a gateway adapter: streaming and tool behavior depend on the configured upstream model and route, so verify those capabilities independently. Use --credential-env for the proxy's configured master key or other bearer token."] })
+  ], ["https://docs.litellm.ai/", "https://docs.litellm.ai/docs/proxy/quick_start", "https://github.com/BerriAI/litellm/blob/main/litellm/proxy/anthropic_endpoints/endpoints.py"]),
+  preset("groq", "Groq", [route("openai-chat", "https://api.groq.com/openai/v1"), route("openai-responses", "https://api.groq.com/openai/v1", { notes: ["Responses is an upstream beta API."] })], ["https://console.groq.com/docs/api-reference"], "GROQ_API_KEY"),
+  preset("cerebras", "Cerebras", [route("openai-chat", "https://api.cerebras.ai/v1")], ["https://inference-docs.cerebras.ai/api-reference/chat-completions"], "CEREBRAS_API_KEY"),
+  preset("mistral", "Mistral", [route("openai-chat", "https://api.mistral.ai/v1", { catalogFormat: "mistral" })], ["https://docs.mistral.ai/api/endpoint/chat", "https://docs.mistral.ai/api/endpoint/models"], "MISTRAL_API_KEY"),
+  preset("together", "Together AI", [route("openai-chat", "https://api.together.ai/v1", { catalogFormat: "together" })], ["https://docs.together.ai/docs/inference/openai-compatibility", "https://docs.together.ai/reference/models"], "TOGETHER_API_KEY"),
+  preset("fireworks", "Fireworks AI", [
+    route("openai-chat", "https://api.fireworks.ai/inference/v1", { catalogFormat: "fireworks", notes: ["Model discovery uses GET /v1/accounts/{account_id}/models; provide --catalog-account-id or --catalog-url."] }),
+    route("openai-responses", "https://api.fireworks.ai/inference/v1", { catalogFormat: "fireworks", notes: ["Model discovery uses GET /v1/accounts/{account_id}/models; provide --catalog-account-id or --catalog-url."] }),
+    route("anthropic-messages", "https://api.fireworks.ai/inference/v1", { catalogFormat: "fireworks", notes: ["Model discovery uses GET /v1/accounts/{account_id}/models; provide --catalog-account-id or --catalog-url."] })
+  ], ["https://docs.fireworks.ai/getting-started/quickstart", "https://docs.fireworks.ai/tools-sdks/python-client/api-reference", "https://docs.fireworks.ai/api-reference/anthropic-messages", "https://docs.fireworks.ai/api-reference/post-chatcompletions", "https://docs.fireworks.ai/api-reference/list-models"], "FIREWORKS_API_KEY"),
+  preset("moonshot", "Moonshot AI (Kimi)", [route("openai-chat", "https://api.moonshot.ai/v1", { catalogBaseUrl: "https://api.moonshot.ai/v1" })], ["https://platform.kimi.ai/docs/api/chat", "https://platform.kimi.ai/docs/api/list-models"], "MOONSHOT_API_KEY"),
+  preset("dashscope", "Alibaba Cloud Model Studio (Qwen)", [route("openai-chat", "https://dashscope-us.aliyuncs.com/compatible-mode/v1", {
+    catalogFormat: "none",
+    notes: ["Inference keys and endpoints are region/workspace-specific. Model discovery uses GET /api/v1/models on a documented region or workspace catalog URL; pass --catalog-url and --catalog-format dashscope."]
+  })], ["https://help.aliyun.com/en/model-studio/base-url", "https://help.aliyun.com/en/model-studio/compatibility-of-openai-with-dashscope", "https://help.aliyun.com/en/model-studio/list-models"], "DASHSCOPE_API_KEY"),
+  preset("zai", "Z.AI", [route("openai-chat", "https://api.z.ai/api/paas/v4", {
+    catalogFormat: "none",
+    notes: ["The published API reference documents inference endpoints but no model-list endpoint; use manual models or provide an explicit catalog URL and parser."]
+  })], ["https://docs.z.ai/api-reference/introduction", "https://docs.z.ai/devpack/quick-start"], "ZAI_API_KEY"),
+  preset("minimax", "MiniMax", [
+    route("openai-chat", "https://api.minimax.cn/v1", { catalogBaseUrl: "https://api.minimax.cn/v1", notes: ["The Open Platform contract uses api.minimax.cn and Bearer auth. Token Plan documentation uses api.minimaxi.com; select that authority explicitly with --url and matching auth/credential settings."] }),
+    route("anthropic-messages", "https://api.minimax.cn/anthropic/v1", { authStyle: "x-api-key", catalogBaseUrl: "https://api.minimax.cn/anthropic/v1", catalogAuthStyle: "x-api-key", notes: ["The Open Platform contract uses api.minimax.cn/anthropic/v1 and X-Api-Key. Token Plan documentation uses api.minimaxi.com/anthropic; select that authority explicitly with --url and matching auth/credential settings."] })
+  ], ["https://platform.minimaxi.com/docs/api-reference/text-chat-anthropic", "https://platform.minimaxi.com/docs/api-reference/models/anthropic/list-models", "https://platform.minimaxi.com/docs/api-reference/models/openai/list-models", "https://platform.minimaxi.com/docs/token-plan/other-tools"], "MINIMAX_API_KEY"),
+  preset("siliconflow", "SiliconFlow", [route("openai-chat", "https://api.siliconflow.cn/v1", { catalogBaseUrl: "https://api.siliconflow.cn/v1", catalogFormat: "openai", notes: ["The official SiliconCloud OpenAPI contract defines GET /models with Bearer auth and data[] model rows; optional type and sub_type filters are available at the upstream endpoint."] })], ["https://github.com/siliconflow/siliconcloud/blob/main/openapi.yaml", "https://docs.siliconflow.cn/docs/userguide/quickstart", "https://docs.siliconflow.cn/docs/api/chat-completions-post"], "SILICONFLOW_API_KEY"),
+  ...["anthropic-messages", "openai-responses", "openai-chat"].map((protocol) => preset(`generic-${protocol}`, `Custom ${protocol}`, [route(protocol)], []))
+];
+function getProviderPreset(id) {
+  const entry = providerPresets.find((p) => p.id === id);
+  if (!entry)
+    throw new Fault(404, "preset_not_found", "Unknown provider preset. Use switcher providers presets to list available presets.");
+  return structuredClone(entry);
+}
+function providerFromPreset(presetId, options = {}) {
+  const preset2 = getProviderPreset(presetId);
+  const selected = preset2.protocols.find((p) => (!options.protocol || p.protocol === options.protocol) && (!options.harness || compatible(options.harness, p.protocol)));
+  if (!selected)
+    throw new Fault(422, "protocol_mismatch", "This provider preset has no native protocol compatible with the requested harness. Choose an explicitly compatible gateway.");
+  const baseUrl = options.baseUrl ?? selected.baseUrl;
+  if (!baseUrl)
+    throw new Fault(400, "endpoint_required", "This preset requires an explicit --url for its inference endpoint.");
+  if (options.baseUrl && selected.baseUrl && new URL(endpoint(options.baseUrl)).origin !== new URL(selected.baseUrl).origin && preset2.credentialEnv && !options.credentialEnv)
+    throw new Fault(422, "credential_authority", "An endpoint on another origin requires an explicit --credential-env reference.");
+  const suffix = selected.protocol === "anthropic-messages" ? "messages" : selected.protocol === "openai-responses" ? "responses" : "chat";
+  if (presetId === "fireworks" && !options.catalogBaseUrl && !options.catalogAccountId)
+    throw new Fault(400, "catalog_account_required", "Fireworks model discovery requires --catalog-account-id or an explicit --catalog-url.");
+  if (selected.catalogFormat === "none" && options.catalogFormat && !options.catalogBaseUrl)
+    throw new Fault(400, "catalog_url_required", "This preset requires an explicit --catalog-url when enabling a catalog parser.");
+  const catalogBaseUrl = options.catalogBaseUrl ?? (presetId === "fireworks" && options.catalogAccountId ? `https://api.fireworks.ai/v1/accounts/${encodeURIComponent(options.catalogAccountId)}` : options.baseUrl ? undefined : selected.catalogBaseUrl);
+  return parse(providerInputSchema, {
+    id: options.id ?? `${preset2.id}-${suffix}`,
+    name: preset2.name,
+    baseUrl,
+    protocol: selected.protocol,
+    credentialEnv: options.credentialEnv ?? preset2.credentialEnv,
+    authStyle: options.authStyle ?? selected.authStyle,
+    catalogBaseUrl,
+    catalogCredentialEnv: options.catalogCredentialEnv,
+    catalogAuthStyle: options.catalogAuthStyle ?? selected.catalogAuthStyle,
+    catalogFormat: options.catalogFormat ?? selected.catalogFormat,
+    catalogAccountId: options.catalogAccountId,
+    modelsPath: options.modelsPath ?? selected.modelsPath
+  });
+}
+
+// src/sdk.ts
 class SwitcherError extends Error {
   status;
   code;
@@ -131,7 +276,7 @@ class SwitcherClient {
     this.options = { ...options };
   }
   async request(method, path, body, options = {}) {
-    if (!/^\/v1\/[a-zA-Z0-9/?&=._%+-]+$/.test(path) || path.includes(".."))
+    if (!/^\/v1\/[a-zA-Z0-9/?&=._%+-]+$/.test(path) && !["/health", "/ready", "/version"].includes(path) || path.includes(".."))
       throw new Error("Invalid API path.");
     const apiKey = typeof this.options.apiKey === "function" ? this.options.apiKey() : this.options.apiKey;
     if (!apiKey || /[\r\n]/.test(apiKey))
@@ -162,6 +307,21 @@ class SwitcherClient {
   query(options = {}) {
     return new URLSearchParams(Object.entries(options).filter(([, v]) => v !== undefined).map(([k, v]) => [k, String(v)])).toString();
   }
+  health() {
+    return this.request("GET", "/health");
+  }
+  ready() {
+    return this.request("GET", "/ready");
+  }
+  version() {
+    return this.request("GET", "/version");
+  }
+  listProviderPresets() {
+    return this.request("GET", "/v1/provider-presets");
+  }
+  getProviderPreset(id) {
+    return this.request("GET", `/v1/provider-presets/${encodeURIComponent(id)}`);
+  }
   listProviders(options = {}) {
     return this.request("GET", `/v1/providers?${this.query(options)}`);
   }
@@ -221,6 +381,7 @@ function clientFromEnv(env = process.env) {
   return new SwitcherClient({ baseUrl: env.HASNA_SWITCHER_API_URL, apiKey: credential });
 }
 export {
+  providerFromPreset,
   clientFromEnv,
   SwitcherError,
   SwitcherClient