@nemesis-oss/ollama-sdk 1.18.0 → 1.19.0

package/dist/index.cjs

@@ -193,6 +193,11 @@ var OllamaAuthError = class extends OllamaClientError {
     super(message, { ...options, code: "auth_error", retryable: false });
   }
 };
+var OllamaPaymentRequiredError = class extends OllamaClientError {
+  constructor(message, options) {
+    super(message, { ...options, code: "payment_required", retryable: false, status: 402 });
+  }
+};
 var OllamaNotFoundError = class extends OllamaClientError {
   constructor(message, options) {
     super(message, { ...options, code: "not_found", retryable: false });
@@ -427,6 +432,9 @@ function statusToError(status, message, options) {
   if (status === 401 || status === 403) {
     return new OllamaAuthError(message, { ...options, status });
   }
+  if (status === 402) {
+    return new OllamaPaymentRequiredError(message, options);
+  }
   if (status === 404) {
     return new OllamaNotFoundError(message, { ...options, status });
   }
@@ -4425,14 +4433,25 @@ var OllamaRuntime = class {
       const hasImages = bodyHasImages(body);
       const maxBytesWithImages = req.operation.constraints?.maxRequestBytesWithImages;
       const maxBytesBase = req.operation.constraints?.maxRequestBytes;
-      const applicableLimit = hasImages && maxBytesWithImages !== void 0 ? maxBytesWithImages : maxBytesBase;
-      if (applicableLimit !== void 0 && serialized.byteLength > applicableLimit) {
+      const textAndSchemaBytes = hasImages ? serializeWithoutImages(body).byteLength : serialized.byteLength;
+      if (maxBytesBase !== void 0 && textAndSchemaBytes > maxBytesBase) {
         throw new OllamaRequestTooLargeError(
-          `Operation ${req.operation.operationId} (${req.operation.method} ${req.operation.path}) request body is ${serialized.byteLength} bytes, exceeding the contract limit of ${applicableLimit} bytes` + (hasImages ? " (images limit)." : "."),
+          `Operation ${req.operation.operationId} (${req.operation.method} ${req.operation.path}) text and schema (images excluded) is ${textAndSchemaBytes} bytes, exceeding the contract limit of ${maxBytesBase} bytes.`,
+          {
+            operationId: req.operation.operationId,
+            actualBytes: textAndSchemaBytes,
+            maxBytes: maxBytesBase,
+            request: { method: req.operation.method, url: req.operation.path }
+          }
+        );
+      }
+      if (maxBytesWithImages !== void 0 && serialized.byteLength > maxBytesWithImages) {
+        throw new OllamaRequestTooLargeError(
+          `Operation ${req.operation.operationId} (${req.operation.method} ${req.operation.path}) request body is ${serialized.byteLength} bytes, exceeding the contract limit of ${maxBytesWithImages} bytes (whole request, images included).`,
           {
             operationId: req.operation.operationId,
             actualBytes: serialized.byteLength,
-            maxBytes: applicableLimit,
+            maxBytes: maxBytesWithImages,
             request: { method: req.operation.method, url: req.operation.path }
           }
         );
@@ -4574,6 +4593,13 @@ function bodyHasImages(body) {
   const images = obj.images;
   return Array.isArray(images) && images.length > 0;
 }
+function serializeWithoutImages(body) {
+  if (typeof body === "object" && body !== null && !Array.isArray(body)) {
+    const { images: _images, ...rest } = body;
+    return serializeForByteCount(rest);
+  }
+  return serializeForByteCount(body);
+}
 function resolvePathParams(pathTemplate, pathParams) {
   if (!pathTemplate.includes("{")) return pathTemplate;
   const params = pathParams ?? {};
@@ -5326,7 +5352,7 @@ var systemOneOp = {
   path: "/v1/systemone",
   request: { $ref: "#/schemas/SystemOneRequest" },
   response: { $ref: "#/schemas/SystemOneResponse" },
-  environment: { local: true, cloud: false },
+  environment: { local: true, cloud: true },
   transport: { mode: "json", streaming: false },
   capabilities: {
     thinking: "unsupported",
@@ -5344,8 +5370,10 @@ var systemOneOp = {
   status: { documented: true, experimental: true },
   domain: "native",
   notes: [
-    "Local-only; Ollama Cloud rejects /v1/systemone.",
-    "64 KiB request limit without images, 32 MiB with images (includes JSON/base64 overhead).",
+    "Local and cloud: model accepts cloud models (e.g. clef:cloud); cloud models require Ollama >= 0.40.5 or the hosted https://ollama.com endpoint.",
+    "Local dual cap: text and schema (excluding images) within 64 KiB AND the whole request within 32 MiB, including base64 and JSON overhead.",
+    "Cloud requests: paid plan or usage credits required; up to 64 questions named with up to 100 ASCII letters, digits, `_`, `.`, or `-`; up to 4 images of at most 4 MiB and 16 megapixels each (8 MiB total); requests within 13 MiB; input within the model's 65,536-token context window.",
+    "401 when a cloud model is requested without signing in or with an invalid API key (through a local server the response includes a signin_url); 402 when the cloud model is not included in the Free plan or a payment fails.",
     "No tools, no generation controls (temperature, top_p, etc.).",
     "Vision is model-dependent: images are accepted by vision-weighted models (Clef/Clef Flash) but not by text-only models.",
     "Experimental: behavioral surface may change between Ollama releases."
@@ -6301,11 +6329,15 @@ var OllamaClient = class _OllamaClient {
    * 1–64 typed questions (choice, noul, or score) against a provided state
    * and returns typed answers with probabilities and confidence.
    *
-   * Local-only; requires Ollama >= 0.35.0. The runtime enforces the
-   * version constraint automatically (lazily fetching `/api/version` when
-   * `enforceVersion` is `'auto'` or `'strict'`) and rejects cloud-mode
-   * calls. Request size is checked client-side: 64 KiB without images,
-   * 32 MiB with images.
+   * Local and cloud (since the 2026-10-16 upstream System One cloud
+   * launch): `model` also accepts cloud models such as `clef:cloud` —
+   * cloud models require Ollama >= 0.40.5 or the hosted
+   * `https://ollama.com` endpoint. Requires Ollama >= 0.35.0 locally.
+   * The runtime enforces the version constraint automatically (lazily
+   * fetching `/api/version` when `enforceVersion` is `'auto'` or
+   * `'strict'`). Request size is checked client-side as a dual cap:
+   * text and schema (images excluded) within 64 KiB AND the whole
+   * request within 32 MiB.
    *
    * The generic `Q` parameter captures the caller's question map at
    * compile time, giving key-safe answer access:
@@ -8256,8 +8288,9 @@ function createModelAffinityScheduler(client, options) {
 // src/system-one.ts
 var MAX_SYSTEM_ONE_IMAGES_BYTES = 32 * 1024 * 1024;
 var MAX_SYSTEM_ONE_REQUEST_BYTES = 64 * 1024;
-function estimateSystemOneRequestBytes(request) {
-  return new TextEncoder().encode(JSON.stringify(request)).byteLength;
+function estimateSystemOneRequestBytes(request, options) {
+  const body = options?.excludeImages && request.images !== void 0 ? { ...request, images: void 0 } : request;
+  return new TextEncoder().encode(JSON.stringify(body)).byteLength;
 }
 
 exports.ATTR_GEN_AI_OPERATION_NAME = ATTR_GEN_AI_OPERATION_NAME;
@@ -8325,6 +8358,7 @@ exports.OllamaModelRoutingError = OllamaModelRoutingError;
 exports.OllamaNetworkError = OllamaNetworkError;
 exports.OllamaNotFoundError = OllamaNotFoundError;
 exports.OllamaOverloadedError = OllamaOverloadedError;
+exports.OllamaPaymentRequiredError = OllamaPaymentRequiredError;
 exports.OllamaQuotaExceededError = OllamaQuotaExceededError;
 exports.OllamaRateLimitError = OllamaRateLimitError;
 exports.OllamaRequestTooLargeError = OllamaRequestTooLargeError;