@nemesis-oss/ollama-sdk 1.19.1 → 1.20.1

package/dist/index.cjs

@@ -303,6 +303,17 @@ var OllamaToolTimeoutError = class extends OllamaClientError {
     this.timeoutMs = options.timeoutMs;
   }
 };
+var OllamaCloudModelVersionError = class extends OllamaClientError {
+  model;
+  serverVersion;
+  minRequiredVersion;
+  constructor(message, options) {
+    super(message, { ...options, code: "cloud_model_version", retryable: false });
+    this.model = options.model;
+    this.serverVersion = options.serverVersion;
+    this.minRequiredVersion = options.minRequiredVersion;
+  }
+};
 var OllamaUnsupportedCapabilityError = class extends OllamaClientError {
   capability;
   constructor(message, options) {
@@ -1597,16 +1608,23 @@ async function checkEndpointHealth(endpoint, fetchImpl, timeoutMs = 5e3, middlew
 
 // src/capabilities/capabilities.ts
 function inferRuntimeMode(baseUrl) {
+  let host;
   try {
-    const url = new URL(baseUrl);
-    const host = url.hostname.toLowerCase().replace(/^\[|\]$/g, "");
-    if (host === "localhost" || host === "127.0.0.1" || host === "::1" || host.endsWith(".local") || host.startsWith("192.168.") || host.startsWith("10.")) {
-      return "local";
-    }
-    return "cloud";
+    host = new URL(baseUrl).hostname.toLowerCase().replace(/^\[|\]$/g, "");
   } catch {
     return "unknown";
   }
+  if (host === "ollama.com" || host.endsWith(".ollama.com")) return "cloud";
+  if (isPrivateNetworkHost(host)) return "local";
+  return "unknown";
+}
+function isPrivateNetworkHost(host) {
+  if (!host.includes(".") && !host.includes(":")) return true;
+  if (host.endsWith(".localhost") || host.endsWith(".local") || host.endsWith(".internal"))
+    return true;
+  if (host === "::1" || host === "0.0.0.0") return true;
+  if (host.includes(":") && /^f[cd][0-9a-f]{0,2}:/.test(host)) return true;
+  return /^(?:127|10)\.\d{1,3}\.\d{1,3}\.\d{1,3}$/.test(host) || /^192\.168\.\d{1,3}\.\d{1,3}$/.test(host) || /^172\.(?:1[6-9]|2\d|3[01])\.\d{1,3}\.\d{1,3}$/.test(host);
 }
 function extractContextLength(modelInfo) {
   if (!modelInfo) return void 0;
@@ -4233,6 +4251,9 @@ var ShowResponseSchema = zod.z.object({
   modified_at: zod.z.string().optional(),
   details: zod.z.object({}).optional(),
   template: zod.z.string().optional(),
+  renderer: zod.z.string().optional(),
+  parser: zod.z.string().optional(),
+  template_source: zod.z.string().optional(),
   capabilities: zod.z.array(zod.z.string()).optional(),
   model_info: zod.z.object({}).optional()
 });
@@ -4324,9 +4345,26 @@ var UsageMetricsSchema = zod.z.object({
   output_tokens: zod.z.number().optional()
 });
 
+// src/generated/models/UsageGroup.schema.ts
+var UsageGroupSchema = zod.z.intersection(
+  UsageMetricsSchema,
+  zod.z.object({
+    model: zod.z.string()
+  })
+);
+
+// src/generated/models/UsageSummary.schema.ts
+var UsageSummarySchema = zod.z.intersection(
+  UsageMetricsSchema,
+  zod.z.object({
+    groups: zod.z.array(UsageGroupSchema).optional(),
+    remainder: UsageMetricsSchema.optional()
+  })
+);
+
 // src/generated/models/UsageBucket.schema.ts
 var UsageBucketSchema = zod.z.intersection(
-  UsageMetricsSchema,
+  UsageSummarySchema,
   zod.z.object({
     from: zod.z.string(),
     until: zod.z.string(),
@@ -4339,9 +4377,10 @@ var UsageResponseSchema = zod.z.object({
   range: zod.z.enum(["24h", "7d", "30d"]),
   scope: zod.z.enum(["self", "team"]),
   granularity: zod.z.enum(["hour", "day"]),
+  group_by: zod.z.enum(["model"]).optional(),
   from: zod.z.string(),
   until: zod.z.string(),
-  totals: UsageMetricsSchema,
+  totals: UsageSummarySchema,
   buckets: zod.z.array(UsageBucketSchema)
 });
 var VersionResponseSchema = zod.z.object({
@@ -5784,6 +5823,29 @@ function createDecision(client) {
   };
 }
 
+// src/system-one.ts
+var MAX_SYSTEM_ONE_IMAGES_BYTES = 32 * 1024 * 1024;
+var MAX_SYSTEM_ONE_REQUEST_BYTES = 64 * 1024;
+var CLOUD_MODEL_SUFFIX = ":cloud";
+var CLOUD_MODEL_MIN_OLLAMA_VERSION = "0.40.5";
+function isCloudModel(model) {
+  return model.endsWith(CLOUD_MODEL_SUFFIX);
+}
+function compareOllamaVersions(a, b) {
+  const pa = a.split(".").map((s) => Number.parseInt(s, 10) || 0);
+  const pb = b.split(".").map((s) => Number.parseInt(s, 10) || 0);
+  for (let i = 0; i < Math.max(pa.length, pb.length); i += 1) {
+    const va = pa[i] ?? 0;
+    const vb = pb[i] ?? 0;
+    if (va !== vb) return va - vb;
+  }
+  return 0;
+}
+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;
+}
+
 // src/client.ts
 var logicalRequestSequence = 0;
 function createLogicalRequestId() {
@@ -6410,9 +6472,13 @@ var OllamaClient = class _OllamaClient {
    * 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
+   * Both version constraints are enforced client-side before the request
+   * is sent: the blanket 0.35.0 floor by the generated runtime (lazily
    * fetching `/api/version` when `enforceVersion` is `'auto'` or
-   * `'strict'`). Request size is checked client-side as a dual cap:
+   * `'strict'`), and the model-dependent 0.40.5 cloud gate by this
+   * wrapper — a cloud model through a local server older than 0.40.5
+   * throws {@link OllamaCloudModelVersionError} without a round-trip
+   * (ADR 0036). 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.
    *
@@ -6437,10 +6503,44 @@ var OllamaClient = class _OllamaClient {
    * `client.native.systemOne(request)` directly.
    */
   async systemOne(request) {
+    await this.assertCloudModelSupported(request.model);
     return this.native.systemOne(
       request
     );
   }
+  /**
+   * Cloud-model version gate for {@link systemOne} — throws
+   * {@link OllamaCloudModelVersionError} when a cloud model is requested
+   * through a local server older than {@link CLOUD_MODEL_MIN_OLLAMA_VERSION}
+   * (0.40.5). No-op for non-cloud models and for cloud-mode clients
+   * (hosted `ollama.com`). The `/api/version` probe goes through the normal
+   * failover machinery and is cached per client instance (reset on failure,
+   * so a later gated call retries), so at most one extra round-trip happens
+   * across all gated calls.
+   */
+  async assertCloudModelSupported(model) {
+    if (!isCloudModel(model)) return;
+    const candidates = this.registry.candidates();
+    const baseUrl = candidates[0]?.baseUrl ?? DEFAULT_BASE_URL;
+    if (inferRuntimeMode(baseUrl) !== "local") return;
+    let version;
+    try {
+      if (this.serverVersionCache === void 0) {
+        this.serverVersionCache = this.native.version().then((v) => v.version);
+      }
+      version = await this.serverVersionCache;
+    } catch {
+      this.serverVersionCache = void 0;
+      return;
+    }
+    if (version && compareOllamaVersions(version, CLOUD_MODEL_MIN_OLLAMA_VERSION) < 0) {
+      throw new OllamaCloudModelVersionError(
+        `System One cloud model "${model}" requires Ollama >= ${CLOUD_MODEL_MIN_OLLAMA_VERSION} when routed through a local server (this server reports ${version}). Upgrade Ollama, or point an endpoint at https://ollama.com.`,
+        { model, serverVersion: version, minRequiredVersion: CLOUD_MODEL_MIN_OLLAMA_VERSION }
+      );
+    }
+  }
+  serverVersionCache;
   // --- Web Endpoints ---
   /**
    * Ollama's hosted web search tool (`POST https://ollama.com/api/web_search`) — a fixed
@@ -6485,13 +6585,20 @@ var OllamaClient = class _OllamaClient {
    * endpoint's 429 with a `Retry-After` header ("Seconds to wait before retrying"),
    * which the SDK parses into `OllamaRateLimitError.retryAfterMs` and honors verbatim
    * when retrying. Query parameters mirror the official endpoint: `range`
-   * (`'24h' | '7d' | '30d'`, server default `'7d'`) and `scope`
-   * (`'self' | 'team'`, server default `'self'`; team scope requires a team admin).
+   * (`'24h' | '7d' | '30d'`, server default `'7d'`), `scope`
+   * (`'self' | 'team'`, server default `'self'`; team scope requires a team
+   * admin), and — since the 2026-10-22 pin (upstream 2476d5e) — `groupBy`
+   * (`group_by=model`), which makes `totals` and every bucket carry up to 10
+   * per-model `UsageGroup` entries plus a `remainder` roll-up, with the
+   * selection echoed back as `group_by` on the response.
    * Omitted options are left unset so the server applies its own defaults.
    *
    * ```ts
-   * const usage = await client.usage({ range: '24h' });
+   * const usage = await client.usage({ range: '24h', groupBy: 'model' });
    * console.log(usage.totals.request_count, usage.totals.usage_usd);
+   * for (const group of usage.totals.groups ?? []) {
+   *   console.log(group.model, group.usage_usd);
+   * }
    * for (const bucket of usage.buckets) {
    *   if (bucket.partial) continue; // current hour, still in progress
    *   // ...
@@ -6502,6 +6609,7 @@ var OllamaClient = class _OllamaClient {
     const params = new URLSearchParams();
     if (req.range !== void 0) params.set("range", req.range);
     if (req.scope !== void 0) params.set("scope", req.scope);
+    if (req.groupBy !== void 0) params.set("group_by", req.groupBy);
     const query = params.toString();
     return this.executeCloudRequest(
       (http, signal) => http.request({
@@ -8385,14 +8493,6 @@ function createModelAffinityScheduler(client, options) {
   return new ModelAffinityScheduler(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, 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;
 exports.ATTR_GEN_AI_REQUEST_MODEL = ATTR_GEN_AI_REQUEST_MODEL;
 exports.ATTR_GEN_AI_RESPONSE_MODEL = ATTR_GEN_AI_RESPONSE_MODEL;
@@ -8413,6 +8513,8 @@ exports.ATTR_URL_FULL = ATTR_URL_FULL;
 exports.Agent = Agent;
 exports.AnthropicCompatClient = AnthropicCompatClient;
 exports.AnthropicMessagesStream = AnthropicMessagesStream;
+exports.CLOUD_MODEL_MIN_OLLAMA_VERSION = CLOUD_MODEL_MIN_OLLAMA_VERSION;
+exports.CLOUD_MODEL_SUFFIX = CLOUD_MODEL_SUFFIX;
 exports.ConversationSession = ConversationSession;
 exports.DEFAULT_BACKOFF = DEFAULT_BACKOFF;
 exports.DEFAULT_BASE_URL = DEFAULT_BASE_URL;
@@ -8451,6 +8553,7 @@ exports.OllamaAuthError = OllamaAuthError;
 exports.OllamaBadGatewayError = OllamaBadGatewayError;
 exports.OllamaClient = OllamaClient;
 exports.OllamaClientError = OllamaClientError;
+exports.OllamaCloudModelVersionError = OllamaCloudModelVersionError;
 exports.OllamaGenericClientError = OllamaGenericClientError;
 exports.OllamaIncompatibleModelError = OllamaIncompatibleModelError;
 exports.OllamaMcpError = OllamaMcpError;
@@ -8496,6 +8599,7 @@ exports.checkEndpointHealth = checkEndpointHealth;
 exports.checkGenerateContext = checkGenerateContext;
 exports.compactConversationHistory = compactConversationHistory;
 exports.compareDraftPerformance = compareDraftPerformance;
+exports.compareOllamaVersions = compareOllamaVersions;
 exports.composeMiddleware = composeMiddleware;
 exports.computeProgressPercent = computeProgressPercent;
 exports.contextWarningMessage = contextWarningMessage;
@@ -8528,6 +8632,7 @@ exports.formatWebFetchResult = formatWebFetchResult;
 exports.formatWebSearchResult = formatWebSearchResult;
 exports.imageStringNeedsResolution = imageStringNeedsResolution;
 exports.inferRuntimeMode = inferRuntimeMode;
+exports.isCloudModel = isCloudModel;
 exports.isKeepAliveSugar = isKeepAliveSugar;
 exports.isKnownDoneReason = isKnownDoneReason;
 exports.isModelResident = isModelResident;