@superlinked/sie-sdk 0.3.3 → 0.3.4

package/dist/index.cjs

@@ -109,6 +109,15 @@ var ModelLoadFailedError = class extends ServerError {
     this.attempts = options?.attempts ?? 1;
   }
 };
+var InputTooLongError = class extends RequestError {
+  /** The model that was requested */
+  model;
+  constructor(message, options) {
+    super(message, "INPUT_TOO_LONG", 400);
+    this.name = "InputTooLongError";
+    this.model = options?.model;
+  }
+};
 
 // src/internal/constants.ts
 var MSGPACK_CONTENT_TYPE = "application/msgpack";
@@ -380,6 +389,14 @@ async function throwIfModelLoadFailed(response, model) {
     attempts
   });
 }
+async function throwIfInputTooLong(response, model) {
+  if (response.status !== 400) return;
+  const detail = await getErrorDetail(response.clone());
+  if (!detail) return;
+  if (detail.code !== "INPUT_TOO_LONG") return;
+  const message = typeof detail.message === "string" ? detail.message : "Input exceeds the model's maximum token capacity";
+  throw new InputTooLongError(message, { model });
+}
 async function handleError(response, gpu) {
   const { status } = response;
   const detail = await getErrorDetail(response.clone());
@@ -413,6 +430,9 @@ async function handleError(response, gpu) {
     throw new ProvisioningError(message, gpu, retryAfter);
   }
   if (status >= HTTP_CLIENT_ERROR_MIN && status <= HTTP_CLIENT_ERROR_MAX) {
+    if (status === 400 && code === "INPUT_TOO_LONG") {
+      throw new InputTooLongError(message);
+    }
     throw new RequestError(message, code, status);
   }
   if (status >= HTTP_SERVER_ERROR_MIN && status <= HTTP_SERVER_ERROR_MAX) {
@@ -531,7 +551,7 @@ function parseCapacityInfo(data, gpuFilter) {
 }
 
 // src/version.ts
-var SDK_VERSION = "0.3.3";
+var SDK_VERSION = "0.3.4";
 
 // src/client.ts
 function sleep(ms) {
@@ -820,6 +840,9 @@ var SIEClient = class {
     if (options.threshold !== void 0) {
       params.threshold = options.threshold;
     }
+    if (options.adapterOptions !== void 0) {
+      params.options = options.adapterOptions;
+    }
     body.params = params;
     const waitForCapacity = options.waitForCapacity ?? this.defaultWaitForCapacity;
     const { pool, gpu } = this.parseGpuParam(options.gpu);
@@ -1236,6 +1259,7 @@ var SIEClient = class {
         continue;
       }
       await throwIfModelLoadFailed(response, model);
+      await throwIfInputTooLong(response, model);
       if (response.status === 503) {
         const clonedResponse = response.clone();
         const errorCode = await getErrorCode(clonedResponse);
@@ -1557,6 +1581,7 @@ function detectImageFormat(bytes) {
   return "unknown";
 }
 
+exports.InputTooLongError = InputTooLongError;
 exports.LoraLoadingError = LoraLoadingError;
 exports.ModelLoadFailedError = ModelLoadFailedError;
 exports.ModelLoadingError = ModelLoadingError;