@ai-sdk/openai 3.0.106 → 3.0.107

package/dist/internal/index.d.mts

@@ -298,23 +298,24 @@ declare const openaiResponsesChunkSchema: _ai_sdk_provider_utils.LazySchema<{
 } | {
     type: "response.completed" | "response.incomplete";
     response: {
-        usage: {
+        incomplete_details?: {
+            reason: string;
+        } | null | undefined;
+        usage?: (Record<string, JSONValue | undefined> & {
             input_tokens: number;
             output_tokens: number;
-            input_tokens_details?: {
+            input_tokens_details?: (Record<string, JSONValue | undefined> & {
                 cached_tokens?: number | null | undefined;
                 cache_write_tokens?: number | null | undefined;
                 orchestration_input_tokens?: number | null | undefined;
                 orchestration_input_cached_tokens?: number | null | undefined;
-            } | null | undefined;
-            output_tokens_details?: {
+            }) | null | undefined;
+            output_tokens_details?: (Record<string, JSONValue | undefined> & {
                 reasoning_tokens?: number | null | undefined;
                 orchestration_output_tokens?: number | null | undefined;
-            } | null | undefined;
-        };
-        incomplete_details?: {
-            reason: string;
-        } | null | undefined;
+            }) | null | undefined;
+            total_tokens?: number | undefined;
+        }) | null | undefined;
         reasoning?: {
             context?: string | null | undefined;
         } | null | undefined;
@@ -331,20 +332,21 @@ declare const openaiResponsesChunkSchema: _ai_sdk_provider_utils.LazySchema<{
         incomplete_details?: {
             reason: string;
         } | null | undefined;
-        usage?: {
+        usage?: (Record<string, JSONValue | undefined> & {
             input_tokens: number;
             output_tokens: number;
-            input_tokens_details?: {
+            input_tokens_details?: (Record<string, JSONValue | undefined> & {
                 cached_tokens?: number | null | undefined;
                 cache_write_tokens?: number | null | undefined;
                 orchestration_input_tokens?: number | null | undefined;
                 orchestration_input_cached_tokens?: number | null | undefined;
-            } | null | undefined;
-            output_tokens_details?: {
+            }) | null | undefined;
+            output_tokens_details?: (Record<string, JSONValue | undefined> & {
                 reasoning_tokens?: number | null | undefined;
                 orchestration_output_tokens?: number | null | undefined;
-            } | null | undefined;
-        } | null | undefined;
+            }) | null | undefined;
+            total_tokens?: number | undefined;
+        }) | null | undefined;
         reasoning?: {
             context?: string | null | undefined;
         } | null | undefined;

package/dist/internal/index.d.ts

@@ -298,23 +298,24 @@ declare const openaiResponsesChunkSchema: _ai_sdk_provider_utils.LazySchema<{
 } | {
     type: "response.completed" | "response.incomplete";
     response: {
-        usage: {
+        incomplete_details?: {
+            reason: string;
+        } | null | undefined;
+        usage?: (Record<string, JSONValue | undefined> & {
             input_tokens: number;
             output_tokens: number;
-            input_tokens_details?: {
+            input_tokens_details?: (Record<string, JSONValue | undefined> & {
                 cached_tokens?: number | null | undefined;
                 cache_write_tokens?: number | null | undefined;
                 orchestration_input_tokens?: number | null | undefined;
                 orchestration_input_cached_tokens?: number | null | undefined;
-            } | null | undefined;
-            output_tokens_details?: {
+            }) | null | undefined;
+            output_tokens_details?: (Record<string, JSONValue | undefined> & {
                 reasoning_tokens?: number | null | undefined;
                 orchestration_output_tokens?: number | null | undefined;
-            } | null | undefined;
-        };
-        incomplete_details?: {
-            reason: string;
-        } | null | undefined;
+            }) | null | undefined;
+            total_tokens?: number | undefined;
+        }) | null | undefined;
         reasoning?: {
             context?: string | null | undefined;
         } | null | undefined;
@@ -331,20 +332,21 @@ declare const openaiResponsesChunkSchema: _ai_sdk_provider_utils.LazySchema<{
         incomplete_details?: {
             reason: string;
         } | null | undefined;
-        usage?: {
+        usage?: (Record<string, JSONValue | undefined> & {
             input_tokens: number;
             output_tokens: number;
-            input_tokens_details?: {
+            input_tokens_details?: (Record<string, JSONValue | undefined> & {
                 cached_tokens?: number | null | undefined;
                 cache_write_tokens?: number | null | undefined;
                 orchestration_input_tokens?: number | null | undefined;
                 orchestration_input_cached_tokens?: number | null | undefined;
-            } | null | undefined;
-            output_tokens_details?: {
+            }) | null | undefined;
+            output_tokens_details?: (Record<string, JSONValue | undefined> & {
                 reasoning_tokens?: number | null | undefined;
                 orchestration_output_tokens?: number | null | undefined;
-            } | null | undefined;
-        } | null | undefined;
+            }) | null | undefined;
+            total_tokens?: number | undefined;
+        }) | null | undefined;
         reasoning?: {
             context?: string | null | undefined;
         } | null | undefined;

package/dist/internal/index.js

@@ -125,6 +125,22 @@ function getGptVersion(modelId) {
 
 // src/openai-stream-error.ts
 var import_provider = require("@ai-sdk/provider");
+function createOpenAIProviderStreamError(frame) {
+  var _a2, _b;
+  const streamError = parseStreamError(frame);
+  if (streamError == null) {
+    return void 0;
+  }
+  const statusCode = getStatusCode(streamError);
+  return {
+    message: streamError.message,
+    type: (_a2 = streamError.type) != null ? _a2 : void 0,
+    code: (_b = streamError.code) != null ? _b : void 0,
+    statusCode,
+    isRetryable: isRetryableStreamError(streamError, statusCode),
+    data: frame
+  };
+}
 async function throwIfOpenAIStreamErrorBeforeOutput({
   stream,
   getError,
@@ -196,16 +212,14 @@ function parseStreamError(frame) {
     return typeof (responseError == null ? void 0 : responseError.message) === "string" ? {
       message: responseError.message,
       code: getStringOrNumber(responseError.code),
-      type: "response.failed",
-      frame
+      type: "response.failed"
     } : void 0;
   }
   const error = (_a2 = asRecord(value.error)) != null ? _a2 : value;
   return typeof error.message === "string" && (asRecord(value.error) != null || typeof error.type === "string" || "code" in error || "param" in error) ? {
     message: error.message,
     code: getStringOrNumber(error.code),
-    type: typeof error.type === "string" ? error.type : void 0,
-    frame
+    type: typeof error.type === "string" ? error.type : void 0
   } : void 0;
 }
 function getStatusCode(error) {
@@ -245,6 +259,15 @@ function getStringOrNumber(value) {
 function isHttpErrorStatusCode(value) {
   return Number.isInteger(value) && value >= 400 && value <= 599;
 }
+function isRetryableStatusCode(statusCode) {
+  return statusCode === 408 || statusCode === 409 || statusCode === 429 || statusCode >= 500;
+}
+function isRetryableStreamError(error, statusCode) {
+  if (error.code === "insufficient_quota" || error.type === "insufficient_quota") {
+    return false;
+  }
+  return isRetryableStatusCode(statusCode);
+}
 
 // src/chat/convert-openai-chat-usage.ts
 function convertOpenAIChatUsage(usage) {
@@ -3952,6 +3975,31 @@ var jsonValueSchema = import_v418.z.lazy(
     import_v418.z.record(import_v418.z.string(), jsonValueSchema.optional())
   ])
 );
+var jsonObjectSchema = import_v418.z.record(import_v418.z.string(), jsonValueSchema.optional());
+var openaiResponsesUsageSchema = import_v418.z.intersection(
+  jsonObjectSchema,
+  import_v418.z.object({
+    input_tokens: import_v418.z.number(),
+    input_tokens_details: import_v418.z.intersection(
+      jsonObjectSchema,
+      import_v418.z.object({
+        cached_tokens: import_v418.z.number().nullish(),
+        cache_write_tokens: import_v418.z.number().nullish(),
+        orchestration_input_tokens: import_v418.z.number().nullish(),
+        orchestration_input_cached_tokens: import_v418.z.number().nullish()
+      })
+    ).nullish(),
+    output_tokens: import_v418.z.number(),
+    output_tokens_details: import_v418.z.intersection(
+      jsonObjectSchema,
+      import_v418.z.object({
+        reasoning_tokens: import_v418.z.number().nullish(),
+        orchestration_output_tokens: import_v418.z.number().nullish()
+      })
+    ).nullish(),
+    total_tokens: import_v418.z.number().optional()
+  })
+);
 function isRecord(value) {
   return value != null && typeof value === "object" && !Array.isArray(value);
 }
@@ -4063,20 +4111,7 @@ var openaiResponsesChunkSchema = (0, import_provider_utils26.lazySchema)(
         type: import_v418.z.enum(["response.completed", "response.incomplete"]),
         response: import_v418.z.object({
           incomplete_details: import_v418.z.object({ reason: import_v418.z.string() }).nullish(),
-          usage: import_v418.z.object({
-            input_tokens: import_v418.z.number(),
-            input_tokens_details: import_v418.z.object({
-              cached_tokens: import_v418.z.number().nullish(),
-              cache_write_tokens: import_v418.z.number().nullish(),
-              orchestration_input_tokens: import_v418.z.number().nullish(),
-              orchestration_input_cached_tokens: import_v418.z.number().nullish()
-            }).nullish(),
-            output_tokens: import_v418.z.number(),
-            output_tokens_details: import_v418.z.object({
-              reasoning_tokens: import_v418.z.number().nullish(),
-              orchestration_output_tokens: import_v418.z.number().nullish()
-            }).nullish()
-          }),
+          usage: openaiResponsesUsageSchema.nullish(),
           reasoning: import_v418.z.object({
             context: import_v418.z.string().nullish()
           }).nullish(),
@@ -4092,20 +4127,7 @@ var openaiResponsesChunkSchema = (0, import_provider_utils26.lazySchema)(
             message: import_v418.z.string()
           }).nullish(),
           incomplete_details: import_v418.z.object({ reason: import_v418.z.string() }).nullish(),
-          usage: import_v418.z.object({
-            input_tokens: import_v418.z.number(),
-            input_tokens_details: import_v418.z.object({
-              cached_tokens: import_v418.z.number().nullish(),
-              cache_write_tokens: import_v418.z.number().nullish(),
-              orchestration_input_tokens: import_v418.z.number().nullish(),
-              orchestration_input_cached_tokens: import_v418.z.number().nullish()
-            }).nullish(),
-            output_tokens: import_v418.z.number(),
-            output_tokens_details: import_v418.z.object({
-              reasoning_tokens: import_v418.z.number().nullish(),
-              orchestration_output_tokens: import_v418.z.number().nullish()
-            }).nullish()
-          }).nullish(),
+          usage: openaiResponsesUsageSchema.nullish(),
           reasoning: import_v418.z.object({
             context: import_v418.z.string().nullish()
           }).nullish(),
@@ -4871,20 +4893,7 @@ var openaiResponsesResponseSchema = (0, import_provider_utils26.lazySchema)(
         context: import_v418.z.string().nullish()
       }).nullish(),
       incomplete_details: import_v418.z.object({ reason: import_v418.z.string() }).nullish(),
-      usage: import_v418.z.object({
-        input_tokens: import_v418.z.number(),
-        input_tokens_details: import_v418.z.object({
-          cached_tokens: import_v418.z.number().nullish(),
-          cache_write_tokens: import_v418.z.number().nullish(),
-          orchestration_input_tokens: import_v418.z.number().nullish(),
-          orchestration_input_cached_tokens: import_v418.z.number().nullish()
-        }).nullish(),
-        output_tokens: import_v418.z.number(),
-        output_tokens_details: import_v418.z.object({
-          reasoning_tokens: import_v418.z.number().nullish(),
-          orchestration_output_tokens: import_v418.z.number().nullish()
-        }).nullish()
-      }).optional()
+      usage: openaiResponsesUsageSchema.nullish()
     })
   )
 );
@@ -6795,7 +6804,7 @@ var OpenAIResponsesLanguageModel = class {
             controller.enqueue({ type: "stream-start", warnings });
           },
           transform(chunk, controller) {
-            var _a3, _b2, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z, _A, _B, _C, _D, _E, _F, _G, _H, _I, _J, _K, _L, _M, _N, _O, _P;
+            var _a3, _b2, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z, _A, _B, _C, _D, _E, _F, _G, _H, _I, _J, _K, _L, _M, _N, _O, _P, _Q, _R, _S;
             if (options.includeRawChunks) {
               controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
             }
@@ -7616,15 +7625,15 @@ var OpenAIResponsesLanguageModel = class {
                   raw: (_z = (_y = value.response.incomplete_details) == null ? void 0 : _y.reason) != null ? _z : void 0
                 };
               }
-              usage = value.response.usage;
+              usage = (_A = value.response.usage) != null ? _A : void 0;
               if (typeof value.response.service_tier === "string") {
                 serviceTier = value.response.service_tier;
               }
-              if (((_A = value.response.reasoning) == null ? void 0 : _A.context) != null) {
+              if (((_B = value.response.reasoning) == null ? void 0 : _B.context) != null) {
                 reasoningContext = value.response.reasoning.context;
               }
             } else if (isResponseFailedChunk(value)) {
-              const incompleteReason = (_B = value.response.incomplete_details) == null ? void 0 : _B.reason;
+              const incompleteReason = (_C = value.response.incomplete_details) == null ? void 0 : _C.reason;
               finishReason = {
                 unified: incompleteReason ? mapOpenAIResponseFinishReason({
                   finishReason: incompleteReason,
@@ -7632,23 +7641,24 @@ var OpenAIResponsesLanguageModel = class {
                 }) : "error",
                 raw: incompleteReason != null ? incompleteReason : "error"
               };
-              usage = (_C = value.response.usage) != null ? _C : void 0;
-              if (((_D = value.response.reasoning) == null ? void 0 : _D.context) != null) {
+              usage = (_D = value.response.usage) != null ? _D : void 0;
+              if (((_E = value.response.reasoning) == null ? void 0 : _E.context) != null) {
                 reasoningContext = value.response.reasoning.context;
               }
               if (!encounteredStreamError && value.response.error != null) {
                 encounteredStreamError = true;
+                const error = {
+                  type: "response.failed",
+                  sequence_number: value.sequence_number,
+                  response: {
+                    error: value.response.error,
+                    incomplete_details: value.response.incomplete_details,
+                    service_tier: value.response.service_tier
+                  }
+                };
                 controller.enqueue({
                   type: "error",
-                  error: {
-                    type: "response.failed",
-                    sequence_number: value.sequence_number,
-                    response: {
-                      error: value.response.error,
-                      incomplete_details: value.response.incomplete_details,
-                      service_tier: value.response.service_tier
-                    }
-                  }
+                  error: (_F = createOpenAIProviderStreamError(error)) != null ? _F : error
                 });
               }
             } else if (isResponseAnnotationAddedChunk(value)) {
@@ -7657,7 +7667,7 @@ var OpenAIResponsesLanguageModel = class {
                 controller.enqueue({
                   type: "source",
                   sourceType: "url",
-                  id: (_G = (_F = (_E = self.config).generateId) == null ? void 0 : _F.call(_E)) != null ? _G : (0, import_provider_utils36.generateId)(),
+                  id: (_I = (_H = (_G = self.config).generateId) == null ? void 0 : _H.call(_G)) != null ? _I : (0, import_provider_utils36.generateId)(),
                   url: value.annotation.url,
                   title: value.annotation.title
                 });
@@ -7665,7 +7675,7 @@ var OpenAIResponsesLanguageModel = class {
                 controller.enqueue({
                   type: "source",
                   sourceType: "document",
-                  id: (_J = (_I = (_H = self.config).generateId) == null ? void 0 : _I.call(_H)) != null ? _J : (0, import_provider_utils36.generateId)(),
+                  id: (_L = (_K = (_J = self.config).generateId) == null ? void 0 : _K.call(_J)) != null ? _L : (0, import_provider_utils36.generateId)(),
                   mediaType: "text/plain",
                   title: value.annotation.filename,
                   filename: value.annotation.filename,
@@ -7681,7 +7691,7 @@ var OpenAIResponsesLanguageModel = class {
                 controller.enqueue({
                   type: "source",
                   sourceType: "document",
-                  id: (_M = (_L = (_K = self.config).generateId) == null ? void 0 : _L.call(_K)) != null ? _M : (0, import_provider_utils36.generateId)(),
+                  id: (_O = (_N = (_M = self.config).generateId) == null ? void 0 : _N.call(_M)) != null ? _O : (0, import_provider_utils36.generateId)(),
                   mediaType: "text/plain",
                   title: value.annotation.filename,
                   filename: value.annotation.filename,
@@ -7697,7 +7707,7 @@ var OpenAIResponsesLanguageModel = class {
                 controller.enqueue({
                   type: "source",
                   sourceType: "document",
-                  id: (_P = (_O = (_N = self.config).generateId) == null ? void 0 : _O.call(_N)) != null ? _P : (0, import_provider_utils36.generateId)(),
+                  id: (_R = (_Q = (_P = self.config).generateId) == null ? void 0 : _Q.call(_P)) != null ? _R : (0, import_provider_utils36.generateId)(),
                   mediaType: "application/octet-stream",
                   title: value.annotation.file_id,
                   filename: value.annotation.file_id,
@@ -7713,7 +7723,10 @@ var OpenAIResponsesLanguageModel = class {
             } else if (isErrorChunk(value)) {
               encounteredStreamError = true;
               finishReason = { unified: "error", raw: "error" };
-              controller.enqueue({ type: "error", error: value });
+              controller.enqueue({
+                type: "error",
+                error: (_S = createOpenAIProviderStreamError(value)) != null ? _S : value
+              });
             }
           },
           flush(controller) {