ai 7.0.22 → 7.0.25

package/CHANGELOG.md

@@ -1,5 +1,35 @@
 # ai
 
+## 7.0.25
+
+### Patch Changes
+
+- 7805e4a: Cancelling the `experimental_streamTranscribe` `fullStream` now also aborts a still-pending `doStream` setup, so a model whose `doStream` has not yet resolved is cancelled instead of leaking.
+- f8e82fd: Update the `experimental_streamTranscribe` unsupported-model error message now that gateway string model IDs can support streaming transcription.
+- Updated dependencies [4be62c1]
+- Updated dependencies [f8e82fd]
+- Updated dependencies [7805e4a]
+- Updated dependencies [cd12954]
+  - @ai-sdk/provider-utils@5.0.9
+  - @ai-sdk/gateway@4.0.19
+
+## 7.0.24
+
+### Patch Changes
+
+- e193290: Cancel the caller's `audio` stream when `experimental_streamTranscribe` fails before or during streaming. Previously, when the model's `doStream` rejected before a stream existed (e.g. missing API key or other auth failure), the audio stream was never consumed or cancelled, so an upstream producer piping into it would hang forever.
+- Updated dependencies [e193290]
+  - @ai-sdk/provider-utils@5.0.8
+  - @ai-sdk/gateway@4.0.18
+
+## 7.0.23
+
+### Patch Changes
+
+- 930f949: feat(ai): wrap embedMany in tracing channel context
+- Updated dependencies [867f80a]
+  - @ai-sdk/gateway@4.0.17
+
 ## 7.0.22
 
 ### Patch Changes

package/dist/index.d.ts

@@ -4274,7 +4274,7 @@ type RerankingModelCallEndEvent = {
 };
 
 declare const AI_SDK_TELEMETRY_TRACING_CHANNEL = "ai:telemetry";
-type TelemetryTracingEventType = 'generateText' | 'streamText' | 'step' | 'languageModelCall' | 'executeTool' | 'embed' | 'rerank';
+type TelemetryTracingEventType = 'generateText' | 'streamText' | 'step' | 'languageModelCall' | 'executeTool' | 'embed' | 'embedMany' | 'rerank';
 type TelemetryTracingChannelMessage<EVENT = unknown> = {
     readonly type: TelemetryTracingEventType;
     readonly event: EVENT;

package/dist/index.js

@@ -1089,7 +1089,7 @@ import {
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "7.0.22" : "0.0.0-test";
+var VERSION = true ? "7.0.25" : "0.0.0-test";
 
 // src/util/download/download.ts
 var download = async ({
@@ -11301,7 +11301,7 @@ async function embedMany({
   experimental_onEnd,
   _internal: { generateCallId = originalGenerateCallId5 } = {}
 }) {
-  var _a22, _b;
+  var _a22;
   const model = resolveEmbeddingModel(modelArg);
   const { maxRetries, retry } = prepareRetries({
     maxRetries: maxRetriesArg,
@@ -11317,112 +11317,33 @@ async function embedMany({
   const telemetryDispatcher = createTelemetryDispatcher({
     telemetry
   });
-  await notify({
-    event: {
-      callId,
-      operationId: "ai.embedMany",
-      provider: model.provider,
-      modelId: model.modelId,
-      value: values,
-      maxRetries,
-      headers: headersWithUserAgent,
-      providerOptions
-    },
-    callbacks: [resolvedOnStart, telemetryDispatcher.onStart]
-  });
-  try {
-    const [maxEmbeddingsPerCall, supportsParallelCalls] = await Promise.all([
-      model.maxEmbeddingsPerCall,
-      model.supportsParallelCalls
-    ]);
-    if (maxEmbeddingsPerCall == null || maxEmbeddingsPerCall === Infinity) {
-      const { embeddings: embeddings2, usage, warnings: warnings2, response, providerMetadata: providerMetadata2 } = await retry(async () => {
-        var _a23, _b2;
-        const embedCallId = generateCallId();
-        await notify({
-          event: {
-            callId,
-            embedCallId,
-            operationId: "ai.embedMany.doEmbed",
-            provider: model.provider,
-            modelId: model.modelId,
-            values
-          },
-          callbacks: [telemetryDispatcher.onEmbedStart]
-        });
-        const modelResponse = await model.doEmbed({
-          values,
-          abortSignal,
-          headers: headersWithUserAgent,
-          providerOptions
-        });
-        const embeddings3 = modelResponse.embeddings;
-        const usage2 = (_a23 = modelResponse.usage) != null ? _a23 : { tokens: NaN };
-        await notify({
-          event: {
-            callId,
-            embedCallId,
-            operationId: "ai.embedMany.doEmbed",
-            provider: model.provider,
-            modelId: model.modelId,
-            values,
-            embeddings: embeddings3,
-            usage: usage2
-          },
-          callbacks: [telemetryDispatcher.onEmbedEnd]
-        });
-        return {
-          embeddings: embeddings3,
-          usage: usage2,
-          warnings: (_b2 = modelResponse.warnings) != null ? _b2 : [],
-          providerMetadata: modelResponse.providerMetadata,
-          response: modelResponse.response
-        };
-      });
-      logWarnings({
-        warnings: warnings2,
-        provider: model.provider,
-        model: model.modelId
-      });
+  const runInTracingChannelSpan = (_a22 = telemetryDispatcher.runInTracingChannelSpan) != null ? _a22 : async ({ execute }) => await execute();
+  const startEvent = {
+    callId,
+    operationId: "ai.embedMany",
+    provider: model.provider,
+    modelId: model.modelId,
+    value: values,
+    maxRetries,
+    headers: headersWithUserAgent,
+    providerOptions
+  };
+  return await runInTracingChannelSpan({
+    type: "embedMany",
+    event: startEvent,
+    execute: async () => {
+      var _a23, _b;
       await notify({
-        event: {
-          callId,
-          operationId: "ai.embedMany",
-          provider: model.provider,
-          modelId: model.modelId,
-          value: values,
-          embedding: embeddings2,
-          usage,
-          warnings: warnings2,
-          providerMetadata: providerMetadata2,
-          response: [response]
-        },
-        callbacks: [resolvedOnEnd, telemetryDispatcher.onEnd]
-      });
-      return new DefaultEmbedManyResult({
-        values,
-        embeddings: embeddings2,
-        usage,
-        warnings: warnings2,
-        providerMetadata: providerMetadata2,
-        responses: [response]
+        event: startEvent,
+        callbacks: [resolvedOnStart, telemetryDispatcher.onStart]
       });
-    }
-    const valueChunks = splitArray(values, maxEmbeddingsPerCall);
-    const embeddings = [];
-    const warnings = [];
-    const responses = [];
-    let tokens = 0;
-    let providerMetadata;
-    const parallelChunks = splitArray(
-      valueChunks,
-      supportsParallelCalls ? maxParallelCalls : 1
-    );
-    for (const parallelChunk of parallelChunks) {
-      const results = await Promise.all(
-        parallelChunk.map((chunk) => {
-          return retry(async () => {
-            var _a23, _b2;
+      try {
+        const [maxEmbeddingsPerCall, supportsParallelCalls] = await Promise.all(
+          [model.maxEmbeddingsPerCall, model.supportsParallelCalls]
+        );
+        if (maxEmbeddingsPerCall == null || maxEmbeddingsPerCall === Infinity) {
+          const { embeddings: embeddings2, usage, warnings: warnings2, response, providerMetadata: providerMetadata2 } = await retry(async () => {
+            var _a24, _b2;
             const embedCallId = generateCallId();
             await notify({
               event: {
@@ -11431,18 +11352,18 @@ async function embedMany({
                 operationId: "ai.embedMany.doEmbed",
                 provider: model.provider,
                 modelId: model.modelId,
-                values: chunk
+                values
               },
               callbacks: [telemetryDispatcher.onEmbedStart]
             });
             const modelResponse = await model.doEmbed({
-              values: chunk,
+              values,
               abortSignal,
               headers: headersWithUserAgent,
               providerOptions
             });
-            const chunkEmbeddings = modelResponse.embeddings;
-            const usage = (_a23 = modelResponse.usage) != null ? _a23 : { tokens: NaN };
+            const embeddings3 = modelResponse.embeddings;
+            const usage2 = (_a24 = modelResponse.usage) != null ? _a24 : { tokens: NaN };
             await notify({
               event: {
                 callId,
@@ -11450,75 +11371,162 @@ async function embedMany({
                 operationId: "ai.embedMany.doEmbed",
                 provider: model.provider,
                 modelId: model.modelId,
-                values: chunk,
-                embeddings: chunkEmbeddings,
-                usage
+                values,
+                embeddings: embeddings3,
+                usage: usage2
               },
               callbacks: [telemetryDispatcher.onEmbedEnd]
             });
             return {
-              embeddings: chunkEmbeddings,
-              usage,
+              embeddings: embeddings3,
+              usage: usage2,
               warnings: (_b2 = modelResponse.warnings) != null ? _b2 : [],
               providerMetadata: modelResponse.providerMetadata,
               response: modelResponse.response
             };
           });
-        })
-      );
-      for (const result of results) {
-        embeddings.push(...result.embeddings);
-        warnings.push(...result.warnings);
-        responses.push(result.response);
-        tokens += result.usage.tokens;
-        if (result.providerMetadata) {
-          if (!providerMetadata) {
-            providerMetadata = { ...result.providerMetadata };
-          } else {
-            for (const [providerName, metadata] of Object.entries(
-              result.providerMetadata
-            )) {
-              providerMetadata[providerName] = {
-                ...(_a22 = providerMetadata[providerName]) != null ? _a22 : {},
-                ...metadata
-              };
+          logWarnings({
+            warnings: warnings2,
+            provider: model.provider,
+            model: model.modelId
+          });
+          await notify({
+            event: {
+              callId,
+              operationId: "ai.embedMany",
+              provider: model.provider,
+              modelId: model.modelId,
+              value: values,
+              embedding: embeddings2,
+              usage,
+              warnings: warnings2,
+              providerMetadata: providerMetadata2,
+              response: [response]
+            },
+            callbacks: [resolvedOnEnd, telemetryDispatcher.onEnd]
+          });
+          return new DefaultEmbedManyResult({
+            values,
+            embeddings: embeddings2,
+            usage,
+            warnings: warnings2,
+            providerMetadata: providerMetadata2,
+            responses: [response]
+          });
+        }
+        const valueChunks = splitArray(values, maxEmbeddingsPerCall);
+        const embeddings = [];
+        const warnings = [];
+        const responses = [];
+        let tokens = 0;
+        let providerMetadata;
+        const parallelChunks = splitArray(
+          valueChunks,
+          supportsParallelCalls ? maxParallelCalls : 1
+        );
+        for (const parallelChunk of parallelChunks) {
+          const results = await Promise.all(
+            parallelChunk.map((chunk) => {
+              return retry(async () => {
+                var _a24, _b2;
+                const embedCallId = generateCallId();
+                await notify({
+                  event: {
+                    callId,
+                    embedCallId,
+                    operationId: "ai.embedMany.doEmbed",
+                    provider: model.provider,
+                    modelId: model.modelId,
+                    values: chunk
+                  },
+                  callbacks: [telemetryDispatcher.onEmbedStart]
+                });
+                const modelResponse = await model.doEmbed({
+                  values: chunk,
+                  abortSignal,
+                  headers: headersWithUserAgent,
+                  providerOptions
+                });
+                const chunkEmbeddings = modelResponse.embeddings;
+                const usage = (_a24 = modelResponse.usage) != null ? _a24 : { tokens: NaN };
+                await notify({
+                  event: {
+                    callId,
+                    embedCallId,
+                    operationId: "ai.embedMany.doEmbed",
+                    provider: model.provider,
+                    modelId: model.modelId,
+                    values: chunk,
+                    embeddings: chunkEmbeddings,
+                    usage
+                  },
+                  callbacks: [telemetryDispatcher.onEmbedEnd]
+                });
+                return {
+                  embeddings: chunkEmbeddings,
+                  usage,
+                  warnings: (_b2 = modelResponse.warnings) != null ? _b2 : [],
+                  providerMetadata: modelResponse.providerMetadata,
+                  response: modelResponse.response
+                };
+              });
+            })
+          );
+          for (const result of results) {
+            embeddings.push(...result.embeddings);
+            warnings.push(...result.warnings);
+            responses.push(result.response);
+            tokens += result.usage.tokens;
+            if (result.providerMetadata) {
+              if (!providerMetadata) {
+                providerMetadata = { ...result.providerMetadata };
+              } else {
+                for (const [providerName, metadata] of Object.entries(
+                  result.providerMetadata
+                )) {
+                  providerMetadata[providerName] = {
+                    ...(_a23 = providerMetadata[providerName]) != null ? _a23 : {},
+                    ...metadata
+                  };
+                }
+              }
             }
           }
         }
+        logWarnings({
+          warnings,
+          provider: model.provider,
+          model: model.modelId
+        });
+        await notify({
+          event: {
+            callId,
+            operationId: "ai.embedMany",
+            provider: model.provider,
+            modelId: model.modelId,
+            value: values,
+            embedding: embeddings,
+            usage: { tokens },
+            warnings,
+            providerMetadata,
+            response: responses
+          },
+          callbacks: [resolvedOnEnd, telemetryDispatcher.onEnd]
+        });
+        return new DefaultEmbedManyResult({
+          values,
+          embeddings,
+          usage: { tokens },
+          warnings,
+          providerMetadata,
+          responses
+        });
+      } catch (error) {
+        await ((_b = telemetryDispatcher.onError) == null ? void 0 : _b.call(telemetryDispatcher, { callId, error }));
+        throw error;
       }
     }
-    logWarnings({
-      warnings,
-      provider: model.provider,
-      model: model.modelId
-    });
-    await notify({
-      event: {
-        callId,
-        operationId: "ai.embedMany",
-        provider: model.provider,
-        modelId: model.modelId,
-        value: values,
-        embedding: embeddings,
-        usage: { tokens },
-        warnings,
-        providerMetadata,
-        response: responses
-      },
-      callbacks: [resolvedOnEnd, telemetryDispatcher.onEnd]
-    });
-    return new DefaultEmbedManyResult({
-      values,
-      embeddings,
-      usage: { tokens },
-      warnings,
-      providerMetadata,
-      responses
-    });
-  } catch (error) {
-    await ((_b = telemetryDispatcher.onError) == null ? void 0 : _b.call(telemetryDispatcher, { callId, error }));
-    throw error;
-  }
+  });
 }
 var DefaultEmbedManyResult = class {
   constructor(options) {
@@ -15953,7 +15961,7 @@ function streamTranscribe({
   if (doStream == null) {
     throw new UnsupportedFunctionalityError4({
       functionality: "streaming transcription",
-      message: `The ${resolvedModel.provider} model "${resolvedModel.modelId}" does not support streaming transcription.` + (typeof model === "string" ? " String model IDs resolve through the global provider (AI Gateway by default), which does not support streaming transcription yet. Pass a provider model instance instead, e.g. openai.transcription('gpt-realtime-whisper')." : "")
+      message: `The ${resolvedModel.provider} model "${resolvedModel.modelId}" does not support streaming transcription.` + (typeof model === "string" ? " String model IDs resolve through the global provider (AI Gateway by default). If that provider does not support streaming transcription, pass a provider model instance instead (e.g. openai.transcription('gpt-realtime-whisper')) or upgrade @ai-sdk/gateway to a version with streaming transcription support." : "")
     });
   }
   const headersWithUserAgent = withUserAgentSuffix11(
@@ -16057,7 +16065,8 @@ function streamTranscribe({
       audio,
       inputAudioFormat,
       providerOptions,
-      abortSignal,
+      // merged so cancelling fullStream also aborts a still-pending doStream
+      abortSignal: mergeAbortSignals(abortSignal, pipeAbortController.signal),
       headers: headersWithUserAgent,
       includeRawChunks
     });
@@ -16072,6 +16081,8 @@ function streamTranscribe({
   })().catch((error) => {
     const reason = error != null ? error : new Error("Transcription stream was cancelled or errored.");
     rejectPendingPromises(reason);
+    audio.cancel(reason).catch(() => {
+    });
     transform.writable.abort(reason).catch(() => {
     });
   });