@anuma/sdk 6.33.2 → 6.35.0

package/dist/telemetry/index.d.mts

@@ -523,10 +523,9 @@ interface MetricsHooksOptions {
     includeErrorMessages?: boolean;
 }
 /**
- * Build {@link RunHooks} that report run lifecycle, model calls, and tool
- * calls to `sink`. The returned hooks are synchronous and never throw — a
- * throwing sink method is swallowed with a console-free `noop` (the tool loop
- * already swallows hook errors; we avoid double-logging here).
+ * Build {@link RunHooks} that report run lifecycle, model calls, and tool calls to `sink`; every run ends in exactly one `run.completed` or `run.failed`, and no message contents or tool arguments are emitted.
+ *
+ * @see TELEMETRY.md for event payloads and metric names.
  */
 declare function createMetricsHooks(sink: TelemetrySink, opts?: MetricsHooksOptions): RunHooks;
 

package/dist/telemetry/index.d.ts

@@ -523,10 +523,9 @@ interface MetricsHooksOptions {
     includeErrorMessages?: boolean;
 }
 /**
- * Build {@link RunHooks} that report run lifecycle, model calls, and tool
- * calls to `sink`. The returned hooks are synchronous and never throw — a
- * throwing sink method is swallowed with a console-free `noop` (the tool loop
- * already swallows hook errors; we avoid double-logging here).
+ * Build {@link RunHooks} that report run lifecycle, model calls, and tool calls to `sink`; every run ends in exactly one `run.completed` or `run.failed`, and no message contents or tool arguments are emitted.
+ *
+ * @see TELEMETRY.md for event payloads and metric names.
  */
 declare function createMetricsHooks(sink: TelemetrySink, opts?: MetricsHooksOptions): RunHooks;
 

package/dist/tools/selection/index.cjs

@@ -1058,7 +1058,7 @@ async function embedOne(text, options, outer) {
 }
 var DEFAULT_EMBEDDING_BATCH_SIZE = 100;
 var DEFAULT_EMBEDDING_BATCH_CONCURRENCY = 3;
-async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage, maskInput, timeoutMs) {
+async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage, maskInput, timeoutMs, outer) {
   const response = await withEmbeddingRetry(
     () => withDeadline(
       (signal) => postApiV1Embeddings({
@@ -1068,8 +1068,10 @@ async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage,
         ...signal && { signal }
       }),
       timeoutMs,
-      "embedding batch request"
-    )
+      "embedding batch request",
+      outer
+    ),
+    outer
   );
   if (response.error) {
     throw embeddingErrorFrom(response);
@@ -1087,6 +1089,11 @@ async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage,
 }
 async function generateEmbeddings(texts, options) {
   if (texts.length === 0) return [];
+  const total = options.totalTimeoutMs;
+  if (total === void 0) return embedMany(texts, options, void 0);
+  return withDeadline((signal) => embedMany(texts, options, signal), total, "embedding batch");
+}
+async function embedMany(texts, options, outer) {
   const { baseUrl = BASE_URL, model, batchSize, cache } = options;
   const timeoutMs = options.timeoutMs ?? DEFAULT_EMBEDDING_REQUEST_TIMEOUT_MS;
   const chunkSize = batchSize ?? DEFAULT_EMBEDDING_BATCH_SIZE;
@@ -1107,7 +1114,8 @@ async function generateEmbeddings(texts, options) {
   if (uncachedTexts.length === 0) {
     return results;
   }
-  const headers = await resolveAuthHeaders(options);
+  const headers = await resolveAuthHeaders(options, outer);
+  if (outer?.aborted) throw outer.reason;
   const embeddingModel = model ?? DEFAULT_API_EMBEDDING_MODEL;
   let newEmbeddings;
   if (uncachedTexts.length <= chunkSize) {
@@ -1118,7 +1126,8 @@ async function generateEmbeddings(texts, options) {
       embeddingModel,
       options.onUsage,
       options.maskInput,
-      timeoutMs
+      timeoutMs,
+      outer
     );
   } else {
     const chunks = [];
@@ -1129,6 +1138,7 @@ async function generateEmbeddings(texts, options) {
     let nextIndex = 0;
     const worker = async () => {
       while (nextIndex < chunks.length) {
+        if (outer?.aborted) throw outer.reason;
         const idx = nextIndex++;
         allEmbeddings[idx] = await generateEmbeddingsBatch(
           chunks[idx],
@@ -1137,7 +1147,8 @@ async function generateEmbeddings(texts, options) {
           embeddingModel,
           options.onUsage,
           options.maskInput,
-          timeoutMs
+          timeoutMs,
+          outer
         );
       }
     };
@@ -1148,6 +1159,7 @@ async function generateEmbeddings(texts, options) {
     await Promise.all(workers);
     newEmbeddings = allEmbeddings.flat();
   }
+  if (outer?.aborted) throw outer.reason;
   for (let i = 0; i < uncachedIndices.length; i++) {
     const f32Embedding = Float32Array.from(newEmbeddings[i]);
     results[uncachedIndices[i]] = Array.from(f32Embedding);

package/dist/tools/selection/index.mjs

@@ -936,7 +936,7 @@ async function embedOne(text, options, outer) {
 }
 var DEFAULT_EMBEDDING_BATCH_SIZE = 100;
 var DEFAULT_EMBEDDING_BATCH_CONCURRENCY = 3;
-async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage, maskInput, timeoutMs) {
+async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage, maskInput, timeoutMs, outer) {
   const response = await withEmbeddingRetry(
     () => withDeadline(
       (signal) => postApiV1Embeddings({
@@ -946,8 +946,10 @@ async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage,
         ...signal && { signal }
       }),
       timeoutMs,
-      "embedding batch request"
-    )
+      "embedding batch request",
+      outer
+    ),
+    outer
   );
   if (response.error) {
     throw embeddingErrorFrom(response);
@@ -965,6 +967,11 @@ async function generateEmbeddingsBatch(texts, headers, baseUrl, model, onUsage,
 }
 async function generateEmbeddings(texts, options) {
   if (texts.length === 0) return [];
+  const total = options.totalTimeoutMs;
+  if (total === void 0) return embedMany(texts, options, void 0);
+  return withDeadline((signal) => embedMany(texts, options, signal), total, "embedding batch");
+}
+async function embedMany(texts, options, outer) {
   const { baseUrl = BASE_URL, model, batchSize, cache } = options;
   const timeoutMs = options.timeoutMs ?? DEFAULT_EMBEDDING_REQUEST_TIMEOUT_MS;
   const chunkSize = batchSize ?? DEFAULT_EMBEDDING_BATCH_SIZE;
@@ -985,7 +992,8 @@ async function generateEmbeddings(texts, options) {
   if (uncachedTexts.length === 0) {
     return results;
   }
-  const headers = await resolveAuthHeaders(options);
+  const headers = await resolveAuthHeaders(options, outer);
+  if (outer?.aborted) throw outer.reason;
   const embeddingModel = model ?? DEFAULT_API_EMBEDDING_MODEL;
   let newEmbeddings;
   if (uncachedTexts.length <= chunkSize) {
@@ -996,7 +1004,8 @@ async function generateEmbeddings(texts, options) {
       embeddingModel,
       options.onUsage,
       options.maskInput,
-      timeoutMs
+      timeoutMs,
+      outer
     );
   } else {
     const chunks = [];
@@ -1007,6 +1016,7 @@ async function generateEmbeddings(texts, options) {
     let nextIndex = 0;
     const worker = async () => {
       while (nextIndex < chunks.length) {
+        if (outer?.aborted) throw outer.reason;
         const idx = nextIndex++;
         allEmbeddings[idx] = await generateEmbeddingsBatch(
           chunks[idx],
@@ -1015,7 +1025,8 @@ async function generateEmbeddings(texts, options) {
           embeddingModel,
           options.onUsage,
           options.maskInput,
-          timeoutMs
+          timeoutMs,
+          outer
         );
       }
     };
@@ -1026,6 +1037,7 @@ async function generateEmbeddings(texts, options) {
     await Promise.all(workers);
     newEmbeddings = allEmbeddings.flat();
   }
+  if (outer?.aborted) throw outer.reason;
   for (let i = 0; i < uncachedIndices.length; i++) {
     const f32Embedding = Float32Array.from(newEmbeddings[i]);
     results[uncachedIndices[i]] = Array.from(f32Embedding);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@anuma/sdk",
-  "version": "6.33.2",
+  "version": "6.35.0",
   "description": "",
   "main": "./dist/index.cjs",
   "module": "./dist/index.mjs",