@mastra/memory 1.17.4 → 1.17.5-alpha.1

package/CHANGELOG.md

@@ -1,5 +1,23 @@
 # @mastra/memory
 
+## 1.17.5-alpha.1
+
+### Patch Changes
+
+- Fixed Observational Memory model resolution for user-defined gateways. Models such as `cloudflare/google/gemini-2.5-flash-lite` now resolve through registered gateways instead of failing with provider-config errors. Closes #13841. ([#16083](https://github.com/mastra-ai/mastra/pull/16083))
+
+- Updated dependencies [[`7679a63`](https://github.com/mastra-ai/mastra/commit/7679a634eae8e8ca459fd87538fdf72b4389b07f), [`1d64a76`](https://github.com/mastra-ai/mastra/commit/1d64a765861a0772ea187bab76e5ed37bf82d042), [`7679a63`](https://github.com/mastra-ai/mastra/commit/7679a634eae8e8ca459fd87538fdf72b4389b07f), [`a0d9b6d`](https://github.com/mastra-ai/mastra/commit/a0d9b6d6b810aeaa9e177a0dcc99a4402e609634)]:
+  - @mastra/core@1.32.0-alpha.4
+
+## 1.17.5-alpha.0
+
+### Patch Changes
+
+- Fixed async reflection buffering incorrectly triggering during idle timeout and provider-change activations when observation tokens are below the reflection activation threshold ([#16076](https://github.com/mastra-ai/mastra/pull/16076))
+
+- Updated dependencies [[`c05c9a1`](https://github.com/mastra-ai/mastra/commit/c05c9a13230988cef6d438a62f37760f31927bc7), [`e24aacb`](https://github.com/mastra-ai/mastra/commit/e24aacba07bd66f5d95b636dc24016fca26b52cf), [`c721164`](https://github.com/mastra-ai/mastra/commit/c7211643f7ac861f83b19a3757cc921487fc9d75), [`1b55954`](https://github.com/mastra-ai/mastra/commit/1b559541c1e08a10e49d01ffc51a634dfc37a286), [`5adc55e`](https://github.com/mastra-ai/mastra/commit/5adc55e63407be8ee977914957d68bcc2a075ceb), [`70017d7`](https://github.com/mastra-ai/mastra/commit/70017d72ab741b5d7040e2a15c251a317782e39e), [`e4942bc`](https://github.com/mastra-ai/mastra/commit/e4942bc7fdc903572f7d84f26d5e15f9d39c763d)]:
+  - @mastra/core@1.32.0-alpha.1
+
 ## 1.17.4
 
 ### Patch Changes

package/dist/{chunk-4C4ERX6N.js → chunk-BPJLUC2F.js}

@@ -3752,6 +3752,7 @@ var ObserverRunner = class {
   observedMessageIds;
   resolveModel;
   tokenCounter;
+  mastra;
   /** Captured prompt/response from the last observer call (for repro capture). */
   lastExchange;
   constructor(opts) {
@@ -3759,9 +3760,13 @@ var ObserverRunner = class {
     this.observedMessageIds = opts.observedMessageIds;
     this.resolveModel = opts.resolveModel;
     this.tokenCounter = opts.tokenCounter;
+    this.mastra = opts.mastra;
+  }
+  __registerMastra(mastra) {
+    this.mastra = mastra;
   }
   createAgent(model, isMultiThread = false) {
-    return new Agent({
+    const agent = new Agent({
       id: isMultiThread ? "multi-thread-observer" : "observational-memory-observer",
       name: isMultiThread ? "multi-thread-observer" : "Observer",
       instructions: buildObserverSystemPrompt(
@@ -3771,6 +3776,10 @@ var ObserverRunner = class {
       ),
       model
     });
+    if (this.mastra) {
+      agent.__registerMastra(this.mastra);
+    }
+    return agent;
   }
   async withAbortCheck(fn, abortSignal) {
     if (abortSignal?.aborted) {
@@ -4290,6 +4299,7 @@ var ReflectorRunner = class {
   persistMarkerToStorage;
   persistMarkerToMessage;
   getCompressionStartLevel;
+  mastra;
   constructor(opts) {
     this.reflectionConfig = opts.reflectionConfig;
     this.observationConfig = opts.observationConfig;
@@ -4302,14 +4312,22 @@ var ReflectorRunner = class {
     this.persistMarkerToStorage = opts.persistMarkerToStorage;
     this.persistMarkerToMessage = opts.persistMarkerToMessage;
     this.getCompressionStartLevel = opts.getCompressionStartLevel;
+    this.mastra = opts.mastra;
+  }
+  __registerMastra(mastra) {
+    this.mastra = mastra;
   }
   createAgent(model) {
-    return new Agent({
+    const agent = new Agent({
       id: "observational-memory-reflector",
       name: "Reflector",
       instructions: buildReflectorSystemPrompt(this.reflectionConfig.instruction),
       model
     });
+    if (this.mastra) {
+      agent.__registerMastra(this.mastra);
+    }
+    return agent;
   }
   getObservationMarkerConfig(record) {
     return {
@@ -4814,6 +4832,13 @@ ${unreflectedContent}` : freshRecord.bufferedReflection;
           `[OM:reflect] blockAfter exceeded (${observationTokens} >= ${this.reflectionConfig.blockAfter}), falling through to sync reflection`
         );
       } else {
+        const activationPoint = reflectThreshold * this.reflectionConfig.bufferActivation;
+        if (observationTokens < activationPoint) {
+          omDebug(
+            `[OM:reflect] skipping async reflection \u2014 observationTokens (${observationTokens}) below activation point (${activationPoint}), triggered by ${activationTriggeredBy}`
+          );
+          return;
+        }
         omDebug(
           `[OM:reflect] async activation failed, no blockAfter or below it (obsTokens=${observationTokens}, blockAfter=${this.reflectionConfig.blockAfter}) \u2014 starting background reflection`
         );
@@ -6588,6 +6613,7 @@ var ObservationalMemory = class _ObservationalMemory {
   buffering;
   shouldObscureThreadIds = false;
   hasher = xxhash();
+  mastra;
   /**
    * Track message IDs observed during this instance's lifetime.
    * Prevents re-observing messages when per-thread lastObservedAt cursors
@@ -6654,20 +6680,10 @@ var ObservationalMemory = class _ObservationalMemory {
     this.scope = config.scope ?? "thread";
     this.retrieval = Boolean(config.retrieval);
     this.onIndexObservations = config.onIndexObservations;
-    const resolveModel = (m) => m === "default" ? OBSERVATIONAL_MEMORY_DEFAULTS.observation.model : m;
-    const observationModel = resolveModel(config.model) ?? resolveModel(config.observation?.model) ?? resolveModel(config.reflection?.model);
-    const reflectionModel = resolveModel(config.model) ?? resolveModel(config.reflection?.model) ?? resolveModel(config.observation?.model);
-    if (!observationModel || !reflectionModel) {
-      throw new Error(
-        `Observational Memory requires a model to be set. Use \`observationalMemory: true\` for the default (google/gemini-2.5-flash), or set a model explicitly:
-
-  observationalMemory: {
-    model: "$provider/$model",
-  }
-
-See https://mastra.ai/docs/memory/observational-memory#models for model recommendations and alternatives.`
-      );
-    }
+    this.mastra = config.mastra;
+    const resolveModel = (model, defaultModel) => model === "default" ? defaultModel : model;
+    const observationModel = resolveModel(config.model, OBSERVATIONAL_MEMORY_DEFAULTS.observation.model) ?? resolveModel(config.observation?.model, OBSERVATIONAL_MEMORY_DEFAULTS.observation.model) ?? resolveModel(config.reflection?.model, OBSERVATIONAL_MEMORY_DEFAULTS.observation.model) ?? OBSERVATIONAL_MEMORY_DEFAULTS.observation.model;
+    const reflectionModel = resolveModel(config.model, OBSERVATIONAL_MEMORY_DEFAULTS.reflection.model) ?? resolveModel(config.reflection?.model, OBSERVATIONAL_MEMORY_DEFAULTS.reflection.model) ?? resolveModel(config.observation?.model, OBSERVATIONAL_MEMORY_DEFAULTS.reflection.model) ?? OBSERVATIONAL_MEMORY_DEFAULTS.reflection.model;
     const messageTokens = config.observation?.messageTokens ?? OBSERVATIONAL_MEMORY_DEFAULTS.observation.messageTokens;
     const observationTokens = config.reflection?.observationTokens ?? OBSERVATIONAL_MEMORY_DEFAULTS.reflection.observationTokens;
     const isSharedBudget = config.shareTokenBudget ?? false;
@@ -6753,7 +6769,8 @@ Async buffering is enabled by default \u2014 this opt-out is only needed when us
       observationConfig: this.observationConfig,
       observedMessageIds: this.observedMessageIds,
       resolveModel: (inputTokens) => this.resolveObservationModel(inputTokens),
-      tokenCounter: this.tokenCounter
+      tokenCounter: this.tokenCounter,
+      mastra: config.mastra
     });
     this.buffering = new BufferingCoordinator({
       observationConfig: this.observationConfig,
@@ -6771,13 +6788,19 @@ Async buffering is enabled by default \u2014 this opt-out is only needed when us
       persistMarkerToStorage: (m, t, r) => this.persistMarkerToStorage(m, t, r),
       persistMarkerToMessage: (m, ml, t, r) => this.persistMarkerToMessage(m, ml, t, r),
       getCompressionStartLevel: (rc) => this.getCompressionStartLevel(rc),
-      resolveModel: (inputTokens) => this.resolveReflectionModel(inputTokens)
+      resolveModel: (inputTokens) => this.resolveReflectionModel(inputTokens),
+      mastra: config.mastra
     });
     this.validateBufferConfig();
     omDebug(
       `[OM:init] new ObservationalMemory instance created \u2014 scope=${this.scope}, messageTokens=${JSON.stringify(this.observationConfig.messageTokens)}, obsAsyncEnabled=${this.buffering.isAsyncObservationEnabled()}, bufferTokens=${this.observationConfig.bufferTokens}, bufferActivation=${this.observationConfig.bufferActivation}, blockAfter=${this.observationConfig.blockAfter}, reflectionTokens=${this.reflectionConfig.observationTokens}, refAsyncEnabled=${this.buffering.isAsyncReflectionEnabled()}, refAsyncActivation=${this.reflectionConfig.bufferActivation}, refBlockAfter=${this.reflectionConfig.blockAfter}`
     );
   }
+  __registerMastra(mastra) {
+    this.mastra = mastra;
+    this.observer.__registerMastra(mastra);
+    this.reflector.__registerMastra(mastra);
+  }
   /**
    * Get the current configuration for this OM instance.
    * Used by the server to expose config to the UI when OM is added via processors.
@@ -6886,7 +6909,7 @@ Async buffering is enabled by default \u2014 this opt-out is only needed when us
     if (!modelToResolve) {
       return void 0;
     }
-    const resolved = await resolveModelConfig(modelToResolve, requestContext);
+    const resolved = await resolveModelConfig(modelToResolve, requestContext, this.mastra);
     return {
       provider: resolved.provider,
       modelId: resolved.modelId
@@ -9343,5 +9366,5 @@ function getObservationsAsOf(activeObservations, asOf) {
 }
 
 export { ModelByInputTokens, OBSERVER_SYSTEM_PROMPT, ObservationalMemory, ObservationalMemoryProcessor, TokenCounter, buildObserverPrompt, buildObserverSystemPrompt, combineObservationGroupRanges, deriveObservationGroupProvenance, extractCurrentTask, formatMessagesForObserver, formatToolResultForObserver, getObservationsAsOf, hasCurrentTaskSection, injectAnchorIds, optimizeObservationsForContext, parseAnchorId, parseObservationGroups, parseObserverOutput, reconcileObservationGroupsFromReflection, renderObservationGroupsForReflection, resolveToolResultValue, stripEphemeralAnchorIds, stripObservationGroups, truncateStringByTokens, wrapInObservationGroup };
-//# sourceMappingURL=chunk-4C4ERX6N.js.map
-//# sourceMappingURL=chunk-4C4ERX6N.js.map
+//# sourceMappingURL=chunk-BPJLUC2F.js.map
+//# sourceMappingURL=chunk-BPJLUC2F.js.map