@remnic/core 9.30.1 → 9.31.0

package/dist/{chunk-QJAQNXWI.js → chunk-ONHRRJZM.js}

@@ -1,6 +1,11 @@
 import {
   resolveEnsureCollectionArgs
 } from "./chunk-YV32N67O.js";
+import {
+  EMPTY_QMD_STATUS,
+  embedQmdFiles,
+  fetchQmdStatus
+} from "./chunk-IUM3VSHI.js";
 import {
   QMD_PROBE_RETRY_BACKOFF_MS,
   QMD_PROBE_TIMEOUT_MS,
@@ -13,11 +18,6 @@ import {
   qmdVersionToString,
   resolveQmdCapabilities
 } from "./chunk-64UDOIMI.js";
-import {
-  EMPTY_QMD_STATUS,
-  embedQmdFiles,
-  fetchQmdStatus
-} from "./chunk-IUM3VSHI.js";
 import {
   launchProcess
 } from "./chunk-O75CRYGF.js";
@@ -2228,4 +2228,4 @@ export {
   sharedDaemonSessionCountForTest,
   QmdClient
 };
-//# sourceMappingURL=chunk-QJAQNXWI.js.map
+//# sourceMappingURL=chunk-ONHRRJZM.js.map

package/dist/{chunk-FDAMS4PK.js → chunk-PNPOWBGC.js}

@@ -1,6 +1,6 @@
 import {
   runPostConsolidationMaterialize
-} from "./chunk-NHWB6HG2.js";
+} from "./chunk-NYCGIQRA.js";
 import {
   discoverMemoryExtensions,
   renderExtensionsBlock,
@@ -220,4 +220,4 @@ export {
   buildExtensionsBlockForConsolidation,
   materializeAfterSemanticConsolidation
 };
-//# sourceMappingURL=chunk-FDAMS4PK.js.map
+//# sourceMappingURL=chunk-PNPOWBGC.js.map

package/dist/{chunk-IC6CBHCA.js → chunk-QGENAYVS.js}

@@ -1,6 +1,6 @@
 import {
   StorageManager
-} from "./chunk-RGIWINDN.js";
+} from "./chunk-7G5MH4NT.js";
 import {
   namespaceIdentityToken,
   normalizeNamespaceIdentity
@@ -388,4 +388,4 @@ export {
   resolveNamespaceStorageRoot,
   NamespaceStorageRouter
 };
-//# sourceMappingURL=chunk-IC6CBHCA.js.map
+//# sourceMappingURL=chunk-QGENAYVS.js.map

package/dist/{chunk-GBJ4RDFR.js → chunk-QIHIHZYC.js}

@@ -21,22 +21,23 @@ import {
 } from "./chunk-WLZYGLJ4.js";
 import {
   LocalLlmClient
-} from "./chunk-PHWUFPCO.js";
+} from "./chunk-ZCPP5OW6.js";
 import {
-  applyExtractionSourceGrounding
-} from "./chunk-RPW2NOU6.js";
+  applyExtractionSourceGrounding,
+  attachExtractionProvenance
+} from "./chunk-YIAA2X3D.js";
 import {
   classifyExtractionThrownError,
   classifyFallbackParseFailure
 } from "./chunk-BLDBT6HZ.js";
-import {
-  delinearize
-} from "./chunk-VEWZZM3H.js";
 import {
   buildExtensionsFooterForSummary,
   formatDaySummaryMemories,
   loadDaySummaryPrompt
 } from "./chunk-RPSYDM6C.js";
+import {
+  delinearize
+} from "./chunk-VEWZZM3H.js";
 import {
   applyWorkExtractionBoundary
 } from "./chunk-EI6V5UXY.js";
@@ -61,9 +62,6 @@ import {
 import {
   extractJsonCandidates
 } from "./chunk-RGMVMVMF.js";
-import {
-  buildFactProvenance
-} from "./chunk-5I4M26Q7.js";
 import {
   sanitizeMemoryContent
 } from "./chunk-5GKPFL3S.js";
@@ -183,6 +181,11 @@ function normalizeProfileUpdateKey(update) {
 function shouldEnableLocalExtractionThinking(config, conversationChars) {
   return config.localLlmDisableThinking && config.localLlmThinkingThresholdChars > 0 && conversationChars < config.localLlmThinkingThresholdChars;
 }
+function throwIfExtractionAborted(signal) {
+  if (!signal?.aborted) return;
+  const reason = signal.reason;
+  throw reason instanceof Error ? reason : new Error("extraction aborted");
+}
 var ExtractionEngine = class {
   constructor(config, profilerArg, localLlm, gatewayConfig, modelRegistry) {
     this.config = config;
@@ -474,18 +477,19 @@ var ExtractionEngine = class {
     const entityUpdates = (Array.isArray(result.entityUpdates) ? result.entityUpdates : []).map((entity) => this.normalizeEntityUpdate(entity)).filter((entity) => entity !== void 0 && entity.name.length > 0);
     return { items, profileUpdates, entityUpdates };
   }
-  async applyProactiveQuestionPass(conversation, base, groundingSource, assertionSource, messageTimestamp, roleAssertionSources) {
+  async applyProactiveQuestionPass(conversation, base, groundingSource, assertionSource, messageTimestamp, roleAssertionSources, signal) {
     if (!resolvePipelineProcessingCapabilities(this.config).proactiveExtraction) return base;
     const maxAdditional = Math.max(0, Math.floor(this.config.maxProactiveQuestionsPerExtraction));
     if (!shouldRunProactivePass(this.config, maxAdditional, this.shouldUseLocalLlm, this.localLlm)) return base;
     try {
-      const proactive = await this.generateProactiveQuestions(conversation, base, maxAdditional);
+      const proactive = await this.generateProactiveQuestions(conversation, base, maxAdditional, signal);
       if (proactive.length === 0) return base;
       const proactiveAdditions = await this.answerProactiveQuestions(
         conversation,
         base,
         proactive,
-        maxAdditional
+        maxAdditional,
+        signal
       );
       const sanitizedAdditions = this.sanitizeExtractionResult(proactiveAdditions, messageTimestamp);
       const groundedAdditions = this.applySourceGrounding(
@@ -498,6 +502,7 @@ var ExtractionEngine = class {
       if (!this.hasExtractionOutputs(groundedAdditions)) return base;
       return this.mergeProactiveExtractionPass(base, groundedAdditions, maxAdditional);
     } catch (err) {
+      if (signal?.aborted) throw err;
       log.debug(`proactive extraction question pass failed (ignored): ${err}`);
       return base;
     }
@@ -526,7 +531,7 @@ var ExtractionEngine = class {
     }
     return null;
   }
-  async generateProactiveQuestions(conversation, base, maxAdditional) {
+  async generateProactiveQuestions(conversation, base, maxAdditional, signal) {
     const existingQuestionKeys = new Set(
       (base.questions ?? []).map((q) => q.question.trim().toLowerCase()).filter((q) => q.length > 0)
     );
@@ -562,9 +567,11 @@ var ExtractionEngine = class {
             maxTokens: this.config.proactiveExtractionMaxTokens,
             timeoutMs: this.config.proactiveExtractionTimeoutMs,
             operation: "proactive_extraction",
-            priority: "background"
+            priority: "background",
+            signal
           }
         );
+        throwIfExtractionAborted(signal);
         if (localResponse?.content) {
           const localParsed = this.parseProactiveQuestionsFromText(
             localResponse.content.trim(),
@@ -578,7 +585,7 @@ var ExtractionEngine = class {
           return [];
         }
       } catch (err) {
-        if (!this.config.localLlmFallback) {
+        if (signal?.aborted || !this.config.localLlmFallback) {
           throw err;
         }
       }
@@ -595,13 +602,14 @@ var ExtractionEngine = class {
       this.withGatewayAgent({
         temperature: 0.2,
         maxTokens: this.config.proactiveExtractionMaxTokens,
-        timeoutMs: this.config.proactiveExtractionTimeoutMs
+        timeoutMs: this.config.proactiveExtractionTimeoutMs,
+        signal
       })
     );
     if (!fallbackResult?.questions) return [];
     return fallbackResult.questions.map((q) => normalizeQuestion(q)).filter((q) => q.question.length > 0).filter((q) => !existingQuestionKeys.has(q.question.toLowerCase())).slice(0, maxAdditional);
   }
-  async answerProactiveQuestions(conversation, base, proactiveQuestions, maxAdditional) {
+  async answerProactiveQuestions(conversation, base, proactiveQuestions, maxAdditional, signal) {
     const factsPreview = base.facts.slice(0, 8).map((f) => `- (${f.category}) ${f.content}`).join("\n");
     const entitiesPreview = base.entities.slice(0, 8).map((entity) => `- (${entity.type}) ${entity.name}: ${entity.facts.join("; ") || "(no facts)"}`).join("\n");
     const proactivePreview = proactiveQuestions.slice(0, maxAdditional).map((question, index) => `${index + 1}. ${question.question}${question.context ? `
@@ -646,9 +654,11 @@ var ExtractionEngine = class {
             maxTokens: this.config.proactiveExtractionMaxTokens,
             timeoutMs: this.config.proactiveExtractionTimeoutMs,
             operation: "proactive_extraction",
-            priority: "background"
+            priority: "background",
+            signal
           }
         );
+        throwIfExtractionAborted(signal);
         if (localResponse?.content) {
           const parsed = this.parseProactiveExtractionResultFromText(localResponse.content.trim());
           if (parsed) {
@@ -659,7 +669,7 @@ var ExtractionEngine = class {
           return { facts: [], profileUpdates: [], entities: [], questions: [] };
         }
       } catch (err) {
-        if (!this.config.localLlmFallback) {
+        if (signal?.aborted || !this.config.localLlmFallback) {
           throw err;
         }
       }
@@ -676,7 +686,8 @@ var ExtractionEngine = class {
       this.withGatewayAgent({
         temperature: 0.2,
         maxTokens: this.config.proactiveExtractionMaxTokens,
-        timeoutMs: this.config.proactiveExtractionTimeoutMs
+        timeoutMs: this.config.proactiveExtractionTimeoutMs,
+        signal
       })
     );
     if (!fallbackResult) {
@@ -818,54 +829,6 @@ var ExtractionEngine = class {
     }
     return null;
   }
-  /**
-   * Attach claim-level provenance spans to each fact in the extraction result
-   * (issue #1575 PR 2). Runs once at write time, after sanitize + proactive
-   * pass so ALL facts (base + proactive additions) get verified spans before
-   * the result is returned for persistence.
-   *
-   * The validator locates each fact's LLM-provided `quote` in the buffered
-   * turns and builds a `ProvenanceSource[]` with verified offsets. Never
-   * throws, never drops a fact — an unverifiable span is a tagged state, not
-   * a silent failure (rule 34 spirit). When `provenance.enabled` is false,
-   * this is a no-op (byte-identical to pre-feature behavior, rule 39).
-   */
-  attachProvenanceToResult(result, turns) {
-    if (!this.config.provenance?.enabled) {
-      if (result.facts.length === 0) return result;
-      return {
-        ...result,
-        facts: result.facts.map((fact) => {
-          if (fact.quote === void 0) return fact;
-          const { quote: _stripped, ...rest } = fact;
-          return rest;
-        })
-      };
-    }
-    if (result.facts.length === 0) return result;
-    const provenanceTurns = turns.map((t) => ({
-      content: t.content,
-      sessionKey: t.sessionKey,
-      logicalSessionKey: t.logicalSessionKey,
-      timestamp: t.timestamp,
-      turnId: t.turnFingerprint
-    }));
-    const facts = result.facts.map((fact) => {
-      const built = buildFactProvenance(
-        fact.quote,
-        provenanceTurns,
-        this.config.provenance
-      );
-      const { quote: _stripped, ...factWithoutQuote } = fact;
-      return {
-        ...factWithoutQuote,
-        ...built.sources && built.sources.length > 0 ? { sources: built.sources } : {},
-        ...built.provenance !== "none" ? { provenance: built.provenance } : {},
-        ...built.requireSpansPending ? { requireSpansPending: true } : {}
-      };
-    });
-    return { ...result, facts };
-  }
   applySourceGrounding(result, sourceText, assertionSourceText = sourceText, roleAssertionSources, messageTimestamp) {
     const capabilities = resolvePipelineProcessingCapabilities(this.config);
     return applyExtractionSourceGrounding(
@@ -881,9 +844,9 @@ var ExtractionEngine = class {
     );
   }
   finalizeExtractionResult(result, turns) {
-    return this.attachProvenanceToResult(result, turns);
+    return attachExtractionProvenance(result, turns, this.config.provenance);
   }
-  async extract(turns, existingEntities) {
+  async extract(turns, existingEntities, signal) {
     const lifecycleCaps = resolveMemoryLifecycleCapabilities(this.config);
     const substantiveTurns = turns.filter((t) => t.content.trim().length > 0);
     if (substantiveTurns.length === 0) {
@@ -937,7 +900,8 @@ var ExtractionEngine = class {
         this.profiler.startSpan("local-llm", extractionTraceId);
         primaryExtractorAttempted = true;
         try {
-          const localResult = await this.extractWithLocalLlm(conversation, existingEntities);
+          const localResult = await this.extractWithLocalLlm(conversation, existingEntities, signal);
+          throwIfExtractionAborted(signal);
           if (localResult) {
             const durationMs = Date.now() - startTime;
             this.profiler.endSpan("local-llm", extractionTraceId);
@@ -957,7 +921,8 @@ var ExtractionEngine = class {
               groundingSource,
               assertionSource,
               messageTimestamp,
-              roleAssertionSources
+              roleAssertionSources,
+              signal
             );
             return this.finalizeExtractionResult(finalResult, boundedTurns);
           }
@@ -974,6 +939,7 @@ var ExtractionEngine = class {
           }
           log.info("extraction: local LLM unavailable, falling back to gateway default AI");
         } catch (err) {
+          if (signal?.aborted) throw err;
           if (!this.config.localLlmFallback) {
             log.warn("extraction: local LLM error and fallback disabled:", err);
             return {
@@ -997,7 +963,7 @@ var ExtractionEngine = class {
         this.profiler.startSpan("direct-client", extractionTraceId);
         primaryExtractorAttempted = true;
         try {
-          const directResult = await this.extractWithDirectClient(conversation, existingEntities);
+          const directResult = await this.extractWithDirectClient(conversation, existingEntities, signal);
           if (directResult) {
             const durationMs = Date.now() - startTime;
             this.profiler.endSpan("direct-client", extractionTraceId);
@@ -1017,7 +983,8 @@ var ExtractionEngine = class {
               groundingSource,
               assertionSource,
               messageTimestamp,
-              roleAssertionSources
+              roleAssertionSources,
+              signal
             );
             return this.finalizeExtractionResult(finalResult, boundedTurns);
           }
@@ -1035,6 +1002,7 @@ var ExtractionEngine = class {
           closedDirectTrace = true;
           log.info("extraction: direct client returned no result, falling back to gateway AI");
         } catch (err) {
+          if (signal?.aborted) throw err;
           try {
             this.emit({
               kind: "llm_error",
@@ -1090,7 +1058,8 @@ var ExtractionEngine = class {
           this.withGatewayAgent({
             temperature: 0.3,
             maxTokens: this.config.extractionMaxOutputTokens,
-            timeoutMs: this.config.localLlmTimeoutMs
+            timeoutMs: this.config.localLlmTimeoutMs,
+            signal
           })
         );
         const fallbackDurationMs = Date.now() - fallbackStartTime;
@@ -1122,7 +1091,8 @@ var ExtractionEngine = class {
             groundingSource,
             assertionSource,
             messageTimestamp,
-            roleAssertionSources
+            roleAssertionSources,
+            signal
           );
           return this.finalizeExtractionResult(finalResult, boundedTurns);
         }
@@ -1149,6 +1119,7 @@ var ExtractionEngine = class {
           extractionFailureClass: fallbackParseFailureClass
         };
       } catch (err) {
+        if (signal?.aborted) throw err;
         this.emit({
           kind: "llm_error",
           traceId: fallbackTraceId,
@@ -1181,7 +1152,7 @@ var ExtractionEngine = class {
    * Extract memories using local LLM with JSON mode.
    * Uses a minimal prompt to fit within local model context limits (typically 4k-8k).
    */
-  async extractWithLocalLlm(conversation, existingEntities) {
+  async extractWithLocalLlm(conversation, existingEntities, signal) {
     const lifecycleCaps = resolveMemoryLifecycleCapabilities(this.config);
     log.debug(
       `extractWithLocalLlm: starting extraction, localLlmEnabled=${this.shouldUseLocalLlm}, model=${this.config.localLlmModel}`
@@ -1241,7 +1212,8 @@ ${truncatedConversation}`;
         maxTokens: contextSizes.maxOutputTokens,
         operation: "extraction",
         disableThinking: shouldEnableLocalExtractionThinking(this.config, conversation.length) ? false : void 0,
-        priority: "background"
+        priority: "background",
+        signal
       }
     );
     if (!response?.content) {
@@ -1279,26 +1251,29 @@ ${truncatedConversation}`;
    * Extract memories using direct OpenAI-compatible client (Chat Completions API).
    * Works with Scryr, OpenRouter, and other OpenAI-compatible endpoints.
    */
-  async extractWithDirectClient(conversation, existingEntities) {
+  async extractWithDirectClient(conversation, existingEntities, signal) {
     if (!this.client) return null;
     const tokenParams = buildChatCompletionTokenLimit(this.config.model, this.config.extractionMaxOutputTokens, {
       assumeOpenAI: this.directClientUsesOpenAiTokenSemantics()
     });
     log.debug(`extractWithDirectClient: calling model=${this.config.model} tokenParams=${JSON.stringify(tokenParams)}`);
-    const response = await this.client.chat.completions.create({
-      model: this.config.model,
-      messages: [
-        {
-          role: "system",
-          content: this.buildExtractionInstructions(existingEntities) + `
+    const response = await this.client.chat.completions.create(
+      {
+        model: this.config.model,
+        messages: [
+          {
+            role: "system",
+            content: this.buildExtractionInstructions(existingEntities) + `
 
 Return only valid JSON matching this shape. Placeholder text describes field shape only and is never source evidence:
 ${EXTRACTION_RESPONSE_SHAPE}`
-        },
-        { role: "user", content: conversation }
-      ],
-      ...tokenParams
-    });
+          },
+          { role: "user", content: conversation }
+        ],
+        ...tokenParams
+      },
+      signal ? { signal } : void 0
+    );
     const content = response.choices?.[0]?.message?.content?.trim();
     if (!content) {
       log.info(`extractWithDirectClient: empty response \u2014 choices=${JSON.stringify(response.choices?.length ?? 0)} finishReason=${response.choices?.[0]?.finish_reason ?? "n/a"}`);
@@ -2507,4 +2482,4 @@ export {
   shouldEnableLocalExtractionThinking,
   ExtractionEngine
 };
-//# sourceMappingURL=chunk-GBJ4RDFR.js.map
+//# sourceMappingURL=chunk-QIHIHZYC.js.map