@remnic/core 9.50.0 → 9.50.2

package/dist/{chunk-3LD4N2TF.js → chunk-S3AYIQCU.js}

@@ -6,7 +6,7 @@ import {
   headerConnector,
   renderExtractionConversation,
   resolveSourceConnector
-} from "./chunk-ZDAYWJV6.js";
+} from "./chunk-NO37MR45.js";
 import {
   ConsolidationResultSchema,
   DaySummaryResultSchema,
@@ -93,6 +93,9 @@ import {
   resolvePipelineProcessingCapabilities,
   resolveRecallAuxiliaryCapabilities
 } from "./chunk-2XRE6FLM.js";
+import {
+  isAbortError
+} from "./chunk-BTTPKDIL.js";
 import {
   log
 } from "./chunk-3WCASNQZ.js";
@@ -146,61 +149,26 @@ function normalizeVerdicts(parsed, dependents) {
     return verdict ?? { memoryId: memory.id, verdict: "uncertain" };
   });
 }
-async function revalidateDependentsViaLlm(deps, superseded, replacement, dependents, signal) {
+async function revalidateDependentsViaLlm(deps, superseded, replacement, dependents, options = {}) {
   const { input, instructionText } = buildPrompt(superseded, replacement, dependents);
   const maxTokens = Math.max(1024, dependents.length * 256);
-  let rawContent = null;
-  if (deps.shouldUseLocalLlm) {
-    try {
-      const response = await deps.localLlm.chatCompletion(
-        [
-          { role: "system", content: instructionText },
-          { role: "user", content: input }
-        ],
-        {
-          temperature: 0.2,
-          maxTokens,
-          operation: "dependency_revalidation",
-          priority: "background",
-          signal
-        }
-      );
-      rawContent = response?.content ?? null;
-      if (!rawContent && !deps.config.localLlmFallback) {
-        throw new Error("local LLM returned no dependency revalidation output");
-      }
-    } catch (error) {
-      if (!deps.config.localLlmFallback || signal?.aborted) throw error;
-    }
-  }
-  if (rawContent === null && !deps.shouldUseDirectClient) {
-    const response = await deps.fallbackLlm.chatCompletion(
-      [
-        { role: "system", content: instructionText },
-        { role: "user", content: input }
-      ],
-      deps.withGatewayAgent({ temperature: 0.2, maxTokens, signal })
-    );
-    rawContent = response?.content ?? null;
-    if (!rawContent) throw new Error("fallback LLM returned no dependency revalidation output");
-  }
-  if (rawContent === null && deps.shouldUseDirectClient) {
-    const chatClient = deps.client;
-    const response = await chatClient.chat.completions.create(
-      {
-        model: deps.config.model,
-        messages: [
-          { role: "system", content: instructionText },
-          { role: "user", content: input }
-        ],
-        max_tokens: maxTokens
-      },
-      { signal }
-    );
-    rawContent = response.choices?.[0]?.message?.content ?? null;
-    if (!rawContent?.trim()) {
-      throw new Error("direct client returned no dependency revalidation output");
+  const response = await deps.fastChatCompletion(
+    [
+      { role: "system", content: instructionText },
+      { role: "user", content: input }
+    ],
+    {
+      temperature: 0.2,
+      maxTokens,
+      operation: "dependency_revalidation",
+      priority: "background",
+      timeoutMs: options.timeoutMs,
+      signal: options.signal
     }
+  );
+  const rawContent = response?.content ?? null;
+  if (!rawContent?.trim()) {
+    throw new Error("fast completion returned no dependency revalidation output");
   }
   return { verdicts: normalizeVerdicts(deps.parseJsonObject(rawContent), dependents) };
 }
@@ -256,8 +224,9 @@ function throwIfExtractionAborted(signal) {
   throw reason instanceof Error ? reason : new Error("extraction aborted");
 }
 var ExtractionEngine = class {
-  constructor(config, profilerArg, localLlm, gatewayConfig, modelRegistry) {
+  constructor(config, profilerArg, localLlm, gatewayConfig, modelRegistry, fastChatCompletion) {
     this.config = config;
+    this.fastChatCompletion = fastChatCompletion;
     this.profiler = profilerArg ?? new ProfilingCollector({ enabled: false, storageDir: "/tmp/engram-profiler-disabled", maxTraces: 0 });
     if (config.openaiApiKey) {
       this.client = new OpenAI({
@@ -285,29 +254,18 @@ var ExtractionEngine = class {
     }
   }
   config;
+  fastChatCompletion;
   client;
   localLlm;
   fallbackLlm;
   modelRegistry;
   profiler;
-  /**
-   * Whether LLM calls should be routed through the gateway model chain
-   * instead of the plugin's own local/OpenAI clients.
-   */
   get useGatewayModelSource() {
     return this.config.modelSource === "gateway";
   }
-  /**
-   * Whether the local LLM path should be attempted.
-   * Disabled when gateway model source is active (gateway chain replaces local).
-   */
   get shouldUseLocalLlm() {
     return resolveLocalLlmCapabilities(this.config).localLlm && !this.useGatewayModelSource;
   }
-  /**
-   * Whether the direct OpenAI client should be used.
-   * Disabled when gateway model source is active.
-   */
   get shouldUseDirectClient() {
     return !this.useGatewayModelSource && this.client !== null;
   }
@@ -2121,27 +2079,55 @@ Respond with valid JSON matching this schema:
     }
   }
   async revalidateDependents(superseded, replacement, dependents, signal) {
+    const complete = async (messages, options) => {
+      try {
+        const fast = await this.fastChatCompletion?.(messages, options);
+        if (fast) return fast;
+      } catch (error) {
+        if (isAbortError(error) || options.signal?.aborted) throw error;
+        log.warn(`dependency revalidation fast route failed: ${error}`);
+      }
+      if (this.shouldUseLocalLlm) {
+        try {
+          const local = await this.localLlm.chatCompletion(messages, options);
+          if (local || !this.config.localLlmFallback) return local;
+        } catch (error) {
+          if (!this.config.localLlmFallback) throw error;
+        }
+      }
+      if (!this.shouldUseDirectClient) {
+        return this.fallbackLlm.chatCompletion(messages, this.withGatewayAgent(options));
+      }
+      const client = this.client;
+      if (!client) return null;
+      if (!shouldAssumeOpenAiChatCompletions(client.baseURL ?? this.config.openaiBaseUrl)) {
+        const response2 = await client.chat.completions.create(
+          {
+            model: this.config.model,
+            messages,
+            max_tokens: options.maxTokens
+          },
+          { signal: options.signal }
+        );
+        const content = response2.choices[0]?.message?.content;
+        return typeof content === "string" ? { content } : null;
+      }
+      const response = await client.responses.create({
+        model: this.config.model,
+        instructions: messages.find((message) => message.role === "system")?.content,
+        input: messages.filter((message) => message.role !== "system").map((message) => message.content).join("\n\n"),
+        max_output_tokens: options.maxTokens
+      }, { signal: options.signal });
+      return typeof response.output_text === "string" ? { content: response.output_text } : null;
+    };
     return revalidateDependentsViaLlm(
-      {
-        shouldUseLocalLlm: this.shouldUseLocalLlm,
-        shouldUseDirectClient: this.shouldUseDirectClient,
-        localLlm: this.localLlm,
-        fallbackLlm: this.fallbackLlm,
-        client: this.client,
-        config: this.config,
-        withGatewayAgent: this.withGatewayAgent.bind(this),
-        parseJsonObject: this.parseJsonObject
-      },
+      { fastChatCompletion: complete, parseJsonObject: this.parseJsonObject },
       superseded,
       replacement,
       dependents,
-      signal
+      { signal, timeoutMs: this.config.dependencyPropagation.timeoutMs }
     );
   }
-  /**
-   * Suggest links between a new memory and existing memories (Phase 3A).
-   * Called during extraction to build the knowledge graph.
-   */
   async suggestLinks(newMemory, candidateMemories) {
     if (candidateMemories.length === 0) {
       return { links: [] };
@@ -2456,4 +2442,4 @@ export {
   shouldEnableLocalExtractionThinking,
   ExtractionEngine
 };
-//# sourceMappingURL=chunk-3LD4N2TF.js.map
+//# sourceMappingURL=chunk-S3AYIQCU.js.map