trivial turn patch

packages/orchestration-service/src/chat/index.js

@@ -2,7 +2,7 @@ const memory = require("../services/memory");
 const inference = require("../services/inference");
 const embedding = require("../services/embedding");
 const qdrant = require("../services/qdrant");
-const { ORCHESTRATION, RETRIEVAL, logger } = require("@nexusai/shared");
+const { ORCHESTRATION, RETRIEVAL, logger, isTrivialTurn } = require("@nexusai/shared");
 const appSettings = require("../config/settings");
 const {triggerSummary} = require('../services/summarization')
 const graph = require('../services/graph');
@@ -283,20 +283,27 @@ async function assembleContext(externalId, userMessage) {
     const isFirstMessage = recentEpisodes.length === 0;
     const recentIds = new Set(recentEpisodes.map(e => e.id));
 
-    // 4. Embed the query once — the vector is shared by semantic episode search
-    //    and entity search, so embedding it twice was a wasted round-trip + Ollama call.
-    let queryVector = null;
-    try {
-        queryVector = await embedding.embed(userMessage);
-    } catch (err) {
-        logger.warn('[orchestration] Query embedding failed; semantic + entity search disabled this turn:', err.message);
-    }
+    // 4. Retrieval — skipped entirely on contentless social turns (greetings,
+    //    sign-offs). On those, recent history alone is the right context; running
+    //    semantic/keyword/entity retrieval only surfaces marginal noise the model
+    //    then confabulates around. Embed once (shared by episode + entity search).
+    let fusedWithScores = [];
+    let entityResults = [];
+    if (!isTrivialTurn(userMessage)) {
+        let queryVector = null;
+        try {
+            queryVector = await embedding.embed(userMessage);
+        } catch (err) {
+            logger.warn('[orchestration] Query embedding failed; semantic + entity search disabled this turn:', err.message);
+        }
 
-    // 4b. Fused retrieval + entity search in parallel (both are independent)
-    const [fusedWithScores, entityResults] = await Promise.all([
-        getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, queryVector, { semanticLimit, scoreThreshold, semanticWeight, keywordWeight }),
-        getRelevantEntities(queryVector, session.project_id ?? null),
-    ]);
+        [fusedWithScores, entityResults] = await Promise.all([
+            getFusedEpisodes(userMessage, session, recentIds, projectSessionIds, queryVector, { semanticLimit, scoreThreshold, semanticWeight, keywordWeight }),
+            getRelevantEntities(queryVector, session.project_id ?? null),
+        ]);
+    } else {
+        logger.debug('[orchestration] Trivial turn — skipping semantic/keyword/entity retrieval');
+    }
 
     // 5. Entity-linked episode IDs for scoring bonus
     const entityIds = entityResults.map(e => e.id);