fixes for summary prompt, chat route logger, QDrant orphan cleanup, llamacpp stream buffering

packages/inference-service/src/providers/llamacpp.js

@@ -66,29 +66,52 @@ async function* completeStream(prompt, options = {}) {
   if (!res.ok)
     throw new Error(`llama.cpp error: ${res.status} ${res.statusText}`);
 
-  for await (const chunk of res.body) {
-    const lines = Buffer.from(chunk)
-      .toString("utf8")
-      .split("\n")
-      .filter((l) => l.startsWith("data: ") && l !== "data: [DONE]");
+  //SSE lines can be split across network chunks, so we need to buffer them until we have a complete line
+  let buffer = '';
 
-    for (const line of lines) {
-        const json = JSON.parse(line.slice(6));
-        const delta = json.choices?.[0]?.delta?.content;
+  function* processLine(line){
+    if (!line.startsWith('data: ') || line === 'data: [DONE]') return;
 
-        if (json.choices?.[0]?.finish_reason === 'stop') {
+    let json;
+
+    try {
+      json = JSON.parse(line.slice(6));
+    } catch (err) {
+      logger.error('[llamacpp] Skipping unparseable SSE line:', line.slice(0,120));
+      return;
+    }
+
+
+
+    const delta = json.choices?.[0]?.delta?.content ?? '';
+
+    if (json.choices?.[0]?.finish_reason === 'stop' ) {
       finalModel = json.model ?? finalModel;
+
     }
 
     // usage arrives in a separate final chunk with empty choices array
-        if (json.usage) {
+    if (json.usage){
       finalTokenCount = (json.usage.completion_tokens ?? 0) + (json.usage.prompt_tokens ?? 0);
     }
 
-        if (delta) yield { response: delta, done: false };
+    for (const line of lines) {
+      yield* processLine(line.trim());
     }
   }
 
+  for await (const chunk of res.body) {
+    buffer += Buffer.from(chunk).toString('utf-8');
+
+    const lines = buffer.split('\n');
+    buffer = lines.pop() ?? ""; // keep the last line in the buffer, it may be incomplete
+  }
+
+  //Flush anything left in buffer after stream closes
+  if (buffer.trim()){
+    yield* processLine(buffer.trim());
+  }
+
   logger.info('[llamacpp] finalTokenCount:', finalTokenCount);
 
   yield { response: '', done: true, model: finalModel, tokenCount: finalTokenCount };

packages/memory-service/src/episodic/index.js

@@ -62,6 +62,11 @@ function touchSession(id) {
 function deleteSession(id) {
     const db = getDB();
     db.prepare(`DELETE FROM sessions WHERE id = ?`).run(id);
+
+    //SQLite cascade removed episode rows, clean up QDrant vectors too
+    // Fire-and-forget - payload-filter delete, so it doesn't depend on the rows existing
+    semantic.deleteEpisodesBySession(id)
+      .catch(err => logger.error(`[Memory] QDrant cleanup failed for session ${id}:`, err.message));
 }
 
 function updateSession(id, { name, projectId } = {}) {

packages/memory-service/src/index.js

@@ -192,10 +192,14 @@ app.get('/entities/:id', (req, res) => {
 });
 
 
-
 // Delete an entity by ID
 app.delete('/entities/:id', (req, res) => {
-  entities.deleteEntity(req.params.id);
+  const id = Number(req.params.id);
+  entities.deleteEntity(id);
+
+  semantic.deleteEntity(id) //fire-and-forget, same pattern as episode delete
+    .catch(err => logger.error(`[Memory] Qdrant delete failed for entity ${id};`, err.message));
+    
   res.status(204).send();
 });
 

packages/memory-service/src/semantic/index.js

@@ -99,6 +99,23 @@ async function deleteEpisode(id) {
   return deleteVector(COLLECTIONS.EPISODES, id);
 }
 
+async function deleteEntity(id) {
+    return deleteVector(COLLECTIONS.ENTITIES, id);
+}
+
+//Delete ALL episode vectors associated with a session via payload filter
+//Used on session delete: SQLite cascade removes the rows,  
+//Filter-based, 
+async function deleteEpisodesBySession(sessionId) {
+    const client = getClient();
+    await client.delete(CoLLECTIONS.EPISODES, {
+        wait: true,
+        filter: {
+            must: [{ key: 'sessionId', match: { value: sessionId } }]
+        }
+    });
+}
+
 
 module.exports = {
     initCollections,
@@ -109,5 +126,7 @@ module.exports = {
     searchEntities,
     searchSummaries,
     deleteVector,
-    deleteEpisode
+    deleteEpisode,
+    deleteEntity,
+    deleteEpisodesBySession
 };

packages/memory-service/src/summarization/project.js

@@ -32,6 +32,9 @@ function buildProjectSummaryPrompt(projectName, sessionSummaries) {
         'Be comprehensive but avoid padding. Do not repeat the same point twice.',
         'Write in third person. Output only the overview text, no headings or labels.',
         '',
+        summaryBlock,
+        '<|im_end|>',
+        '<|im_start|>assistant',
     ].join('\n');
 }
 

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

@@ -1,7 +1,7 @@
 const { Router } = require('express')
 const { chat, chatStream } = require('../chat/index');
 const memory = require('../services/memory')
-const logger = require('@nexusai/shared');
+const { logger } = require('@nexusai/shared');
 
 
 const router = Router();