@tiens.nguyen/gonext-local-worker 1.0.22 → 1.0.25

package/gonext-local-worker.mjs

@@ -162,6 +162,10 @@ async function runChatJob(job) {
   /** Chains debounced chunk POSTs so we never PATCH `completed` while a chunk POST is still in flight. */
   let flushTail = Promise.resolve();
 
+  /** Batch streamed text: fewer HTTPS round-trips to the API than a 12ms debounce per flush. */
+  const CHUNK_DEBOUNCE_MS = 80;
+  const CHUNK_MAX_BUF = 6144;
+
   const flushChunks = async () => {
     const t = buf;
     buf = "";
@@ -199,13 +203,23 @@ async function runChatJob(job) {
     if (!s) return;
     fullText += s;
     buf += s;
+    if (buf.length >= CHUNK_MAX_BUF) {
+      if (flushTimer) {
+        clearTimeout(flushTimer);
+        flushTimer = null;
+      }
+      flushTail = flushTail.then(() => flushChunks()).catch((err) => {
+        console.error("[gonext-worker] chunk flush error:", err);
+      });
+      return;
+    }
     if (!flushTimer) {
       flushTimer = setTimeout(() => {
         flushTimer = null;
         flushTail = flushTail.then(() => flushChunks()).catch((err) => {
           console.error("[gonext-worker] chunk flush error:", err);
         });
-      }, 12);
+      }, CHUNK_DEBOUNCE_MS);
     }
   };
 
@@ -251,6 +265,8 @@ async function runChatJob(job) {
     await flushTail;
     await flushChunks();
 
+    logModelResponseToWorker(jobId, payload.modelId, fullText);
+
     const totalTimeSeconds = (Date.now() - start) / 1000;
     await workerFetch(`/api/worker/jobs/${jobId}`, {
       method: "PATCH",
@@ -300,6 +316,16 @@ function sourceLabelFromBase(base) {
   }
 }
 
+/** Log assistant text to stdout; cap size so huge replies do not flood the terminal. */
+function logModelResponseToWorker(jobId, modelId, text) {
+  const max = 12000;
+  const n = text.length;
+  const body = n <= max ? text : `${text.slice(0, max)}\n… [log truncated: ${n - max} more chars]`;
+  console.log(
+    `[gonext-worker] model reply job=${jobId} model=${modelId} chars=${n}:\n${body}`
+  );
+}
+
 async function checkOllamaTags(base) {
   const endpoint = `${base}/api/tags`;
   try {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tiens.nguyen/gonext-local-worker",
-  "version": "1.0.22",
+  "version": "1.0.25",
   "description": "Polls GoNext cloud API for async local LLM jobs and runs them against Ollama/OpenAI-compatible servers on this Mac",
   "type": "module",
   "license": "MIT",