@tiens.nguyen/gonext-local-worker 1.0.63 → 1.0.64

package/gonext-local-worker.mjs

@@ -1219,13 +1219,6 @@ async function runAgentChatJob(job) {
   console.log(
     `[gonext-worker] agent_chat ${jobId} baseURL=${payload?.agentBaseURL ?? "(none)"} modelId=${payload?.agentModelId ?? "(none)"}`
   );
-  // Send an immediate heartbeat so the web 60-180s no-progress timer doesn't
-  // fire while the local model is loading/generating its first reasoning step.
-  enqueueText("<think>Agent starting…\n\n");
-  flushTail = flushTail.then(() => flushChunks()).catch((err) => {
-    console.error("[gonext-worker] agent_chat heartbeat flush error:", err);
-  });
-
   try {
     const python =
       (process.env.GONEXT_PROBE_PYTHON ?? process.env.GONEXT_MLX_LM_PYTHON ?? "")
@@ -1241,7 +1234,7 @@ async function runAgentChatJob(job) {
     });
     const timeoutMs = 300_000; // 5 min max for an agent run
 
-    let inThink = true; // already opened <think> above
+    let inThink = false;
     let finalText = "";
 
     await runProcessWithStreamingStdout(python, [scriptPath], input, timeoutMs, (event) => {

package/gonext_agent_chat.py

@@ -194,18 +194,18 @@ def run_agent_chat(cfg):
     _log(f"task={task_text[:120]!r}")
 
     # Route: ask the model if this task needs HTTP tool use.
-    _emit({"type": "step", "text": "Routing…"})
     needs_agent = _route(task_text, agent_base_url, agent_api_key, agent_model_id)
 
     if not needs_agent:
         _log("router: plain chat (no HTTP needed)")
-        _emit({"type": "step", "text": "No HTTP request needed — answering directly…"})
         answer = _plain_reply(task_text, agent_base_url, agent_api_key, agent_model_id)
         _log(f"plain reply: {len(answer)} chars")
         _emit({"type": "final", "text": answer})
         return
 
+    # Agent path — from here all step events go into <think>.
     _log("router: agent (HTTP tool use needed)")
+    _emit({"type": "step", "text": "Planning HTTP request…"})
 
     # Prepend explicit tool instructions so small models use http_request correctly
     # and always terminate with final_answer() rather than looping forever.
@@ -284,7 +284,6 @@ def run_agent_chat(cfg):
             executor_kwargs={"timeout_seconds": 60},
             additional_authorized_imports=["json", "urllib", "urllib.request", "urllib.error"],
         )
-        _emit({"type": "step", "text": f"Sending task to {agent_model_id}…"})
         with contextlib.redirect_stdout(sys.stderr):
             result = agent.run(task_with_hint)
         final_text = str(result).strip()

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tiens.nguyen/gonext-local-worker",
-  "version": "1.0.63",
+  "version": "1.0.64",
   "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",