osborn 0.9.242 → 0.9.244

package/dist/claude-llm.d.ts

@@ -125,10 +125,6 @@ export declare const VERIFIER_CHAIN: Record<string, {
     parallelAtStart?: string[];
     afterReminder?: string;
 }>;
-/**
- * Claude LLM - Wraps Claude Agent SDK for LiveKit
- * Research mode: reads anything, writes only to session workspace
- */
 export declare class ClaudeLLM extends llm.LLM {
     #private;
     constructor(opts?: ClaudeLLMOptions);

package/dist/claude-llm.js

@@ -17,7 +17,7 @@ import { getEmbedder } from './embedder.js';
 import { existsSync, readdirSync, readFileSync, writeFileSync, mkdirSync, statSync } from 'node:fs';
 import { join, dirname, resolve, basename } from 'node:path';
 import { fileURLToPath } from 'node:url';
-import { homedir } from 'node:os';
+import { homedir, loadavg, totalmem, freemem, cpus } from 'node:os';
 // Directory of this module — used to locate co-located prompt files (e.g., turn-shape reminder).
 const __claudeLlmDir = dirname(fileURLToPath(import.meta.url));
 const TURN_SHAPE_REMINDER_PATH = join(__claudeLlmDir, 'prompts', 'turn-shape-reminder.md');
@@ -994,6 +994,39 @@ class MessageChannel {
  * Claude LLM - Wraps Claude Agent SDK for LiveKit
  * Research mode: reads anything, writes only to session workspace
  */
+/**
+ * Lightweight machine-load nudge for the MAIN agent's UserPromptSubmit context.
+ * Pure in-memory/synchronous reads (os load average + memory + live sub-agent
+ * dispatch count) — NO embedder, DB, or network — so it stays reliable even when
+ * the box is starving, which is exactly when it matters. ADVISORY ONLY: it never
+ * refuses, caps, or blocks a dispatch — the agent reads it and decides. Returns
+ * '' on any error so it can never break the turn.
+ */
+function buildMachineNudge() {
+    try {
+        const ncpu = Math.max(1, cpus().length);
+        const cpuPct = Math.round((loadavg()[0] / ncpu) * 100); // 1-min load ÷ cores
+        const memPct = Math.round((1 - freemem() / totalmem()) * 100);
+        const subagents = statusManager.getActiveDispatchIds().length;
+        // Recommended concurrent heavy sub-agents, adapted to live headroom. This is a
+        // concrete DEFAULT the agent should follow unless a task genuinely warrants
+        // otherwise — surfaced as an actual number/target, never enforced.
+        const recMax = (cpuPct >= 80 || memPct >= 90) ? Math.max(1, ncpu - 1) // hot: pull back below cores
+            : (cpuPct >= 60) ? ncpu // busy: ~cores
+                : ncpu + 1; // calm: a little headroom
+        const over = subagents > recMax;
+        const warn = over || cpuPct >= 80;
+        const head = `[MACHINE${warn ? ' ⚠️' : ''}] CPU ~${cpuPct}% · MEM ${memPct}% · sub-agents ${subagents}/${recMax} recommended`;
+        if (over)
+            return `${head} — over the recommended concurrency; wind down to ≤${recMax} (finish or stop in-flight sub-agents) unless a task must continue. (recommendation — the agent decides, not enforced)`;
+        if (cpuPct >= 80)
+            return `${head} — load high; hold new sub-agents beyond ${recMax} unless needed. (recommendation — the agent decides, not enforced)`;
+        return head;
+    }
+    catch {
+        return '';
+    }
+}
 export class ClaudeLLM extends llm.LLM {
     #opts;
     #sessionId = null;
@@ -2028,8 +2061,12 @@ class ClaudeLLMStream extends llm.LLMStream {
                                                 buildRecallInjection(sid, this.#opts.workingDirectory, String(input?.prompt || '')),
                                                 new Promise((res) => setTimeout(() => res(''), 6000)),
                                             ]);
-                                        const additionalContext = recalled ? `${reminder}\n\n${recalled}` : reminder;
-                                        console.log(`📌 UserPromptSubmit: injected turn-shape reminder (${reminder.length} chars)${recalled ? ` + recall (${recalled.length} chars)` : ''} for prompt="${promptPreview}..." [tool budget reset to 0/${TOOL_CALL_BUDGET}]`);
+                                        // Lightweight machine-load nudge (main agent only). Cheap synchronous
+                                        // snapshot, prepended so it's present even when recall degrades under load.
+                                        // Advisory only — never caps/refuses a dispatch.
+                                        const machineNudge = fromSubagent ? '' : buildMachineNudge();
+                                        const additionalContext = [machineNudge, reminder, recalled].filter(Boolean).join('\n\n');
+                                        console.log(`📌 UserPromptSubmit: injected turn-shape reminder (${reminder.length} chars)${recalled ? ` + recall (${recalled.length} chars)` : ''}${machineNudge ? ` + machine(${machineNudge.includes('⚠️') ? 'HOT' : 'ok'})` : ''} for prompt="${promptPreview}..." [tool budget reset to 0/${TOOL_CALL_BUDGET}]`);
                                         return {
                                             hookSpecificOutput: {
                                                 hookEventName: 'UserPromptSubmit',

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "osborn",
-  "version": "0.9.242",
+  "version": "0.9.244",
   "description": "Voice AI coding assistant - local agent that connects to Osborn frontend",
   "type": "module",
   "bin": {