@infernetprotocol/cli 0.1.87 → 0.1.88

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/commands/start.js +11 -0
  2. package/package.json +9 -9
package/commands/start.js CHANGED
@@ -227,6 +227,10 @@ async function runDaemon(args, ctx) {
227
227
  const heartbeatMs = Number.parseInt(args.get('heartbeat-interval') ?? '', 10) || DEFAULT_HEARTBEAT_MS;
228
228
  const pollMs = Number.parseInt(args.get('poll-interval') ?? '', 10) || DEFAULT_POLL_MS;
229
229
  const once = args.has('once');
230
+ // Chat jobs run at once. Ollama on one GPU serves a single request at a
231
+ // time unless OLLAMA_NUM_PARALLEL is raised, so extra jobs only queue
232
+ // inside it; raise this together with that.
233
+ const maxConcurrentJobs = Number.parseInt(process.env.INFERNET_MAX_CONCURRENT_JOBS ?? '', 10) || 1;
230
234
 
231
235
  const p2pDisabled = args.has('no-p2p');
232
236
  const noAdvertise = args.has('no-advertise') || node.address === null;
@@ -618,6 +622,12 @@ async function runDaemon(args, ctx) {
618
622
  return;
619
623
  }
620
624
  if (stats.draining) return;
625
+ // Polls fire every pollMs whether or not the last batch finished, so
626
+ // without a cap a node ran one job per tick at once. The extras sat
627
+ // in Ollama's queue until fetch gave up ("fetch failed" at ~300 s,
628
+ // cullen3 2026-11-02). Jobs stay assigned here and run when a slot
629
+ // frees up.
630
+ if (stats.activeJobIds.size >= maxConcurrentJobs) return;
621
631
  const polledAt = Date.now();
622
632
  try {
623
633
  const result = await client.pollJobs({ limit: 5 });
@@ -635,6 +645,7 @@ async function runDaemon(args, ctx) {
635
645
  // new ones" (bonita, 0.1.83 -> 0.1.84). They stay assigned
636
646
  // and the restarted daemon picks them up.
637
647
  if (stats.draining) break;
648
+ if (stats.activeJobIds.size >= maxConcurrentJobs) break;
638
649
  await processJob(job);
639
650
  }
640
651
  } catch (err) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@infernetprotocol/cli",
3
- "version": "0.1.87",
3
+ "version": "0.1.88",
4
4
  "description": "GPU-node CLI / daemon for the Infernet Protocol — register a GPU server with a Supabase control plane and start earning.",
5
5
  "license": "MIT",
6
6
  "type": "module",
@@ -32,14 +32,14 @@
32
32
  "ora": "^9.4.1",
33
33
  "react": "^17.0.2",
34
34
  "react-blessed": "^0.7.2",
35
- "@infernetprotocol/auth": "0.1.87",
36
- "@infernetprotocol/config": "0.1.87",
37
- "@infernetprotocol/deploy-providers": "0.1.87",
38
- "@infernetprotocol/gpu": "0.1.87",
39
- "@infernetprotocol/engine": "0.1.87",
40
- "@infernetprotocol/liquid-agent": "0.1.87",
41
- "@infernetprotocol/rpc-adapter": "0.1.87",
42
- "@infernetprotocol/training": "0.1.87"
35
+ "@infernetprotocol/auth": "0.1.88",
36
+ "@infernetprotocol/config": "0.1.88",
37
+ "@infernetprotocol/deploy-providers": "0.1.88",
38
+ "@infernetprotocol/engine": "0.1.88",
39
+ "@infernetprotocol/gpu": "0.1.88",
40
+ "@infernetprotocol/liquid-agent": "0.1.88",
41
+ "@infernetprotocol/rpc-adapter": "0.1.88",
42
+ "@infernetprotocol/training": "0.1.88"
43
43
  },
44
44
  "repository": {
45
45
  "type": "git",