tinyhumansai/openhuman · error · Error

RPC ${method} timed out after ${timeoutMs}ms

Error message

RPC ${method} timed out after ${timeoutMs}ms

What it means

rpc() in harness-cache-audit.mjs aborts any JSON-RPC POST exceeding its per-call timeout (default 600000 ms, tunable via --rpc-timeout-ms; the internal core.ping probe uses 10 s) and translates the resulting AbortError into this message naming the method and limit. Each audit turn is a full openhuman.agent_chat delegation, so the timeout primarily bounds model inference plus tool execution time.

Source

Thrown at scripts/debug/harness-cache-audit.mjs:216

  let res;
  try {
    res = await fetch(coreUrl, {
      method: "POST",
      signal: controller.signal,
      headers: {
        "content-type": "application/json",
        authorization: `Bearer ${token}`,
      },
      body: JSON.stringify({
        jsonrpc: "2.0",
        id: `audit-${Date.now()}-${Math.random().toString(16).slice(2)}`,
        method,
        params,
      }),
    });
  } catch (err) {
    if (err?.name === "AbortError") {
      throw new Error(`RPC ${method} timed out after ${timeoutMs}ms`);
    }
    throw err;
  } finally {
    clearTimeout(timeout);
  }
  const bodyText = await res.text();
  let body;
  try {
    body = JSON.parse(bodyText);
  } catch {
    throw new Error(`RPC ${method} returned non-JSON HTTP ${res.status}`);
  }
  if (!res.ok) {
    throw new Error(`RPC ${method} HTTP ${res.status}`);
  }
  if (body.error) {
    throw new Error(`RPC ${method} error`);
  }

View on GitHub (pinned to a221052e0d)

Solutions

  1. Raise the limit: `--rpc-timeout-ms 1200000`
  2. Retry — transient provider stalls often clear
  3. Reduce work: fewer `--turns`, cheaper `--model`, shorter `--prompt`
  4. Check whether the core is alive/responsive (GET /health) to distinguish hung core from slow turn
  5. Rerun with --verbose to see per-turn response characters and where it stalls

Example fix

# before
node scripts/debug/harness-cache-audit.mjs --turns 4 --rpc-timeout-ms 120000
# Error: RPC openhuman.agent_chat timed out after 120000ms

# after
node scripts/debug/harness-cache-audit.mjs --turns 4 --rpc-timeout-ms 900000
Defensive patterns

Strategy: retry

Validate before calling

// pre-flight ping with a short budget to fail fast on connectivity problems
// before burning a 10-minute per-turn timeout
try {
  await rpc(coreUrl, token, "core.ping", {}, 5_000);
} catch {
  console.error("core not answering within 5s — resolve connectivity before the audit");
  process.exit(2);
}

Type guard

const isRpcTimeout = (err) => /RPC .* timed out after \d+ms/.test(err?.message || "");

Try / catch

let attempt = 0;
while (true) {
  try {
    return await rpc(coreUrl, token, "openhuman.agent_chat", params, opts.rpcTimeoutMs);
  } catch (err) {
    attempt += 1;
    if (isRpcTimeout(err) && attempt < 2) {
      await new Promise((r) => setTimeout(r, 5_000)); // one retry: provider stalls are often transient
      continue;
    }
    throw err;
  }
}

Prevention

When it happens

Trigger: An agent_chat turn with tool loops (file reads, skills, subagents) exceeding --rpc-timeout-ms; a wedged core; provider rate-limit backoff; lowering --rpc-timeout-ms for a smoke run and forgetting to restore it before a 3+ turn audit; system sleep suspending the process mid-request.

Common situations: Slow model or long cache-audit prompts pushing past 10 minutes; core also serving desktop traffic; laptop suspend during a run; CI network throttling to the inference provider.

Understand the failure class

Related errors

AI-assisted analysis of tinyhumansai/openhuman@a221052e0d (2026-08-16). Data as JSON: /api/errors/25719e32a103afc6. Report an issue: GitHub.