NousResearch/hermes-agent · error · RuntimeError

Provider has been unresponsive (no response received) for {_

Error message

Provider has been unresponsive (no response received) for {_streak} consecutive stale attempts — aborting this call to avoid an indefinite stall. Switch models or start a new session, then retry.

What it means

Cross-turn stale-call circuit breaker (#58962): the agent tracks _consecutive_stale_streams, incremented on every stale kill and reset only when a call completes or the provider is swapped. Past HERMES_STREAM_STALE_GIVEUP (default 5) consecutive stale attempts, _check_stale_giveup() raises immediately — no network attempt, no stale-timeout wait — to stop a session wedged against an unresponsive provider from looping forever (observed 494 consecutive failures over days).

Source

Thrown at agent/chat_completion_helpers.py:386

        logger.debug("stale status buffering failed", exc_info=True)


def _touch_stale_kill_activity(agent, elapsed: float) -> None:
    try:
        agent._touch_activity(
            f"stale non-streaming call killed after {int(elapsed)}s"
        )
    except Exception:
        logger.debug("stale activity touch failed", exc_info=True)


def _check_stale_giveup(agent) -> None:
    """Raise immediately when the consecutive-stale streak is past the
    give-up threshold — no network attempt, no stale-timeout wait."""
    _giveup = env_int("HERMES_STREAM_STALE_GIVEUP", 5)
    _streak = _stale_streak(agent)
    if _giveup > 0 and _streak >= _giveup:
        raise RuntimeError(
            "Provider has been unresponsive (no response received) for "
            f"{_streak} consecutive stale attempts — aborting this call to "
            "avoid an indefinite stall. Switch models or start a new "
            "session, then retry."
        )


def _derive_stream_stale_timeout(agent, api_kwargs: dict) -> float:
    """Stale-stream patience for a provider that is never a local endpoint.

    Mirrors the main streaming path's derivation — provider config → env base
    → context-size scaling → reasoning-model floor — minus the local-endpoint
    ``float('inf')``/900s disable branch, which cannot apply to Bedrock (its
    endpoint is always the AWS cloud). Factored so the Bedrock streaming
    watchdog shares the exact same patience budget as the OpenAI/Anthropic
    stale-stream detector below.
    """
    _cfg_stale = get_provider_stale_timeout(agent.provider, agent.model)

View on GitHub (pinned to c896c09c42)

Solutions

Switch models/providers (`hermes model`) — swapping the provider resets the streak.
Start a new session, then retry.
Fix the underlying connectivity: check the provider status page, base_url, proxy.
As a last resort for controlled environments, raise HERMES_STREAM_STALE_GIVEUP (or set 0 to disable) — understand this re-enables the indefinite-stall loop the breaker exists to stop.

Example fix

# temporary, diagnostics only
export HERMES_STREAM_STALE_GIVEUP=10

Defensive patterns

Strategy: fallback

Validate before calling

from agent.chat_completion_helpers import _stale_streak
if _stale_streak(agent) >= 4:  # one below default give-up of 5
    switch_to_backup_provider_before_next_call()

Try / catch

try:
    response = call_model(...)
except RuntimeError as e:
    if "consecutive stale attempts" in str(e):
        agent.switch_model(backup_provider)  # resets the streak
        response = call_model(...)

Prevention

Configure a fallback model so `hermes model` swaps are one step.
Watch for repeated stale kills and act on the first one, not the fifth.
Do not raise HERMES_STREAM_STALE_GIVEUP in production — it re-enables the multi-day stall this breaker prevents.

When it happens

Trigger: Five (default) consecutive API calls each stalled with no response bytes until the stale detector killed them; the next call then aborts instantly at the pre-flight check in agent/chat_completion_helpers.py:386.

Common situations: Provider outage or hard-hanging endpoint; a proxy accepting connections but never responding; rate-limited endpoint that holds connections open; switching models resolves it because the streak measured the OLD provider.

Related errors

AI-assisted analysis of NousResearch/hermes-agent@c896c09c42 (2026-08-14). Data as JSON: /api/errors/0fee8e5f32c2defc. Report an issue: GitHub.