Skip to content

Inactivity timeout fires repeatedly with MiniMax 2.7 highspeed #6260

@Helmi

Description

@Helmi

Since switching to MiniMax 2.7 highspeed, the gateway inactivity timeout (default 30 min) fires multiple times per day during active sessions. The agent appears to be waiting on an API response when no tool calls are being made, but sessions resume normally after /reset.

Original error message

⏱️ Agent inactive for 30 min — no tool calls or API responses.
Last activity: API call #6 completed (4091s ago, iteration 6/100). The agent may have been waiting on an API response.
To increase the limit, set agent.gateway_timeout in config.yaml (value in seconds, 0 = no limit) and restart the gateway.
Try again, or use /reset to start fresh.

Environment

  • Model: MiniMax 2.7 highspeed
  • Hermes version: 0.8.0
  • OS: macOS (local), Arch Linux (Archie)

Root Cause Analysis

The inactivity tracker (run_agent.py._touch_activity()) resets on every tool call, API call initiation, and stream token received. If 30 minutes pass with no activity, gateway/run.py (lines 7141–7210) triggers a timeout.

The mechanism was introduced in #4815 specifically to catch genuinely hung API calls — but it does not distinguish between a hung call and normal API latency during streaming responses.

It is unclear whether the repeated timeouts are caused by longer-than-expected MiniMax API response times, network latency, or something else. The pattern began after switching to MiniMax 2.7 highspeed and occurs multiple times per day.

Related issues

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions