hermes - 💡(How to fix) Fix Honcho host-scoped cadence config is ignored by status and runtime [1 pull requests]

Summary

hermes honcho setup writes dialecticCadence under the active host block, for example hosts.hermes.dialecticCadence, but hermes honcho status and the Honcho memory provider runtime read only the root-level dialecticCadence from cfg.raw.

This means host-scoped cadence settings can appear ignored and the runtime falls back to every-turn dialectic calls.

Reproduction

1. Configure Honcho with a host-scoped cadence:

{
  "baseUrl": "http://127.0.0.1:8000",
  "hosts": {
    "hermes": {
      "enabled": true,
      "workspace": "hermes",
      "aiPeer": "hermes",
      "peerName": "dachuan",
      "recallMode": "hybrid",
      "dialecticCadence": 3
    }
  }
}

1. Run:

hermes honcho status

Expected

The status output and runtime provider should use the host-level value:

Dialectic cad:  every 3 turns

Actual

Status reports:

Dialectic cad:  every 1 turn

The provider initialization has the same issue because it reads directly from root raw:

self._injection_frequency = raw.get("injectionFrequency", "every-turn")
self._context_cadence = int(raw.get("contextCadence", 1))
self._dialectic_cadence = int(raw.get("dialecticCadence", 1))

cli.py status similarly uses:

dialectic_cadence = raw.get("dialecticCadence") or 1

Impact

Host-scoped Honcho config mostly works through HonchoClientConfig.from_global_config(), but these cost-awareness fields bypass that resolved config:

• dialecticCadence
• contextCadence
• injectionFrequency

This is especially confusing because the setup wizard writes dialecticCadence to the host block.

Suggested fix

Add these fields to HonchoClientConfig and parse them with host-first precedence:

• injection_frequency
• context_cadence
• dialectic_cadence

Then have provider initialization and status output use the resolved config fields instead of reading root cfg.raw directly.