Problem or Use Case

Gateway already supports /model, but there is no matching /reasoning command for agent.reasoning_effort.

Today, changing reasoning effort requires editing ~/.hermes/config.yaml by hand. That is clumsy for gateway users, and it is also inconsistent with /model, which already has a "takes effect on next message" workflow.

Proposed Solution

Add a gateway /reasoning [xhigh|high|medium|low|minimal|none] command that:

• shows the current reasoning effort when called without arguments
• writes the selected value back to config.yaml
• applies the new setting on the next message without restarting the gateway
• exposes the command in Telegram, Slack, and Discord gateway entrypoints

To make hot reload behavior reliable, gateway should re-read reasoning config before each AIAgent(...) construction, including background tasks, and treat config.yaml as the source of truth for the long-running process.

Alternatives Considered

• Keep reasoning changes as manual config.yaml edits plus gateway restart. This is higher friction and inconsistent with /model.
• Add /reasoning but keep startup-only caching. That still leaves the command with surprising behavior.

Feature Type

Gateway / messaging improvement

Scope

Medium (few files, < 300 lines)

Contribution

• I'd like to implement this myself and submit a PR