dsh-llm-fallbacks

Minimal configuration

Add a fallbacks: section to the dsh settings document (default $DSH_HOME/settings.yaml):

fallbacks:
  enabled: true            # feature switch; defaults to false — set explicitly to enable
  rootChain:               # block 1: root agent's chain; tried in order after the primary model fails
    - anthropic/claude-3-5-sonnet
    - openai/*
  roles:                   # block 2: declare roles first, then let rules reference them
    list:
      - id: reviewer       # role entity: unique id (/^[a-z0-9-]{1,32}$/); "inherit" is reserved
        label: Reviewer
        description: Code-review subagents
        chain:
          - openai/gpt-4o-mini
        fallback: inherit-root   # default: role chain, then append rootChain
    rules:
      - origin: subagent   # all subagents → reviewer role (own chain + inherited root)
        role: reviewer

Roles are declared entities: roles.list holds role cards (id/label/description + optional chain/fallback), and roles.rules match origin/provider/model in order to a declared role id or the built-in inherit (first match wins) — no rule match → inherit → rootChain. A declared role is only ever hit when a rule references it. The legacy chain-key namespace and role-default field are gone (migration table: docs/configuration.md).

Save and restart the web session for the changes to take effect. The feature switch fallbacks.enabled defaults to off (false) — the plugin only engages once it is turned on; triggerCodes defaults to AUTH / QUOTA / RATE_LIMIT; and with no rootChain/role chains configured the behavior is identical to not having the plugin installed. More examples (role entities, fallback strategies, rules referencing inherit) → docs/configuration.md.

Upgrade note (behavior change): an existing fallbacks: section without an explicit enabled key now resolves to false after upgrading — add enabled: true to keep the plugin active.