What problem does it solve?
It lets you dynamically ground responses in named, local persona configurations so the same agent can reliably behave like different experts, with no ambiguity about which “version” you mean.
Core Features & Use Cases
- Auto-discovery of local personas: Reads the live set of YAML persona files from
~/.hermes/pantheon/personas/ on every turn, so changes take effect immediately.
- Proper-noun grounded persona selection: Detects proper nouns in your message that match persona names and uses the corresponding YAML as the first-class source of truth.
- Persona-specific actions: Supports “do X as <persona>,” description of a persona, and comparisons (by reading multiple YAMLs) to keep behavior consistent and locally defined.
- Pantheon visibility: Lists persona
name and job fields from the current pantheon YAMLs.
Quick Start
Tell the assistant: “Orpheus, summarize this PR and suggest the most likely architectural risks.”