Personality: Triggers and Context

Triggers (mention and word triggers)

Add a dedicated triggers block to control when the bot generates (separate from listen mode):

triggers:
  enabled: true
  on_mention: true
  words: ["assistant", "help"]   # case-insensitive substring by default
  use_regex: false               # when true, treat words entries as regex patterns

• DM always triggers.
• Mention triggers when on_mention: true (default).
• Word triggers fire when any pattern matches a message, even without a mention.

Context (last-N messages)

Control how many recent messages are included in the model input:

context:
  include_last_n: 12
  include_non_addressed_messages: true

• include_last_n: number of recent messages to include (hard-capped at 50).
• include_non_addressed_messages: when false, user messages are included only if they targeted the bot (mention or word trigger). Assistant messages are always included.

Personalities

Define personas in YAML files and pass them via --personality path.yml.

Schema - name: string identifier - language: optional; selects locale (en/fr) for built-in messages - developer_prompt: optional; prepended to the system prompt - system_prompt: required; the core behavior - command_prefix: optional; overrides global COMMAND_PREFIX - streaming (optional): pacing controls - rate_hz: messages per second (default 1.0) - min_first: minimum chars before first burst (default 80) - min_next: minimum chars before subsequent bursts (default 120) - environment (optional): context templates injected per message - guild_template: text appended when chatting in a server - dm_template: text appended when chatting in DMs - Placeholders: {guild_name}, {channel_name}, {member_names}, {user_display_name} - messages (optional): override built-in i18n strings to match persona tone - Keys: limit_reached, no_history, context_cleared, all_contexts_cleared, owner_only, generic_error, participants_visible

Example

name: example
language: en
developer_prompt: |
  Stay in character. Be concise, helpful, and safe. Keep replies under 1900 characters.
system_prompt: |
  You are a calm, friendly assistant. Provide clear, copy‑paste‑ready answers with brief
  explanations when useful. Ask a short clarifying question if requirements are ambiguous.
streaming:
  rate_hz: 1.0
  min_first: 80
  min_next: 120
command_prefix: "~"
environment:
  guild_template: |
    \nYou are speaking in a public Discord server. Visible members:\n{member_names}
  dm_template: |
    \nYou are in a direct message on Discord (DM).
messages:
  limit_reached: "Conversation limit of {max_turns} reached. Use '{prefix}reset' to continue."
  no_history: "No history for this channel."
  context_cleared: "Context cleared for this channel."
  all_contexts_cleared: "All contexts have been cleared."
  owner_only: "Owner-only action."
  generic_error: "Sorry, an error occurred while generating the reply."
  participants_visible: "Visible participants: {names}"

Guidelines - Keep system prompts concise and specific; avoid revealing internal rules. - Respect Discord constraints: keep messages short; the bot auto-chunks when needed. - For code-heavy personas, increase min_next to avoid breaking blocks mid-stream.