The Orchestrator Settings page (Orchestrator > Settings) is where administrators configure the core behavior of the AI Colleagues and Master Orchestrator. These settings control the assistant's system instructions, response disclaimers, status messages, feedback handling, session timeouts, and voice capabilities.

Changes made here take effect immediately — the system invalidates its configuration cache on every update, so the next user request will use the new settings.

Settings Reference

Additional Instructions

An optional supplementary instruction field that is appended to the main system prompt. Use this to layer on extra behavioural guidance without modifying the core description.

When to use Additional Instructions:

• Organisational policies — Add company-specific rules the assistant should always follow (e.g., "Always direct salary-related questions to the HR team," or "Do not disclose headcount data").
• Seasonal or temporary guidance — Inject time-bound instructions (e.g., "Open enrollment runs from March 1–15. Proactively inform users asking about benefits") without touching the stable base prompt.
• Tone or formatting overrides — Provide instructions that refine how the assistant communicates (e.g., "Keep responses under 3 sentences when answering FAQ-style questions," or "Always include a follow-up question at the end of your response").
• Guardrails and compliance — Enforce boundaries (e.g., "Never provide legal or medical advice. Always recommend the user consult a professional," or "Do not discuss competitor products").

How it works at runtime: The additional_description is stored alongside the main description on the Orchestrator model. During execution, the system combines both fields — along with dynamically generated runtime context (such as the current timestamp, user profile details, and link formatting rules) — into the final instruction set provided to the LLM for each conversation run.

Best practices:

• Keep instructions clear and specific. The LLM follows concrete directives more reliably than vague guidance.
• Use this field for instructions that may change over time, keeping the main Description stable.
• Avoid contradicting the base Description — if there's a conflict, the LLM may behave unpredictably.

Disclaimer Settings

Controls the automated AI-generated response disclaimer shown to users in chat.

• Enable/Disable — Toggle the disclaimer on or off.
• Disclaimer Text — Customise the message displayed (e.g., "This response was generated by AI and may not be fully accurate"). Only visible when the disclaimer is enabled.

Processing Placeholder Messages

Customise the status messages displayed to users while the assistant is working on a response.

• Thinking Text — Shown when the LLM is processing the user's query (e.g., "Thinking...").
• Running Text — Shown when the assistant is executing a skill or tool call (e.g., "Running...").
• Processing Results Text — Shown when the assistant is assembling the final response (e.g., "Processing results...").

These messages are localizable and will be translated based on the user's preferred language.

Thumbs Up/Down Feedback

Enables emoji-based feedback collection from users.

• Enable/Disable — When enabled, the assistant presents 👍/👎 options after responses.
• Thumbs Up Instructions — Custom instructions for how the assistant should respond when a user sends a thumbs up (e.g., "Acknowledge positively and ask if there's anything else you can help with").
• Thumbs Down Instructions — Custom instructions for how the assistant should respond to a thumbs down (e.g., "Apologise for the inconvenience, ask what went wrong, and offer to escalate to a human agent").

Max Fields to Fill Conversationally

Sets the maximum number of form fields the assistant will collect from the user through conversation before switching to a web-view or form-based interaction. This prevents overly long back-and-forth exchanges for complex forms.

Request Session Timeout

Defines how long a conversation session remains active after the last user message, configured in hours and minutes. Once the timeout expires, the next message from the user starts a fresh conversation thread. The default is 6 hours.

Enable Voice Call

A master toggle for real-time voice call capabilities. Enabling this activates the Voice Orchestrator, which handles spoken interactions, interruption management, and turn-taking. This setting is synchronized with the Core Service — if the sync fails, the update is rolled back to prevent configuration mismatches.