Daritas Docs

The Advisor

Ask plain-language questions about a system and get answers built only from its verified model — with citations, and an honest 'I don't know' when the model can't say.

What the Advisor is

The Continuity Advisor answers questions about a specific system using only that system’s verified model. It is not a general chatbot: it retrieves the relevant verified behaviours and phrases an answer strictly from them, citing each one. It is meant for the people who keep the installation running — operators, maintenance, and automation engineers — who need to know what the system does without reading the code.

Asking a question

Open an assessment and go to Ask the Advisor. Ask in plain language — “what happens on a freeze trip?”, “which dampers close on smoke?”, “how is the supply fan proven?”. Naming a specific asset, alarm, or function sharpens the answer.

Citations & review status

Every answer cites the behaviours it relied on. For the key facts it also states each one’s review status (human-confirmed, unverified, or flagged) and confidence, so you always know whether a statement has been signed off or is still provisional. The sources are listed beneath the answer, each with its evidence.

💡
Treat an unverified citation as provisional. The Advisor never hides that a fact hasn’t been signed off — the status is on every key point.

When it will not answer

If the model has no verified information on your question, the Advisor says so plainly rather than guessing. A “there is no verified information on that” is a real, useful answer — it marks a genuine gap, which you can turn into a clarification or a follow-up assessment.

Saving & sharing

  • Answers are saved to your history, so you can ask on one device and read on another.
  • You can share a question and its answer within your organization, so a useful explanation becomes team knowledge.

Read-only & advisory

⚠️
The Advisor is read-only and advisory. It describes what the system does and what to be careful about; it never instructs anyone to take a control action. It is a source of understanding, not a controller.

Every answer is also recorded in your AI runtime ledger, with the context it was given — part of keeping AI use in and around the installation governed and auditable. See Governance & AI.