Aurelia docs
How it works
Aurelia combines page context, collection data, retrieval, response orchestration, and optional live research into a single answer loop designed for branded hotel experiences.
Engineers and CTO-level readers evaluating system architecture.
8 min
Understand the system loop from page context to retrieval, response generation, and handoff.
The response loop
Every answer starts from context, not from a blank chat box.
- Collect context
Aurelia receives page type, active section, pinned hotel, visible hotel set, and any existing stay preferences.
- Determine intent and scope
The system decides whether the guest is planning broadly, comparing a shortlist, evaluating a specific hotel, or preparing for booking.
- Retrieve evidence
Search runs against the indexed hotel knowledge first, then official or trusted sources only if the question is thin or time-sensitive.
- Generate the response
The model answers from grounded context, shapes the response for the current task, and returns follow-up prompts or widgets when useful.
- Guide the next action
Aurelia keeps the guest moving with compare actions, section links, prompt suggestions, or a rate handoff instead of ending at a text answer.
Why page context matters
The same question should produce different answers depending on where the guest is.
A generic chat interface would treat the whole website as one undifferentiated surface. Aurelia does not. A question on the homepage behaves like trip planning. The same question on a hotel page behaves like property evaluation. A question on the booking page behaves like a final room-fit or arrival check.
- Homepage context biases toward planning and shortlist generation.
- Destination context biases toward the visible hotel set and compare flows.
- Hotel context biases toward pinned follow-ups such as dining, room fit, or nearby guidance.
- Booking context biases toward concise, confidence-building handoff answers.
Answer shaping and guardrails
Different question classes get different grounding and response styles.
Aurelia classifies whether the user is asking a planning question, an amenity question, a policy question, a compare question, or a local-area question. That classification changes how much evidence is required, whether live lookup is allowed, and whether the answer should stay plain or become more structured.
A hotel-policy answer should be more cautious than a general planning answer. A local-area answer may justify live research. A compare answer should stay grounded to the shortlisted options instead of broadening unnecessarily.
The answer is not the end of the flow
Aurelia behaves like a guided product surface, not a static FAQ.
- Prompt suggestions adapt after each turn.
- Pinned-hotel context keeps follow-ups coherent.
- Structured widgets can be used for hotel grids, comparisons, and section navigation.
- The final goal is to reduce uncertainty and increase qualified movement into rates or the next product step.