Chat, web search, voice, real time

The simple chat

This is how the Spotfinder runs today: type a question, the answer appears word by word, below it spot buttons, the way to the map and suggestions to tap. Usable without signing in; with an account the conversations are kept.

Dictation and read-aloud

Available once enabled: dictate the question through the microphone and have the answer read aloud. Only the answer text is read – not the buttons and not the JSON behind them. Minutes heard and characters spoken have limits of their own, per account and per tenant.

The real-time conversation

Available once enabled: a conversation by voice in both directions over WebRTC. The key to the speech AI never leaves the server, and the server remains the only one that starts answers. Because a voice knows no fixed answer format, the Spotfinder’s format becomes a tool through which the spots and the map still reach the page. The length of a conversation is limited.

Web search

Switched on, but targeted: the Spotfinder only searches the internet when someone explicitly asks for something current that its data does not hold – events and competitions, closures, opening hours, getting there or rules at a spot. Wind, statistics, temperature and waves still come from the tools only.

The pages found appear as source buttons below the answer; the text itself contains no links. Web search has daily limits of its own – per address without signing in, per person with an account, plus per tenant and in total. Once it is used up for the day, the Spotfinder answers from its data and says so in one sentence.

The search query leaves the data-protection boundary of the Azure services and goes to Microsoft Bing; the privacy policy describes this. Name, account and IP address are not sent along.