Skip to main content

Design principles and privacy boundaries

Mystilink builds its workflow around local records, charts you can verify, and conversation that continues across turns. These principles help you decide which results you can check directly and which still need your judgment.

Stored locally, inferred according to your access path​

Profile cards, conversations, and related working data live in the workspace; interface preferences and model settings live in your local user directory. Accounts, licensing, and Newton are online services and still process the corresponding requests.

With Newton, the required context is processed by the platform and upstream model services; BYOK goes to the vendor or gateway you configured; local deployment goes to the server you run. Storage location and inference location are separate questions, see Privacy and local data.

Check the input and the calculation before reading the interpretation​

A natal chart is produced by the relevant tool from your input and its rules. If a profile card points at the wrong person, date, hour, or place, fluent prose cannot compensate for a wrong input. When comparing software, align the input and the rules first, then look at the actual differences.

Interpretation is another layer: the model may misread the chart, drop a condition, or make a prediction that cannot be verified. You can ask it to list the records, tool outputs, and assumptions it used before explaining the conclusion; see Charting issues.

Inquiries are information checks​

When information is incomplete, an inquiry helps pin down the person, the conditions, or the candidate explanations. You can state plainly what is unknown or forgotten. A life event matching one explanation does not prove a birth time; see Chart confirmation and inquiries.

What each of the two choices controls​

  • Metaphysics system: controls the charting or divination path, such as BaZi, Zi Wei, tarot, six lines, and astrology.
  • Model: controls the service that understands and interprets, with its own permissions, usage, cost, and data path.

Choosing a theory automatically does not choose a cheaper model automatically; configuring a model fallback does not switch the metaphysics system either. See Model architecture.

The process is visible, the conclusion still needs judgment​

Tool panels, inquiries, and the task list let you inspect the process. A completed task means that step finished, not that the conclusion was verified. When reading results, separate chart data, traditional interpretation, and claims about reality or the future.

As personal records accumulate, back up the workspace periodically, and check for sensitive content before sharing screenshots or diagnostic output.