How Areinum finds and checks information behind an answer

A request identifies a matter, a customer, a person, a date, or another useful clue. In the demo, the graph indexes metadata such as titles, names, dates, and identifiers, helping the assistant find related records.

The graph indexes metadata, not the text of messages or attachments. When more detail is needed, the assistant can read selected record text. The scope and freshness of the available sources still affect the answer.

1. Define the matter and the period you are asking about.
2. Find the relevant records and identify their sources.
3. Read the selected details needed for the question.
4. Check the result and any missing or conflicting information.

Imagine a fictional contractor asks what changed in a deck project. An original scope lists the deck surface; a later customer message asks for new lighting. The useful result names the addition, cites the two records, and states whether approval of the change is visible.

If the available material includes a proposed price but no acceptance, the answer should say that the price is proposed and identify what still needs the customer's decision.

The business tools were rewritten in Rust and can be tested in the platform. They include tools for file work and calculations, alongside graph operations for finding, counting, grouping, and summing relevant records.

Code handles defined operations while the assistant brings the result into the conversation. The time needed depends on the task and the information involved.

Check that the answer covers the requested period and the right customer or job. Inspect the source behind any consequential amount or statement, and compare the date of the record with the data-up-to information when it is provided.

If a source is inaccessible or two records disagree, resolve that gap before relying on the answer. A missing result may reflect incomplete information; check another appropriate source.