Review the data path across documents, embedded files, structured sources and approved business actions. Keep retrieval and computed evidence distinct.
Preserve document versions, parent-child references and structured types. Keep searchable knowledge available while replacements are processed.
Apply the selected topic's retrieval strategy, source permissions and approved read actions. Flash and graph strategies have different capabilities.
Retain document references, transcript timestamps and computed-source provenance so reviewers can assess the answer.
Process supported files embedded in Office documents, PDFs and email files, where enabled. Keep child documents linked to the parent so users can understand where cited evidence came from. Supported formats, nesting and processing limits apply; access and lifecycle controls govern the document family.
Retain declared types, nested relationships and labelled paths in supported JSON records. Retrieve exact record fields alongside ordinary document evidence.
Set document routes and priorities within a topic, with explicit fallback and disclosure settings. Keep the current searchable version available while a replacement is processed.
Combine hybrid search, query decomposition, reranking and grounding checks. Preserve exact phrases, named sections, definitions, table identities and page references, linking individual claims to their supporting evidence.
Architecture and controls depend on the deployment. Review source quality, permissions, privacy, model failure, retention and export behaviour against your own requirements.
Bring representative sources, a useful business question and the controls your organisation needs.
