Document intelligence, connections, and knowledge

What is document intelligence in DataFact?

Document intelligence is the process of turning supported files into source material that can be organized, parsed, retrieved, and inspected in DataFact workflows. The Document Library stores uploaded files and folders. A compatible MemoryLake document resource makes selected material usable by an Agent or report workflow. Parsing and indexing preserve the structure that the selected document path supports, while citations can point back to retrieved source context when the workflow returns them.

Document intelligence is related to connections and knowledge, but it is not the same object. A connection provides access to a source. A MemoryLake resource defines material that a workflow can use. Knowledge adds semantic or business guidance. Uploading a file does not automatically create a usable document resource, and adding knowledge does not alter the source file.

Successful upload is also not a guarantee of successful parsing, indexing, retrieval, or interpretation. File condition, format features, service availability, permissions, and resource configuration can all affect what becomes available to an Agent. Inspect the current status and source output before relying on a document in an important answer.

Why it matters

Keeping source files, processed resources, and business guidance distinct helps a reviewer locate a failure and avoids treating a successful upload as proof that the content is ready for analysis.

How DataFact uses these objects

• A database connection stores the information needed to reach a source and discover schemas.
• The Document Library stores and organizes uploaded files.
• A MemoryLake resource turns selected source material into a resource used by DataFact workflows.
• Knowledge adds semantic or business guidance to a Dataset or Analysis Model.

These objects are related but not interchangeable. Uploading a file does not automatically create Documents (Lite) or Documents (Excel). Creating a database connection does not automatically create a Dataset. Adding knowledge does not change source records.

Example

A team can upload a contract to the Document Library, add it to a document resource, wait for indexing, and link the eligible resource to an Agent. Business knowledge can explain an internal term, but the contract remains the source for its signed clause.

Limitations and edge cases

• Scans, merged cells, damaged files, and unsupported features can affect parsing.
• A connected or indexed source may still be unavailable to a user or Agent.
• Retrieval relevance is not the same as answer correctness.

Related concepts

• Documents and databases
• Evidence and citations
• Datasets and Analysis Models
• Knowledge and answer quality