Skip to Content
SourcesAssets & Metadata

Assets & Metadata

Running a scan turns the contents of a source into assets: one normalised record per meaningful item, each carrying the content to be analysed and a set of metadata describing it. Assets are the result of a source execution and the input to every detector.


What an asset is

An asset is a single item extracted from a source — a document, a file, a wiki page, a chat message, a database table, a video. Whatever the system, every asset shares the same core attributes:

AttributeWhat it is
NameA human-readable label for the item.
KindWhat sort of item it is — file, image, page, comment, table, and so on.
External URLA link back to the item in its original system.
LinksRelated links discovered on or around the item.
ContentThe text analysed by detectors — extracted directly, or via OCR / transcription.
MetadataStructured facts about the item (see below).
Change statusHow this item changed since the last scan (see below).

Each asset is identified consistently across scans, so the same item is tracked over time rather than duplicated — that’s what makes findings stable from one run to the next.


Asset change status

Because Classifyre remembers what it saw last time, every asset in a run is marked with how it changed. This is what powers incremental scanning and the automatic resolution of issues that have gone away:

StatusMeaning
NewThe item appeared for the first time.
UpdatedThe item existed before and its content changed.
UnchangedThe item is identical to the last scan.
DeletedThe item is gone from the source; its findings can be auto-resolved.

The mechanics of how a scan diffs against the previous run live in the Flow section.


Asset metadata

Beyond its content, each asset carries metadata — structured facts that make findings easier to understand, filter, and investigate. Metadata comes in two layers.

Shared, content-based metadata

Some metadata depends on the kind of content, no matter which source it came from. These reusable families are attached automatically:

Content familyTypical fields
FileByte size, MIME type, and a parse-error note if extraction failed
ImagePixel width and height
Document (PDF, DOCX)Page count, paragraph count, table count
Spreadsheet / table (CSV, XLSX, Parquet)Row count and column definitions
Audio / videoDuration and transcript details when transcription is on

Source-specific metadata

On top of that, each source adds facts that only make sense for its system — a chat message’s channel and author, a wiki page’s space and version, a video’s view count and upload date, a database row’s table and schema.

Find the exact list. Every source’s reference page has an Extracted Metadata section listing precisely which fields it attaches to each asset kind. Browse the Source Catalog and open any source to see its metadata.


Why metadata matters

Metadata isn’t just description — it does real work downstream:

  • Findings inherit context. A finding points at the asset that produced it, so its name, location, and metadata travel with it into triage and investigation.
  • Better filtering and grouping. Metadata lets you slice findings by source, kind, location, and more.
  • It helps the AI. When a source is silent, the Config and Detector agents read asset metadata to work out what kind of data they’re looking at and which detectors to enable — the “cold start” described in Autopilot.

From assets to findings

Assets are where source execution ends and detection begins. Each asset’s content is handed to the detectors configured on the source, and anything they flag becomes a finding attached to that asset.

That handoff — which detectors run, on what content, and what they produce — is the subject of the Detectors section.


You’ve covered the essentials

Page
How Sources WorkWhat a source is and its journey to findings
Configuration & FieldsRequired, masked, and optional fields
Sampling StrategiesHow much to read, and which items
OCR & TranscriptionReading images, audio, and video
Testing & SchedulingVerify connections and automate scans
Assets & MetadataWhat a scan produces (you are here)
Last updated on