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:
| Attribute | What it is |
|---|---|
| Name | A human-readable label for the item. |
| Kind | What sort of item it is — file, image, page, comment, table, and so on. |
| External URL | A link back to the item in its original system. |
| Links | Related links discovered on or around the item. |
| Content | The text analysed by detectors — extracted directly, or via OCR / transcription. |
| Metadata | Structured facts about the item (see below). |
| Change status | How 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:
| Status | Meaning |
|---|---|
| New | The item appeared for the first time. |
| Updated | The item existed before and its content changed. |
| Unchanged | The item is identical to the last scan. |
| Deleted | The 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 family | Typical fields |
|---|---|
| File | Byte size, MIME type, and a parse-error note if extraction failed |
| Image | Pixel width and height |
| Document (PDF, DOCX) | Page count, paragraph count, table count |
| Spreadsheet / table (CSV, XLSX, Parquet) | Row count and column definitions |
| Audio / video | Duration 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 Work | What a source is and its journey to findings |
| Configuration & Fields | Required, masked, and optional fields |
| Sampling Strategies | How much to read, and which items |
| OCR & Transcription | Reading images, audio, and video |
| Testing & Scheduling | Verify connections and automate scans |
| Assets & Metadata | What a scan produces (you are here) |