Manage Canon Documents
3 min read
The document table is the state of your Canon. Its job is to make one thing obvious: which documents are actually grounding AI output, and which only look like they are.
The Document Table
Section titled “The Document Table”| Column | What it shows |
|---|---|
| Document | The document title. A failed document shows the parse error underneath, verbatim |
| Status | Where the document is in the ingestion pipeline — Pending, Extracting, Embedding, Ready or Failed |
| Type | The source file type, for example PDF or DOCX |
| Passages | How many searchable passages the document produced. Shown as — until the document is Ready |
| Scope | Project or Workspace — see Scope and permissions |
| Added | When the document was admitted to Canon |
Once a Canon holds more than six documents, a Filter documents box appears above the table.
A document with a low passage count relative to its size is worth opening. It usually means the extractor found less text than you expected — a PDF that is mostly diagrams, or a spreadsheet with one populated sheet out of five.
Re-index a Document
Section titled “Re-index a Document”Canon indexes a document once, when you add it. If the underlying file changes, the index does not change with it.
Select the Re-index action on the row to read the file again and rebuild its passages. The document returns to Pending and works back through the pipeline. Its old passages are replaced wholesale, so there is no risk of stale and fresh text both being retrievable.
Re-index after:
- Replacing the source file with a newer revision
- A failed ingest whose cause you have fixed
- A document that produced far fewer passages than its length suggests
Retry Failed Documents
Section titled “Retry Failed Documents”When one or more documents have failed, a Retry failed documents action appears in the action rail with a count. It re-indexes every failed document in one go.
This is worth trying once after a failure that looks transient. For a parse failure — an unsupported type, a scanned PDF, a corrupt file — retrying will fail again in the same way. Fix the source first.
Failure Codes
Section titled “Failure Codes”A failed document shows its reason on the row, exactly as the extractor wrote it.
| Reason | What it means | What to do |
|---|---|---|
| Unsupported file type | The extension is not one Canon can read | Convert the document to PDF, DOCX or Markdown and add it again |
| Empty file | The file has no content | Check you uploaded the intended file |
| Corrupt file | The document could not be opened | Re-export it from its source application, then re-index |
| Scanned PDF unsupported | The PDF has no text layer — it is images of pages | Re-export a text PDF from the original document |
Remove a Document
Section titled “Remove a Document”Select Remove from Canon on the row. The document leaves Canon and its index is deleted, so it stops grounding AI output immediately.
The file itself is not deleted — it stays in the project’s file library, and anything else referencing it is unaffected.
Remove a document when it is superseded, wrong, or was never authoritative in the first place. A stale spec in Canon is worse than no spec: the AI will cite it, and a reviewer will believe it.