Supported file types & limits
Indexed document types
Section titled “Indexed document types”During ingestion, Insight AI extracts and indexes text from these file types:
| Type | Extensions |
|---|---|
.pdf |
|
| Microsoft Word | .doc, .docx |
| Microsoft PowerPoint | .ppt, .pptx |
| Plain text | .txt |
| Markdown | .md |
| CSV | .csv |
Files with any other extension — including images, spreadsheets (.xlsx),
audio, video, and archives — are skipped without stopping the run.
Automatic skipping rules
Section titled “Automatic skipping rules”To keep your index clean, ingestion also skips:
- Files larger than 50 MB
- Hidden files (names starting with
.) - System files such as
thumbs.db,desktop.ini,.DS_Store,System Volume Information, and$Recycle.Bin - Office temporary/lock files (
~$…,.tmp,.temp)
How documents become searchable
Section titled “How documents become searchable”Each supported file’s text is extracted and split into overlapping passages of roughly a thousand characters. Every passage is converted into a semantic vector and stored in the target collection. When you ask a question in chat, your question is matched against these passages by meaning — so “vacation policy” can find a document that says “annual leave guidelines.”
A document that contains no extractable text (for example, a scanned PDF that is just images) will cause the ingestion run to report an error for that file.