Data sources
Connect what your customers already have
Files, links, drives, databases. They connect it in your product; we keep it indexed.
Sources
Five ways data gets in.
Upload
PDFDOCXTXTMDCSVXLSXPPTXHTMLImages · OCR
Web
Single URLSitemap crawlScheduled recrawl
Cloud storage
Google DriveDropboxOneDriveS3
Apps
NotionConfluenceZendeskSlack · soon
Connected through the same integration layer your agents use for tools.
Direct
API pushDatabase connection · soon
Formats
What each format does.
| Format | Max size | OCR | Notes |
|---|---|---|---|
| 50 MB | If scanned | Text PDFs parsed directly; scans routed through OCR | |
| DOCX | 25 MB | No | Headings and tables preserved as chunk boundaries |
| TXT / MD | 10 MB | No | Markdown structure used for chunking |
| CSV | 25 MB | No | Row-wise chunking; header retained on each chunk |
| XLSX | 25 MB | No | One pass per sheet; formulas read as values |
| PPTX | 25 MB | No | Slide text and speaker notes |
| HTML | 10 MB | No | Navigation and boilerplate stripped before chunking |
| PNG / JPG | 10 MB | Yes | OCR quality depends on the source scan |
Limits are the current defaults and are raised per platform account on enterprise agreements. Confirm anything you intend to design against.
Sync
What refreshes itself, and what doesn’t.
| Source | Syncs | Schedule | Deletions |
|---|---|---|---|
| Uploaded files | No | Replaced on re-upload | Removed from the index immediately |
| Web URL / sitemap | Yes | Configurable — daily by default | Removed pages dropped on the next crawl |
| Google Drive / Dropbox / OneDrive | Yes | Configurable — hourly to daily | Propagate on the next sync |
| S3 | Yes | Configurable | Propagate on the next sync |
| Notion / Confluence / Zendesk | Yes | Configurable — daily by default | Propagate on the next sync |
| API push | No | You control timing | Explicit delete call |
The upload UI
A document manager you don’t have to build.
Your users upload, organise and delete their own documents and watch them index — inside your product, in your branding. This ships with the embed.
Drop files here, or connect a source
refund-policy.pdf41 pages · 2.1 MBIndexed
shipping-terms.md6 pages · 48 KBIndexed
faq-export.csv1,204 rows · 890 KBIndexing 62%
support.deskly.comSitemap · 214 pagesSynced daily
handbook-scan.pdf88 pages · OCRFailed — retry
1.4 GB of 2 GB used
Limits
The numbers, plainly.
- Storage per customer — 2 GB default, set by you per customer
- File size — 50 MB for PDF, 25 MB for office formats, 10 MB for text and images
- Ingest rate — 100 documents per minute per customer
- Documents per customer — no hard cap; storage is the binding constraint
- Crawl depth — 500 pages per sitemap by default
Enterprise agreements raise these. What is written here is what applies by default.
Custom sources
Anything not on the list.
Push arbitrary content through the API and it goes through the same pipeline as an upload — parsed, chunked, embedded, indexed to that customer.
POST /v1/knowledge/documents
{
"title":"<Your Document Title>",
"description": "<Your Document Description>",
"settings": { "strategy": "recursive", "chunkSize": 4000 },
"collection_details": "<Your Collection Id>",
"url": "<Your Document URL>"
}