Adding content
Everything Minaya can answer comes from your knowledge base. There are three source types, and you can mix them freely.
Source types
| Type | Best for | Notes |
|---|---|---|
url | Your website, docs, help centre | Crawls linked pages. Can be re-crawled in place. |
pdf | Manuals, price lists, brochures | Text is extracted; scanned images without a text layer will not work. |
text | Policies, FAQs, anything not yet published | Pasted directly. Fastest way to correct a specific answer. |
Adding a URL
Enter a starting URL and Minaya crawls it plus the pages it links to. See Crawling a site for depth, limits, and what is skipped.
Uploading a PDF
Upload the file and the text layer is extracted, chunked, and embedded. A PDF produced by scanning paper has no text layer — run OCR first, or paste the content as a text source instead.
Pasting text
The most direct way to fix a wrong or missing answer. If visitors keep asking something your site does not state plainly, paste a short, explicit note:
Returns: 30 days from delivery, unworn, with the original packaging.
Refunds are issued to the original payment method within 5 working days.
Exchanges are handled by contacting support@example.com.Write it the way a person would ask about it. Retrieval matches on meaning, so phrasing that mirrors real questions retrieves more reliably.
Ingestion status
| Status | Meaning |
|---|---|
pending | Queued, waiting for the worker. |
processing | Being fetched, chunked, and embedded. |
ready | Searchable now. |
failed | Something went wrong; the error is shown on the source. |
Stuck at
pending? The ingestion worker is a separate process from the API and must be running for any source to be processed.
Re-crawling
URL sources have a re-crawl button. It deletes the stored chunks and fetches the site again — use it when your content changes. Chunks are removed before the new crawl starts, so there is a brief window with no answers from that source rather than answers from stale content.
Text and PDF sources cannot be re-crawled; there is nothing remote to re-fetch. Delete and re-add them instead.
Plan limits
Each plan caps how many sources you can keep. Removing a source deletes its chunks immediately and frees a slot.