Mira AI answers from the knowledge sources you connect — a web scrape, an XML product feed, text content, and custom responses. When a source fails, Mira may stop answering product questions or reply with outdated information. This guide covers the most common errors and how to fix them. You can see each source's state in the Knowledge base section — a status icon next to the source name shows details when you hover over it.
Product feed requirements
The feed URL must be publicly accessible (no login required). If your server uses a firewall, allow our IP addresses 94.130.187.58 and 188.245.65.74.
Maximum size is 50 MB, in XML or CSV format (Google Merchant Center or Heureka structure).
Every item needs the required tags: id, name, description, price, and link. Items missing any of them are skipped during processing.
Every item should have a category assigned — items without one can be dropped, and Mira then can't answer product and category questions.
Tip: Before adding your feed, run it through a feed validator to catch missing tags early. When a feed keeps failing, the validator output usually shows exactly which items are the problem. You can validate your feed here.
Common product feed errors
"Failed to process data" / the upload never finishes
The usual causes: the feed is larger than 50 MB, required tags are missing or malformed, or the URL isn't publicly accessible. Validate the feed, check its size, and make sure the URL opens in an incognito browser window without logging in. If your server filters traffic, allow our IP addresses listed above.
"The time limit for downloading the file expired"
Your feed hosting responded too slowly and the download timed out. Try adding the feed again — the check runs anew with every attempt — and verify your server's connection speed. If the error keeps coming back, contact our support and we'll investigate.
Warning about missing categories
Items without a category are skipped, and Mira can't answer product or category questions from them. First add a category to every item in the feed. If the feed's first build ends with a category warning even though your items do have categories, open the feed and run Update — the second build usually completes successfully. If not, contact our support.
The feed is Ready, but Mira can't find some products
Items missing a required tag (id, name, description, price, link) are skipped silently, so the feed can be Ready while some products aren't in it. Validate the feed, fix the flagged items, and run Update.
An update is stuck on Pending
Feed builds can take a while for large catalogs. If an update stays Pending for more than a few hours, contact our support — we can restart the processing for you.
Web scrape issues
Mira can't find pages of my website
Check that the pages are publicly accessible without logging in, and that your firewall allows our IP addresses (94.130.187.58, 188.245.65.74). One scrape covers up to 20 pages — pick the key ones (FAQ, shipping, terms, main product pages) rather than everything.
Scraping runs for hours
A scrape that runs for hours is stuck. Contact our support with the URL you're trying to scan and we'll check the processing on our side.
Mira answers with outdated information
Sources refresh on their Update interval — you can see and change it directly in the sources table (web scrape: every month or every 2 weeks; product feed: up to twice a day). After changing your website or catalog, either wait for the next automatic update or open the source and run Update manually.
Publishing and source states
A single failed source doesn't block publishing your AI agent. Publishing is blocked only when no web scrape is ready, or while selected sources are still being created or updated. If the Publish button is disabled, check the Knowledge base section for sources still in progress.
What's next?
How to start train Mira AI — manage knowledge sources and custom responses.
How to export product feeds — get a feed URL from your e-commerce platform.
Introduction to Mira AI Shopping Assistant — full setup from scratch.