Knowledge Problems

Overview

The Asiyst Knowledge Engine uses dense vector embeddings and hybrid keyword search to ground assistant replies strictly in your verified documentation.

Assistant Hallucinating or Missing Context

Cause

1. The relevant page or document has not finished vector indexing.

1. The search similarity threshold is set too strictly, filtering out valid passages.

1. The content is formatted in complex multi-nested tables or image-only PDFs without OCR.

Resolution

1. Check Knowledge > Sources > Status in the dashboard to confirm all sources are Indexed (Ready).

1. Re-index the specific source if recent edits were published.

1. Add explicit FAQ entries or Markdown definitions for specialized product terms.

Website Crawler Blocked or Timed Out

If the automatic website crawler cannot index your site:

1. Check robots.txt: Ensure your robots.txt allows the user agent AsiystBot:

      User-agent: AsiystBot
   
      Allow: /docs/
      Allow: /help/

1. Cloudflare / WAF Rules: If you use Cloudflare, add an allow rule for Asiyst crawler IP ranges.

1. SPA Rendering: For JavaScript-rendered SPAs, enable Prerender / Headless Crawl in Knowledge Source settings.

Document Parsing Failures (PDF / DOCX)

• Scanned PDFs: Ensure PDFs contain selectable text. Scanned images without OCR cannot be embedded.

• File size limits: Max file size is 50MB per document on Pro plans.

• Password protection: Remove password encryption before uploading.

Vector Indexing Stuck in Processing

If a source remains in Processing for more than 10 minutes:

1. Click Re-sync Source in the dashboard.

1. Check for circular redirect loops in your sitemap or crawled URLs.