Skip to content
Learning Outcomes
  • Troubleshoot common Knowledge Builder access and ingestion issues.
  • Understand why a Knowledge Builder assistant may be unavailable in AI Insight.

Troubleshooting Knowledge Builder

Knowledge Builder isn’t visible in the UI

Possible causes:

  • Knowledge Builder is not enabled for your account
  • You do not have access to Cortex

Solutions:

  1. Contact your administrator to confirm the feature is enabled (for example, the FF_MODEL_SMITH feature flag).
  2. Refresh the page after access is granted.

“Create an integration token…” / no token options available

Cause: You don’t have an AI provider integration token configured.

Solution: Create an AI integration token under Account → Profile → Integration Tokens.

See Integration Tokens.

Assistant stays in “Processing”

Possible causes:

  • Large files (especially PDFs/HTML with charts/images)
  • Provider rate limits or temporary outages

Solutions:

  • Wait a few minutes and refresh the Knowledge Builder page.
  • Try fewer/smaller sources or prefer MD/TXT.
  • Verify your AI provider account has available quota.

Assistant status is “Failed”

Cause: Ingestion failed for one or more sources.

What to check:

  • Review the Error column for details.
  • Confirm your files are within limits (20 MB each; supported extensions).

Common fixes:

  • Convert scanned PDFs to OCR’d text PDFs, or export to TXT/MD.
  • Remove extremely large embedded images or use a text-first version of the document.
  • Retry with a different integration token (if your provider is rate-limiting).

Knowledge Builder assistant does not appear in AI Insight

Possible causes:

  • The assistant is not Ready
  • The assistant has no ready sources
  • You don’t have permission to view the assistant

Solutions:

  1. Confirm the assistant status is Ready in Cortex → Knowledge Builder.
  2. If the assistant is Failed, fix the ingestion error first.
  3. Confirm your permissions with the assistant owner/admin.

Knowledge Builder assistant is disabled

Cause: The assistant was created with a foundation model that isn’t available for the currently selected AI Insight integration token.

Solutions:

  • In AI Insight, switch to an integration token that provides the required model, then re-select the assistant.
  • If you need broad availability, create assistants using a model that is available across your commonly used tokens.

Claude token warning about “embeddings provider”

If you’re using a Claude (Anthropic) token, you may see a warning that an embeddings provider token is missing. Knowledge Builder retrieval depends on embeddings.

Solution: Edit the Claude token and select an OpenAI or Gemini token as the embeddings provider, then try again.