Skip to content
Learning Outcomes
  • Create a Knowledge Builder assistant.
  • Upload knowledge sources and understand ingestion status.
  • Use a Knowledge Builder assistant from AI Insight.

Using Knowledge Builder

Creating a Knowledge Builder assistant

  1. Navigate to Cortex → Knowledge Builder
  2. Click Add Assistant

  3. Complete the form:

  4. Integration Token: The AI provider token used to list models and ingest (embed) your knowledge sources.

  5. Assistant Name: A descriptive name (for example “Lab SOP Assistant”).
  6. Foundation Model: The model this assistant will use when answering questions.
  7. Prompts & Instructions: Describe the assistant’s role and how it should answer.
  8. Knowledge Sources (optional): Upload up to 20 files (PDF/MD/TXT/HTML), 20 MB max each.
  9. Set Permissions: Choose who can use the assistant.

  10. Group Permission to Write: Choose which groups have write access (for example, to delete the assistant, or manage sources where available).

  11. Click Create Assistant

After creation, ingestion starts in the background. You can continue using Foundry Connect while files process.

Understanding assistant status

In the Knowledge Builder table, each assistant shows an overall status:

  • Processing: Files are being ingested (text extraction + embedding generation).
  • Ready: At least one knowledge source is ready and the assistant can be used from AI Insight.
  • Failed: Ingestion failed. Check the Error column for details.
Sources tooltip

The Knowledge Sources count includes a tooltip that lists the uploaded filenames.

Using your Knowledge Builder assistant from AI Insight

Once an assistant is Ready, you can use it inside AI Insight.

  1. Open AI Insight (the AI Insight button in the top header)
  2. Select an Integration Token and AI Model
  3. (Optional) In Select Knowledge Builder, choose your Knowledge Builder assistant
  4. Ask your question as usual

When a Knowledge Builder assistant is selected, Foundry Connect will:

  1. Retrieve the most relevant sections from your uploaded sources
  2. Use those sections (plus your assistant’s instructions) to generate the answer
Model compatibility

Knowledge Builder assistants are tied to the Foundation Model you selected at creation time. If AI Insight is using an integration token that doesn’t provide that model, the assistant may appear disabled with a “requires ” label. Switch to a compatible token/model and try again.

Example questions

  • “Based on our SOP, what are the acceptable RNA integrity thresholds?”
  • “Summarize the key steps for running the preprocessing workflow.”
  • “What does our internal guideline say about handling low-quality samples?”