Skip to main content
Knowledge Base is the centralized workspace where you collect every document, note, and integration that an AI agent should rely on. It acts as the agent’s shared memory—organized, searchable, and governed inside the CometChat Dashboard—so responses stay aligned with your domain expertise, policies, and historical conversations.
Knowledge Base

​
Key capabilities

  • Source management – Upload, index, and curate PDFs, DOCX, TXT, and inline text entries in one place.
  • Status tracking – See whether each source is Indexed, Indexing, or Failed before exposing it to users.
  • Granular attachments – Attach or detach individual sources per agent without duplicating content.
  • Flexible ingestion – Mix manual uploads with links or integrations (coming soon) to keep knowledge fresh.

​
Supported source types

  • Files – Batch upload up to 10 PDF, DOCX, or TXT files. Every file is chunked and indexed so agents can cite specific passages.
  • Text – Create lightweight notes or SOPs (up to 5,000 characters) directly in the dashboard—perfect for quick updates.

​
Add a source

1

Open Knowledge Base

Navigate to AI Agent → Agent Builder → Knowledge Base within the CometChat Dashboard, then click + Add Source.
2

Choose the ingestion method

Select from Files, Text, Links (coming soon), or Integrate (coming soon). The drawer updates its fields based on your selection.
3

Provide content and context

Upload files via drag-and-drop, paste text with the formatting toolbar, or enter metadata such as title and description so teammates can identify the source later.
4

Add and index

Click Add. The Knowledge Base queues the source, displays its status, and refreshes the list when indexing finishes.

​
Attach sources to agents

  1. Open the Knowledge Base list once your source shows Indexed.
  2. Locate the source row and toggle Attach to Agent for the agent you are configuring.
  3. Repeat for every agent that needs the same knowledge—no need to re-upload.
  4. To detach, toggle the switch off. The source stays in the library for future use.

​
Monitor source status

StatusMeaningRecommended action
IndexedProcessing succeeded and the content is ready for agents.Attach to the relevant agents or leave it in the library for later.
IndexingThe source is still being chunked and embedded.Wait for completion; large files may take longer.
FailedThe indexer could not process the file or payload.Click the ↻ retry icon, verify the format, or upload a clean copy.
When a source fails repeatedly, delete it (🗑️) to keep the workspace tidy, correct the underlying issue, and re-ingest it. Reviewing statuses regularly ensures agents only pull from healthy data.

​
Use Knowledge Base sources inside conversations

  • Only attached sources become part of the agent’s retrieval context.
  • Attached sources appear in testing and production conversations so you can verify citations.
  • Detaching a source immediately removes it from the agent’s context without deleting the data.
  • You can reuse the same source across multiple agents to keep shared knowledge consistent.

​
Best practices

  • Group related documents before uploading so you can map them to the right agents.
  • Review indexing results after every bulk upload to catch failures early.
  • Archive outdated sources or detach them from agents to prevent stale information from leaking into responses.
  • Document why each source exists using the title and description fields—future teammates will thank you.