Knowledge Base

What is the Knowledge Base?

The Knowledge Base stores your company-specific documents so the AI voice agent can answer questions accurately. Documents are chunked, embedded, and indexed for semantic search (RAG — Retrieval Augmented Generation).

Supported document types

• faq
• manual
• policy
• pricing
• call summary
• job summary
• technician note

How to upload documents

• Go to Knowledge Base in your dashboard.
• Click "Add document" and provide a title and content.
• Select the appropriate source type (FAQ, manual, policy, etc.).
• Submit — the document enters processing and indexing.
• Status changes from "processing" to "active" when ready.

How search test works

Use the Search Test tab to query your indexed knowledge without making a phone call. Enter a question and review matching chunks with relevance scores. This helps verify your content is indexed correctly before customers call.

How RAG works (simple explanation)

When a caller asks a question, the system searches your indexed documents for the most relevant passages. Those passages are provided as context to the AI, which generates a natural-language answer grounded in your company knowledge — not generic responses.

Do not upload secrets

• Never upload API keys, passwords, or credentials.
• Do not include customer PII (names, addresses, SSNs) in knowledge documents.
• Avoid internal-only credentials or webhook signing secrets.
• Use policy and FAQ content intended for customer-facing answers.

Troubleshooting failed documents

• Check the Failed / Recovery tab for documents with "failed" status.
• Click "Retry index" to re-process a failed document.
• Ensure content is not empty and source type is valid.
• Delete and re-upload if retry does not resolve the issue.
• Contact support if failures persist across multiple documents.