Knowledge Base Overview

Create a knowledge base from your content and use it to power accurate AI responses in bots and voice agents.

What is a Knowledge Base?

A knowledge base is a collection of content — web pages, documents, and URLs — that Swiftsell indexes for semantic search. When an AI bot or voice agent needs to answer a question, it retrieves the most relevant passages from the knowledge base and uses them to generate an accurate, grounded response.

This technique is called Retrieval-Augmented Generation (RAG). Instead of relying solely on the LLM’s training data (which may be outdated or inaccurate for your specific products), the agent first searches your content, then synthesises an answer from what it finds.

---

Source types

You can add content to a knowledge base from three sources:

Source type	Description
Domain crawl	Provide a root URL and Swiftsell crawls all pages under that domain
Single URL	Index a single web page
File upload	Upload PDF, DOCX, or TXT files

See Add Sources for step-by-step instructions for each type.

---

Training status

After adding a source, Swiftsell processes and indexes the content. Each source has a training status:

Status	Meaning
Pending	Content is queued for processing
Completed	Content is indexed and available for retrieval
Failed	Processing failed — check the URL is accessible or the file format is supported

Sources with Completed status are immediately available for retrieval. If a source shows Failed, try re-adding it or check that the content is accessible.

---

Using a knowledge base

Once a knowledge base has completed sources, you can attach it to:

• AI Agent block (WhatsApp automation) — the bot uses the knowledge base to answer questions within the AI node. See AI Agent block.
• Voice Agents — the phone agent retrieves relevant content before responding to callers. See Attach a Knowledge Base.

One knowledge base can be attached to multiple bots and voice agents simultaneously.

---

Limits and recommendations

• Domain crawls are depth-limited — very large sites may not have all pages indexed. Use specific URLs for important individual pages if needed.
• File size — large files (>10 MB) may be rejected. Split large documents into smaller files.
• Content quality — well-structured, factual content retrieves better than dense legal text or content with lots of tables and formatting.
• Freshness — content is indexed at the time of upload or crawl. If your website or documents change, re-add the source to re-index.

---

Next steps

Add Sources

Add domains, URLs, and files to start building your knowledge base.

Manage Sources

Delete, re-index, and inspect your knowledge base sources.