Knowledge Bases
Knowledge Bases

Knowledge Bases

Ground AI Responses in Your Data

RAG-Powered Knowledge

Knowledge Bases use Retrieval-Augmented Generation (RAG) to ground your AI agent's responses in your actual business data. Instead of relying on the AI's general training data, the agent retrieves relevant information from your uploaded documents before generating each response, ensuring accuracy and reducing hallucinations.

1. Creating a Knowledge Base

To create a knowledge base in the Flowk Dashboard:

  1. Navigate to Knowledge Bases in the sidebar.
  2. Click Create Knowledge Base and give it a descriptive name (e.g., "Product Documentation", "Support Policies").
  3. Optionally add a description to help your team understand what this knowledge base contains.
  4. Configure the chunking settings (see section 3 below).
  5. Create the knowledge base — it will be empty until you upload documents.

2. Supported Document Types

Flowk supports the following document formats for knowledge bases:

  • PDF — Ideal for product manuals, whitepapers, and formal documentation.
  • Markdown (.md) — Great for internal wikis, runbooks, and developer documentation.
  • Plain Text (.txt) — Simple text files for notes and lists.
  • CSV — Structured data like product catalogs, pricing tables, and FAQ lists.
  • HTML — Web pages and help center articles.
  • DOCX — Microsoft Word documents.

Documents are automatically processed and indexed when uploaded. Processing time depends on file size and complexity, typically ranging from a few seconds to a few minutes.

3. Connecting to an Agent

Once your knowledge base has documents indexed, you can connect it to one or more agents:

  1. Navigate to the AI Studio and select the agent you want to connect.
  2. In the Virtual Builder, add Knowledge Base Search Tool, in the tool's configuration sidebar, select your Knowledge base you want to connect to.
  3. The tool you just added to the canvas, connect to it to guildeline(s) you want to use the tool.
  4. Republish your agent. The agent will now retrieve relevant information from these knowledge bases when responding to customer inquiries that matches the guideline it's attached to.

You can connect multiple knowledge bases to a single agent. The agent will search across all connected knowledge bases to find the most relevant information.

5. Search & Test

You can test your knowledge base directly from the Virtual Builder in the AI Studio:

  • Test with Agent — Open the conversation chat widget and ask questions that should be answered from knowledge base content.

6. Best Practices

  • Keep documents focused — Each knowledge base should cover a specific domain. Avoid mixing unrelated content in a single knowledge base.
  • Use descriptive file names — File names are used as source citations when the agent references information.
  • Update regularly — Keep your knowledge bases current. Outdated information leads to incorrect responses.
  • Monitor retrieval quality — Review which documents are being retrieved most often and ensure they are accurate and well-structured.
  • Test before deploying — Always test your knowledge base with sample questions in the simulator before going live.

7. Managing Knowledge Bases

From the Knowledge Bases page, you can:

  • View all knowledge bases and their document counts.
  • Upload new documents to existing knowledge bases.
  • Remove documents and re-index updated versions.
  • Monitor indexing status and troubleshoot failed documents.
  • Delete knowledge bases (this will not affect connected agents, but the agent will lose access to that data).

Build Your Agent

Learn how to create guideline-driven agents that use your knowledge bases.

Agent Builder