How to Configure Chunking in FastLinePro Knowledge Base?

Chunking is the process of splitting a document into fragments (chunks) for search. Proper chunking configuration ensures better search accuracy and token usage efficiency.

When Needed

  • You need to configure file processing parameters after uploading to the knowledge base.
  • You need to optimize information search in the knowledge base.
  • You need to configure a balance between search accuracy and token costs.

What's Important to Know

  • Segment identifier — separator used in the file to separate topics (for example, --- or !--).
  • Hybrid search — enhanced search mode that combines vector and full-text analysis to improve accuracy.
  • Top K — parameter that determines the number of chunks the AI agent considers when formulating a response.
  • The larger Top K, the more chance to find an answer, but the more tokens will be used.

Prerequisites

  • You are logged in with integrator or administrator rights.
  • You have navigated to the FastLinePro module via Apps -> FastLinePro.
  • You have uploaded a file to the knowledge base and proceeded to chunking configuration.

Step-by-Step Instructions

  1. In the FastLinePro module, go to the "Knowledge Base" section.
  2. Select the knowledge base for which you need to configure chunking, or create a new one.
  3. After uploading the file, the chunking configuration screen opens.
  4. In the "Segment Identifier" field, specify the separator used in your file:
    • Most commonly used is --- (three dashes)
    • You can use other separators, for example !-- or ***
    • The separator must exactly match what is used in the file
  5. Enable the "Hybrid Search" option (recommended to keep enabled):
    • Hybrid search combines vector and full-text analysis
    • Improves search accuracy for relevant information
    • Recommended to keep enabled for better results
  6. In the "Top K" field, specify the number of chunks the AI agent considers when formulating a response:
    • Recommended value: 2–3
    • The larger the value, the more chance to find an answer
    • The larger the value, the more tokens will be used
    • For simple questions, 2 is enough, for complex ones — 3–4
  7. Click the "Next" or "Save" button to save settings.

Configuration Recommendations

Segment Identifier

  • Use --- as the standard separator for best compatibility.
  • Make sure the separator exactly matches what is used in the file.
  • If the file has no separators, the system will automatically split the document into fragments.
  • Recommended: always enabled
  • Combines advantages of vector and full-text search
  • Ensures better search accuracy for relevant information

Top K

  • For simple questions: 2
  • For complex questions: 3–4
  • For very complex questions: 4–5 (increases token costs)

What Happens After

After saving settings, the system processes the file, splitting it into chunks according to the specified parameters. The AI agent will be able to use these chunks to search for relevant information and formulate answers.

How to Verify It Worked

  • Check that the segment identifier matches the separator in the file.
  • Make sure Hybrid Search is enabled (recommended).
  • Check that Top K is set within reasonable limits (2–3).
  • Test the search to make sure chunking works correctly.