Documentation
Training & Knowledge Base
The way you structure your files and the content inside them is critical for chatbot performance and accuracy. Proper organization ensures the AI can retrieve exactly the information it needs, minimizing irrelevant or incorrect responses.
Use Separate Files for Each Topic
Maintaining a dedicated file for each topic ensures that the AI can retrieve information efficiently and reduces the likelihood of irrelevant or incomplete answers.
Best Practices:
One topic per file
Descriptive, consistent file names
Avoid combining unrelated subjects in a single document
Example:
Clear file separation improves retrieval precision and simplifies maintenance.
Organize Content Inside Each File
Structured content improves the AI’s ability to understand context, which enhances retrieval accuracy and response relevance.
Best Practices:
Use clear headings and subheadings
H1: Main topic
H2: Subsections
H3: Detailed steps or explanations
Break text into concise paragraphs
Each paragraph should cover a single concept or idea
Avoid huge blocks of text
Incorporate lists, tables, and highlights
Bullet points or numbered lists improve readability
Tables or bolded text emphasize critical information
Provide brief introductions for each section
Helps the AI understand context within a chunk
Example File Structure:
Each heading and section represents a logical unit that can become an independently retrievable chunk.
Splitting Chunks
A chunk is a coherent piece of text extracted from a file, typically 200–500 words, that the AI can index and retrieve independently.
Best Practices:
By headings: Each heading and its associated text becomes a chunk
By paragraphs: For unstructured documents, group related paragraphs into a single chunk
By topic: Ensure each chunk represents a single concept or instruction
Example Chunks from Installation_Instructions.pdf:
Each chunk is focused on a single concept, enabling precise retrieval by the AI.