Documentation Knowledge Search & Retrieval

Search & Retrieval

Learn how your chatbot finds the right information from your knowledge base to answer user questions.

Overview

When a user asks a question, your chatbot searches your knowledge collections to find relevant information. The system uses intelligent search technology that understands meaning, not just keywords, to find the best answers.

How It Works

The system finds relevant information even when users phrase questions differently than your documentation. For example, asking "How do I get my money back?" will find information about refunds, returns, and reimbursements.

Search Modes

The system offers three ways to search your knowledge base:

Hybrid (Default)

Combines keyword matching with semantic understanding for the best results.

Recommended for most use cases

Semantic

Finds content by meaning, great for natural language questions.

Best for: Conceptual queries

Keyword

Traditional search by exact terms and phrases.

Best for: Specific terms, IDs

When to Use Each Mode

Hybrid Search

Use for general Q&A, mixed query types, and when you want the most reliable results. This is the default and works well in most situations.

Semantic Search

Use when users ask questions in their own words and might not use your documentation's terminology. Good for "How do I..." or "What is..." questions.

Keyword Search

Use when searching for specific product names, error codes, or technical terms where exact matching is important.

Configuration Options

You can adjust these settings in your RAG nodes to fine-tune search behavior:

Number of Results

How many relevant chunks to retrieve from your knowledge base.

  • Default: 10 chunks
  • Recommended: 5-20 chunks depending on your use case
  • More results give the AI more context but can slow down response generation and increase costs

Quality Threshold

Filters out low-quality content from search results.

  • Default: 0.5 (moderate quality filter)
  • Higher values (0.7-0.9): Only return high-quality, detailed content
  • Lower values (0.3-0.5): Include more results, even if less detailed

Start with Defaults

The default settings work well for most use cases. Only adjust if you're experiencing specific issues like too few results or irrelevant matches.

Testing Your Search

The best way to ensure good search results is to test with real questions:

  1. Think about common questions your users will ask
  2. Test those questions using the Knowledge Agent or RAG nodes in your flow
  3. Review which content is being retrieved
  4. Adjust your settings or improve your content organization if needed

Tip: Content Quality Matters

Well-organized, clearly written documents will always perform better in search than poorly structured content. Focus on content quality first before adjusting search settings.

Improving Search Results

If your chatbot isn't finding the right information, here are ways to improve search performance:

Agentic Optimization

Enable AI-powered query optimization for complex or ambiguous questions. When enabled, the system:

  • Analyzes the user's question to understand intent
  • Reformulates queries for better search results
  • Can perform multiple searches to gather comprehensive information
  • Evaluates if retrieved information is sufficient

Best For

  • Vague or unclear questions
  • Complex multi-part queries
  • When users don't know the right terminology
  • Questions requiring context from conversation

Trade-off

Agentic optimization takes a few extra seconds and costs more (multiple AI calls) but provides more accurate results for complex questions. For simple questions, standard search is usually sufficient.

Content Organization Tips

Better content organization leads to better search results:

  • Use clear headings: Well-structured documents with descriptive headings help the system understand content
  • Separate topics: Create different collections for different subject areas
  • Avoid duplicates: Duplicate content can confuse the search system
  • Include context: Make sure information is understandable on its own, not just in context of a full document

Content Quality is Key

100 well-organized, clearly written documents will outperform 1,000 poorly structured ones. Focus on improving your content before adjusting technical settings.