Skip to main content

Knowledge Retrieval Action

The Knowledge Retrieval Action searches through your Knowledge Bases to find relevant information based on a query.

What does this integration do?​

This action allows workflows to search through configured Knowledge Bases using semantic search. It retrieves relevant documents, chunks, or information that match the provided query, making it perfect for building AI assistants that need access to your company's knowledge.

Typical Use Cases:

  • AI Assistant Support: Provide context for LLM actions
  • Information Lookup: Find specific documents or data
  • Content Research: Gather information for analysis
  • Question Answering: Build knowledge-based chatbots

User Configuration​

Search Query

  • Purpose: The text to search for in the Knowledge Bases
  • Source: Variable from previous workflow steps or direct text
  • Example: "How to reset user passwords"

Knowledge Bases

  • Purpose: Select which Knowledge Bases to search
  • Multiple Selection: Can search across multiple Knowledge Bases
  • Requirement: At least one Knowledge Base must be selected

Result Count (Optional)

  • Purpose: Maximum number of results to return
  • Default: 5 results
  • Range: 1-20 results

Output Variable

  • Purpose: Variable name to store the search results
  • Contains: Array of relevant documents with metadata
  • Example Variable: searchResults

How it Works​

The Knowledge Retrieval Action performs semantic search across your selected Knowledge Bases and returns the most relevant information chunks with confidence scores and metadata.

Workflow Integration​

Search Results Format​

[
  {
    "content": "To reset a user password, navigate to Admin > Users...",
    "source": "admin-manual.pdf",
    "score": 0.92,
    "metadata": {
      "page": 15,
      "section": "User Management"
    }
  },
  {
    "content": "Password requirements include minimum 8 characters...",
    "source": "security-policy.md",
    "score": 0.87,
    "metadata": {
      "category": "Security"
    }
  }
]

Usage with LLM Actions​

1. Knowledge Retrieval Action
   Query: userQuestion
   Output: relevantInfo

2. LLM Action
   Prompt: "Based on this information: {{relevantInfo}}, answer: {{userQuestion}}"

Technical Details​

This action leverages vector embeddings to find semantically similar content in your Knowledge Bases, providing accurate and contextual information retrieval.