Skip to main content

Knowledge Bases

Knowledge Bases are a powerful feature of the NeuralTrust platform that allow you to create and manage specialized knowledge repositories for AI model testing and evaluation. They provide a structured way to organize domain-specific information that can be used to enhance your model testing capabilities.

info

Knowledge bases typically serve as vector databases used for Retrieval-Augmented Generation (RAG).

With Knowledge Bases, you can:

  • Create specialized knowledge repositories for different domains
  • Seed knowledge bases with initial topics and data
  • Use knowledge bases to generate more targeted and relevant test cases
  • Manage credentials and access controls for different knowledge sources
  • Integrate with various data sources and knowledge systems

Knowledge Bases are particularly useful for:

  • Domain-specific testing of AI models
  • Generating contextually relevant test cases
  • Maintaining consistent testing knowledge across teams
  • Supporting specialized compliance and security testing scenarios

Supported Knowledge Base Types​

NeuralTrust currently supports the following knowledge base types:

  • Upstash - Vector database for real-time data storage and retrieval
  • Azure AI Search - Cognitive search service for fast and sophisticated data indexing
  • Company Documents - Support for PDF documents and other company materials
    • Automatically indexes and processes PDF content
    • Maintains document hierarchy and relationships
    • Extracts key information and metadata

Knowledge Base API Methods​

from neuraltrust import NeuralTrustApi

client = NeuralTrustApi(api_key="YOUR_API_KEY")

# Create a new knowledge base
client.knowledge_base.create(
    type="your_kb_type",  # "upstash", "azure_search", or "document"
    credentials={"key": "value"},
    seed_topics=["topic1", "topic2"]
)

# Get a specific knowledge base
client.knowledge_base.get(id="kb_123")

# Delete a knowledge base
client.knowledge_base.delete(id="kb_123")

For more information on how to use Knowledge Bases, please refer to the Knowledge Bases documentation.