Skip to main content

Knowledge Discovery

Welcome to Knowledge Discovery - a powerful AI-driven platform that enables applications to search documents intelligently, extract insights, and deliver accurate answers using advanced AI capabilities.

What is Knowledge Discovery?​

Knowledge Discovery enables organizations to build intelligent systems capable of understanding and extracting insights from their document repositories. The platform combines:

  • 🤖 Intelligent Agents - Configure AI agents with specific instructions, data sources, and filtering rules
  • 🔍 Semantic Search - Find relevant information using advanced embedding models and vector search
  • 💬 Question & Answer - Get accurate answers to natural language questions backed by your documents
  • 🎯 Contextual Filtering - Apply static and dynamic filters to control data access and relevance
  • 🔒 Access Control - Manage permissions at the user and group level for secure information retrieval

Key Features​

Agent Management​

  • Create and configure AI agents tailored to specific use cases
  • Support for multiple LLM models (Amazon Nova, Claude, Llama4, and more)
  • Custom instructions
  • Version control for agent configurations
  • Avatar customization with pre-signed URLs

Advanced Filtering​

  • Static Filters - Permanent filters applied to all agent queries (e.g., department, document type)
  • Dynamic Filters - Runtime filters provided with each question (e.g., date ranges, user context)
  • Complex logical operators (And, Or) and comparison operators

Flexible Integration​

  • RESTful APIs for agent management and question answering
  • Asynchronous question processing
  • Multiple data source integration

Architecture Overview​

Component Overview​

  1. Client Application - Your application that integrates with Knowledge Discovery APIs
  2. Agent API - Manage agent configurations, access rights, data source mappings, and filtering rules
  3. QnA API - Submit questions and retrieve answers with source citations and feedback
  4. Agent Configuration Store - Persistent storage for agent definitions and settings
  5. Agent Platform - Orchestrates agent execution, retrieves relevant document chunks using vector similarity, and coordinates LLM processing
  6. Content Lake - Document repository containing indexed content for semantic search
  7. QnA Store - Tracks questions, answers, and user feedback for analytics and improvement

Use Cases​

Enterprise Knowledge Base​

Build an intelligent knowledge base where employees can ask questions and get instant answers from company documentation, policies, and procedures.

Customer Support Automation​

Create specialized agents that help support teams find relevant information quickly, improving response times and accuracy.

Document Intelligence​

Enable users to search across large document repositories using natural language, with results filtered based on user permissions and context.

Configure agents with specific data sources and filters to help teams find relevant compliance documents, regulations, or legal precedents.

Quick Navigation​

🚀 Getting Started

New to Knowledge Discovery? Start here to set up your environment and make your first API call.

Get Started

🔐 Authentication

Learn how to authenticate with OAuth 2.0 and secure your API requests.

Authentication Guide

📚 Tutorials

Step-by-step guides for common workflows with complete code examples.

❌ Error Reference

Troubleshooting guide with error codes, causes, and solutions.

View Error Reference

API Overview​

Knowledge Discovery provides two main APIs:

Agent API​

Manage AI agent configurations including creation, updates, access rights, filtering rules, and avatar management.

Base Path: /agent

Q&A API​

Submit questions to agents and retrieve answers with source citations, relevance scoring, and context-aware filtering.

Base Path: /qna

Response Quality​

Knowledge Discovery employs multiple strategies to ensure high-quality responses:

  • RAG (Retrieval-Augmented Generation) - Grounds answers in actual document content
  • Source Citations - Every answer includes references to source documents
  • Rank Scoring - References are ranked by relevance
  • Response Completeness - Indicates whether the answer fully addresses the question
  • Guardrails - Built-in safety measures to prevent hallucinations and ensure accuracy

Last Updated: November 2025