Introduction

Orka is the context engine that makes AI actually understand your business.

The Problem with RAG

RAG retrieves documents. It doesn't understand them.

When you ask "which contracts are high-risk?", basic RAG returns contracts containing the word "risk"—but it has no idea what "high-risk" actually means in your organization. Is it liability over $1M? A termination clause? Both?

The result: answers that look right but aren't useful.

The Context Engine

Orka assembles the right context for every query through four layers:

  1. Memory — Session history, user preferences, conversation context
  2. Hybrid Search — Vector similarity + keyword matching + neural reranking
  3. Entity Resolution — "This customer" → Acme Corp → Enterprise tier
  4. Business Rules — Your definitions, metrics, and domain logic

One query. Four layers. Zero config.

Getting Started

New to Orka? Start with the Quickstart guide to build an AI agent that reasons about your documents.

Core Concepts

Agents

Agents are AI-powered interfaces that reason about your business data. Each agent can be configured with:

  • System prompts that define behavior and reasoning approach
  • Connected datastores that provide document context
  • Logic definitions that encode your domain knowledge
  • Custom settings for retrieval, generation, and more

Datastores

Datastores are document collections your agents can reason about. You can:

  • Upload PDFs, Word documents, and other file types
  • Organize documents into logical collections
  • Monitor processing status and document health
  • Configure metadata fields, parsing, and chunking strategies

Logic

Logic definitions teach your agents domain-specific knowledge through three types:

  • Metrics: Mathematical calculations and KPIs (e.g., "MRR = sum of monthly subscription values")
  • Terms: What concepts mean in your context (e.g., "strategic account = ARR > $100K")
  • Rules: How to behave or process data (e.g., "high-risk contract = liability > $1M OR termination clause")

Prompts

Save and organize reusable prompts for your RAG agents:

  • Create templates with variable placeholders
  • Organize prompts into folders
  • Scope to workspace or specific datastores

Chat

The chat interface enables natural conversation with agents that:

  • Answer questions grounded in your documents
  • Apply logic definitions automatically
  • Resolve entities and relationships
  • Provide responses with verifiable citations

Quick Links

Support

Need help? Here are some resources:

  • Documentation: You're already here!
  • GitHub Issues: Report bugs and request features
  • Email Support: Contact us at support@orka.ai