👋Welcome to Kallia

Welcome to the Kallia documentation! Kallia is an advanced semantic document processing library that transforms documents into intelligent, queryable knowledge bases.

What is Kallia?

Kallia is a powerful document processing framework that:

  • Converts documents to structured markdown with intelligent parsing using Docling

  • Generates semantic chunks optimized for retrieval and understanding

  • Creates contextual memories that capture conversation history and insights

  • Provides RESTful APIs for easy integration with FastAPI

  • Supports interactive playground with Chainlit for document Q&A

  • Handles multiple document formats with extensible architecture

Quick Start

  1. Installation: Set up Kallia using Docker or pip

  2. Configuration: Configure your environment variables

  3. API Usage: Start processing documents via REST API

  4. Integration: Integrate with your applications

Documentation Structure

  • Overview - Introduction and core concepts

  • Fundamentals - Setup, API reference, and Docker deployment

  • Use Cases - Practical examples and implementations

Key Features

🔄 Document Processing

Transform any document into structured, searchable content with intelligent parsing and formatting using Docling library.

🧠 Memory Generation

Create contextual memories from conversation histories that understand relationships and provide intelligent insights.

🔍 Semantic Chunking

Break documents into meaningful chunks with summaries, questions, and answers optimized for retrieval and comprehension.

🚀 RESTful API

FastAPI-based service with comprehensive endpoints for document processing, chunking, and memory creation.

🎮 Interactive Playground

Chainlit-powered chat interface for real-time document Q&A with memory management and source citations.

📊 Benchmark Performance

Benchmark Results

Proven superior performance with 4.6/5.0 mean score and 81% perfect score rate in comprehensive RAG evaluations.

Getting Started

Prerequisites

  • Python 3.11 or higher

  • Docker (optional, for containerized deployment)

Quick Installation

pip install kallia

Docker Deployment

# Core API Service
cd kallia/core
docker-compose up -d

# Interactive Playground
cd kallia/playground
docker-compose up -d

Core Concepts

Document Processing Pipeline

  1. Input: PDF documents (extensible to other formats)

  2. Conversion: Document to structured markdown using Docling

  3. Chunking: Semantic segmentation with AI-powered analysis

  4. Output: Structured chunks with summaries, Q&A pairs

Memory System

  • Short-term Memory: Recent conversation context

  • Long-term Memory: Extracted insights and patterns

  • Contextual Understanding: Relationship mapping between conversations

API Architecture

  • RESTful Design: Standard HTTP methods and status codes

  • Pydantic Models: Type-safe request/response validation

  • Error Handling: Comprehensive exception management

  • Scalable: Designed for production deployment

Next Steps

Support

For issues, questions, or contributions:

Built with ❤️ for intelligent document processing

NextIntroduction

Last updated