kaygraph 0.2.0

A context-graph framework for building production-ready AI applications.

KayGraph

An opinionated framework for building context-aware AI applications with production-ready graphs.

What is KayGraph?

KayGraph provides powerful abstractions for orchestrating complex AI workflows through Context Graphs - a pattern that seamlessly integrates operations, LLM calls, and state management into production-ready applications.

Core Philosophy

• Context-Aware Graphs: Build sophisticated AI systems where every node has access to shared context
• Opinionated Patterns: Production-tested patterns for common AI workflows
• Zero Dependencies: Pure Python implementation with no external dependencies
• Bring Your Own Tools: Integrate any LLM, database, or service you prefer

Installation

Using uv (Recommended)

uv is a fast Python package manager that provides better dependency resolution and faster installations:

# Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install KayGraph
uv pip install kaygraph

# Or install from source with development dependencies
git clone https://github.com/KayOS-AI/KayGraph.git
cd KayGraph/kaygraph-library
uv pip install -e ".[dev]"

Using pip

pip install kaygraph

Or install from source:

git clone https://github.com/KayOS-AI/KayGraph.git
cd KayGraph/kaygraph-library
pip install -e .

Quick Start

🚀 Scaffold a New Project

Use our zero-dependency scaffolding tool to generate production-ready boilerplate:

# Generate a basic node pattern
python scripts/kaygraph_scaffold.py node DataProcessor

# Generate an interactive chat application
python scripts/kaygraph_scaffold.py chat CustomerSupport

# Generate a supervisor-worker pattern
python scripts/kaygraph_scaffold.py supervisor TaskManager

# See all available templates
python scripts/kaygraph_scaffold.py --help

📝 Write Your First Graph

from kaygraph import Node, Graph

# Define a simple node
class AnalyzeNode(Node):
    def prep(self, shared):
        # Read from shared context
        return shared.get("input_text")

    def exec(self, text):
        # Process data (e.g., LLM call)
        return analyze_sentiment(text)

    def post(self, shared, prep_res, exec_res):
        # Write to shared context
        shared["sentiment"] = exec_res
        return "complete"  # Next action

# Create and run a graph
analyze = AnalyzeNode("analyzer")
graph = Graph(analyze)

shared = {"input_text": "KayGraph makes AI development intuitive!"}
graph.run(shared)
print(shared["sentiment"])  # Output: "positive"

Key Features

🏗️ Core Abstractions

• Node: Atomic unit of work with 3-phase lifecycle (prep → exec → post)
• Graph: Orchestrates node execution through labeled actions
• Shared Store: Context-aware state management across nodes

🎯 Production Patterns

• Agent: Autonomous decision-making systems
• RAG: Retrieval-augmented generation pipelines
• Workflows: Multi-step task orchestration
• Batch Processing: Efficient data processing at scale
• Async Operations: Non-blocking I/O operations

🚀 Enterprise Features

• ValidatedNode: Input/output validation
• MetricsNode: Performance monitoring
• Comprehensive Logging: Built-in debugging support
• Error Handling: Graceful failure recovery
• Resource Management: Context managers for cleanup

Examples & Workbooks

Explore 71 production-ready examples in the workbooks/ directory:

• Quick Start: kaygraph-hello-world, kaygraph-workflow, kaygraph-chat
• AI Agents: kaygraph-agent, kaygraph-multi-agent, kaygraph-agent-tools
• RAG Systems: kaygraph-rag, kaygraph-text2sql, kaygraph-tool-embeddings
• Production: kaygraph-production-ready-api, kaygraph-realtime-monitoring

See workbooks/QUICK_FINDER.md for task-based navigation or workbooks/guides/WORKBOOK_INDEX.md for the complete catalog.

Documentation

Core Concepts

• Node Design - Understanding the 3-phase lifecycle
• Graph Orchestration - Connecting nodes with actions
• Shared Store - Managing shared context

Common Patterns

• Building Agents
• RAG Pipelines
• Workflows
• Map-Reduce

Advanced Topics

• Async Operations
• Batch Processing
• Parallel Execution
• Production Best Practices

Development with AI Assistants

KayGraph is designed for Agentic Coding - where humans design and AI agents implement.

Generate Cursor Rules

# Generate AI coding assistant rules from documentation
python utils/update_kaygraph_mdc.py

This creates .cursor/rules/ with context-aware guidance for AI assistants.

Project Structure

kaygraph-library/
├── kaygraph/          # Core framework
│   └── __init__.py    # All abstractions in one file
├── docs/              # Comprehensive documentation
├── tests/             # Unit tests
├── utils/             # Helper scripts
└── .cursor/rules/     # AI assistant guidance

Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

MIT License - see LICENSE for details.

---

Built with ❤️ by the KayOS Team