struktx-ai 0.0.3b6

A configurable, typed AI framework with swappable LLM, classifier, handlers, and optional memory

A configurable, typed AI framework with swappable LLM, classifier, handlers, and memory for Natural Language to Action apps

Quick Install

Using uv (Recommended)

# Install with core dependencies only
uv pip install struktx-ai

# Install with LLM support (LangChain)
uv pip install struktx-ai[llm]

# Install with vector store support
uv pip install struktx-ai[vector]

# Install with all optional dependencies
uv pip install struktx-ai[all]

# Install for development
uv pip install struktx-ai[dev]

Using pip

# Install with core dependencies only
pip install struktx-ai

# Install with LLM support (LangChain)
pip install struktx-ai[llm]

# Install with vector store support
pip install struktx-ai[vector]

# Install with all optional dependencies
pip install struktx-ai[all]

Source

# Create virtual env
uv venv

# Activate environment
source .venv/bin/activate

# Sync and install pyproject dependencies
uv sync

# Install finall Strukt package
uv pip install struktx-ai[all]

Quick Start with LangChain

from strukt import create, StruktConfig, HandlersConfig, LLMClientConfig, ClassifierConfig
from strukt.classifiers.llm_classifier import DefaultLLMClassifier
from strukt.prompts import DEFAULT_CLASSIFIER_TEMPLATE
from strukt.examples.time_handler import TimeHandler

from langchain_openai import ChatOpenAI
from strukt.langchain_helpers import LangChainLLMClient

# Create LangChain LLM client
langchain_llm = ChatOpenAI(api_key="your-openai-api-key")
llm = LangChainLLMClient(langchain_llm)

app = create(StruktConfig(
    llm=LLMClientConfig(llm),
    classifier=ClassifierConfig(
        factory=DefaultLLMClassifier(
            llm=llm,
            prompt_template=DEFAULT_CLASSIFIER_TEMPLATE,
            allowed_types=["time_service", "weather", "general"],
            max_parts=4,
        )
    ),
    handlers=HandlersConfig(
        registry={"time_service": TimeHandler(llm)},
        default_route="general",
    ),
))

app.invoke("What is the time in Tokyo?")

What is StruktX?

StruktX is a Python framework for building AI applications with focus on natural language -> actions

• 🔄 Swappable Components: LLM clients, classifiers, handlers, and memory engines
• 📝 Type Safety: Full type hints and Pydantic models
• ⚙️ Configurable: Factory-based configuration system
• 🔌 Extensible: Easy to add custom components
• 🧠 Memory Support: Optional conversation memory
• 🔧 Middleware: Pre/post processing hooks

Features

• LLM Integration: Support for any LLM via the LLMClient interface
• Query Classification: Route requests to appropriate handlers
• Structured Outputs: Pydantic model integration
• Memory Engines: Conversation history and context
• Middleware System: Pre/post processing hooks
• LangChain Helpers: Easy integration with LangChain ecosystem

Requirements

• Python 3.10+
• Core: pydantic>=2.0.0, python-dotenv>=1.0.0
• Optional: LangChain packages, vector stores, etc.

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please report any issues, bugs, improvements, or features on GitHub.

Support

• 📖 Documentation
• 🐛 Issue Tracker
• 💬 Discussions