tinygent 0.1.1008

Tiny agentic framework for building AI agents which can do more than tiny just tasks.

✨ Tiny platform for amazing agents ✨

🏠 Homepage | 📚 Documentation | 💡 Examples | 🚀 Quick Start

🤖 Tinygent is a tiny agentic framework - lightweight, easy to use (hopefully), and efficient (also hopefully ;-0) library for building and deploying generative AI applications. It provides a simple interface for working with various models and tools, making it ideal for developers who want to quickly prototype and deploy AI solutions.

🎯 Create an agent

# uv sync --extra openai

from tinygent.tools import tool
from tinygent.core.factory import build_agent

@tool
def get_weather(location: str) -> str:
    """Get the current weather in a given location."""
    return f'The weather in {location} is sunny with a high of 75°F.'

agent = build_agent(
    'react',
    llm='openai:gpt-4o-mini',
    tools=[get_weather],
)

print(agent.run('What is the weather like in Prague?'))

🚀 Getting Started

📋 Prerequisites

Before you begin using tinygent, ensure that you meet the following software prerequisites.

• Install Git
• Install uv

💻 Install From Source

1. Clone the tinygent repository to your local machine.

   git clone [e-mail]:filchy/tinygent.git tinygent
   cd tinygent
   

2. Create a Python environment.

   uv venv --seed .venv
   source .venv/bin/activate
   

3. Install the tinygent library. To install only the core tinygent library without any optional dependencies, run the following:

   uv sync
   

   To install the tinygent library along with all of the optional dependencies. Including developer tools (--all-groups), additional packages and all of the dependencies needed for profiling and plugins (--all-extras) in the source repository, run the following:

   uv sync --all-groups --all-extras
   

   [!NOTE] Not all packages are included in the default installation to keep the library lightweight. You can customize your installation by specifying the optional dependencies you need.

4. Install tinygent in editable mode (development mode), so that changes in the source code are immediately reflected:

   uv pip install -e .
   

🎬 See It In Action

🏗️ Architecture

%%{init: {'theme': 'base', 'themeVariables': {'primaryColor': '#e1f5ff', 'primaryTextColor': '#0d47a1', 'primaryBorderColor': '#42a5f5', 'lineColor': '#1976d2', 'secondaryColor': '#fff4e1', 'tertiaryColor': '#f0e1ff'}}}%%
flowchart LR
    User[User Code]:::userNode
    Factory[Factories]:::factoryNode
    Runtime[Runtime Registry]:::runtimeNode
    Components[Agents & Tools & Memory]:::componentNode
    Packages[Provider Packages]:::packageNode

    User a@--> Factory
    Factory b@--> Runtime
    Runtime c@--> Components
    Packages d@-.-> Runtime

    a@{animate: true}
    b@{animate: true}
    c@{animate: true}
    d@{animate: true}

    linkStyle 0 stroke-width:2px
    linkStyle 1 stroke-width:2px
    linkStyle 2 stroke-width:2px
    linkStyle 3 stroke-width:2px,stroke-dasharray: 5 5

    classDef userNode fill:#e1f5ff,stroke:#1976d2,stroke-width:2px,color:#0d47a1
    classDef factoryNode fill:#fff4e1,stroke:#f57c00,stroke-width:2px,color:#e65100
    classDef runtimeNode fill:#f0e1ff,stroke:#7b1fa2,stroke-width:2px,color:#4a148c
    classDef componentNode fill:#e1ffe1,stroke:#388e3c,stroke-width:2px,color:#1b5e20
    classDef packageNode fill:#ffe1e1,stroke:#d32f2f,stroke-width:2px,color:#b71c1c

Tinygent uses a registry-based plugin architecture: Packages register components into the Runtime. Factories query the Runtime to build Components for your code.

💡 Examples (Quick Start)

1. 🔑 Ensure you have set the OPENAI_API_KEY environment variable to allow the example to use OpenAI's API. An API key can be obtained from openai.com.

   export OPENAI_API_KEY="your_openai_api_key"
   

2. ▶️ Run the examples using uv:

   uv run examples/agents/multi-step/main.py
   

3. 🔍 Explore more examples below:

📚 Features & Examples

Name	Type	Description
Tool Usage	Basics	Basic tool creation and usage
LLM Usage	Basics	Direct LLM interaction patterns
Function Calling	Basics	Function calling with LLMs
Chat Buffer Memory	Memory	Store complete conversation history
Summary Buffer Memory	Memory	Summarize older messages to save tokens
Window Buffer Memory	Memory	Keep only recent N messages
Combined Memory	Memory	Combine multiple memory strategies
Basic Tools	Tools	Simple tool definitions with @tool
Reasoning Tools	Tools	Tools that provide reasoning traces
JIT Tools	Tools	Just-in-time compiled tools
Middlewares in Agents	Agents	Add custom processing layers to agents
ReAct Agent	Agents	Reasoning and acting pattern
Multi-Step Agent	Agents	Break down complex tasks into steps
Squad Agent	Agents	Coordinate multiple specialized agents
Modular Agentic Planner	Agents	Advanced planning with modular architecture
Tiny OpenAI	Package	OpenAI LLMs and embeddings
Tiny Anthropic	Package	Anthropic Claude LLMs
Tiny MistralAI	Package	Mistral AI LLMs
Tiny Gemini	Package	Google Gemini LLMs
Tiny VoyageAI	Package	VoyageAI embedding models
Brave Tools	Package	Brave search integration
Tiny Chat	Package	FastAPI-based chat interface
Tiny Graph	Package	Neo4j knowledge graph integration

✅ Linting & Formatting

To ensure code quality, formatting consistency, and type safety, run:

uv run fmt   # Format code Ruff
uv run lint  # Run Ruff linter and Mypy type checks