TinyFlow

TinyFlow is a lightweight, extensible, and provider-agnostic Python framework for building LLM agents. It provides a clean abstraction layer over major LLM providers (OpenAI, Anthropic, Gemini), embedding models, and vector databases, enabling you to build complex agentic workflows with ease.

🚀 Key Features

• Provider Agnostic: Switch seamlessly between OpenAI, Anthropic, and Google Gemini models using a unified LLMFactory.
• Agentic Workflow: Built-in "Thinking -> Acting -> Observing" loop for autonomous task execution.
• Tool Support: Easy-to-use @tool decorator to give your agents capabilities (web search, code execution, etc.).
• Memory & RAG: Integrated Vector Database support (ChromaDB, Qdrant) and Embedding abstraction (OpenAI, SentenceTransformers).
• Unified Configuration: flexible configuration system supporting parameters, environment variables, and settings files with clear precedence.
• Streaming UI: Rich streaming support for building interactive chat interfaces, including reasoning steps and tool execution visibility.
• Modern Stack: Built with Python 3.12+, Pydantic, Asyncio, and managed with uv.

📦 Installation

You can install TinyFlow using pip or uv.

Using pip

pip install tinyflow-llm

Using uv (Recommended)

uv add tinyflow-llm

Installation with Extras

TinyFlow supports optional dependencies for specific features:

• Local Embeddings: pip install "tinyflow-llm[local]"
• Vector Databases: pip install "tinyflow-llm[vector]" (includes ChromaDB and Qdrant)

⚙️ Configuration

TinyFlow uses a unified configuration system. You can configure providers via:

1. Explicit Parameters (passed to Factory create methods) - Highest Priority
2. Environment Variables - Medium Priority
3. Settings / Defaults - Lowest Priority

Common Environment Variables

💡 Usage

1. Basic LLM Usage

Use the LLMFactory to create a provider instance. It automatically handles configuration.

import asyncio
from tinyflow.providers.base.factory import LLMFactory
from tinyflow.core.types import Message

async def main():
    # Automatically loads config from env vars
    llm = LLMFactory.create()

    # Or specify explicitly
    # llm = LLMFactory.create(provider="anthropic", model="claude-3-opus-20240229")

    messages = [
        Message(role="system", content="You are a helpful assistant."),
        Message(role="user", content="Explain quantum computing in one sentence.")
    ]

    response = await llm.generate(messages)
    print(response.content)

if __name__ == "__main__":
    asyncio.run(main())

2. Building an Agent with Tools

Create an Agent equipped with custom tools.

import asyncio
from tinyflow.core.agent import Agent
from tinyflow.providers.base.factory import LLMFactory
from tinyflow.core.tools import tool

# Define a tool
@tool
def get_weather(location: str, unit: str = "celsius") -> str:
    """Get the current weather for a location."""
    # In a real app, call a weather API here
    return f"The weather in {location} is 25°{unit.upper()} and sunny."

async def main():
    # 1. Initialize LLM
    llm = LLMFactory.create()

    # 2. Create Agent with tools
    agent = Agent(
        llm=llm,
        tools=[get_weather],
        system_prompt="You are a helpful assistant with access to weather tools."
    )

    # 3. Run Agent
    response = await agent.run("What's the weather like in Paris?")
    print(response)

if __name__ == "__main__":
    asyncio.run(main())

3. Using Vector Memory

Integrate RAG (Retrieval-Augmented Generation) capabilities.

from tinyflow.vector.factory import VectorDBFactory
from tinyflow.embeddings.factory import EmbeddingFactory
from tinyflow.memory.vector import VectorMemory

# Initialize components
embedding_model = EmbeddingFactory.create()
vector_db = VectorDBFactory.create()

# Create memory interface
memory = VectorMemory(
    vector_db=vector_db,
    embedding_model=embedding_model
)

# Use in Agent
agent = Agent(
    llm=llm,
    memory=memory,
    system_prompt="Use your memory to answer questions."
)

🏗️ Project Structure

tinyflow/
├── tinyflow/
│   ├── config/       # Configuration and helper utilities
│   ├── core/         # Core abstractions (Agent, Tools, Types)
│   ├── providers/    # LLM Provider implementations (OpenAI, Anthropic, Gemini)
│   ├── embeddings/   # Embedding models (OpenAI, Local)
│   ├── vector/       # Vector Database adapters (Chroma, Qdrant)
│   └── memory/       # Memory implementations
├── tests/            # Unit and integration tests
├── main.py           # Entry point example
└── pyproject.toml    # Project dependencies and config

🧪 Development

Running Tests

TinyFlow uses pytest for testing.

# Run all tests
uv run pytest

# Run specific test file
uv run pytest tests/test_factories.py -v

Code Style

The project uses ruff for linting and formatting.

# Lint
uv run ruff check .

# Format
uv run ruff format .

📄 License

MIT License