LLMling-Agent is a powerful framework designed for creating and managing agents powered by large language models (LLMs). Built with modern Python 3.12 features, this tool integrates seamlessly with LLMling's resource system, enabling structured interactions with language models for diverse applications.

Key Features

Modern Architecture

• Async Framework: True asynchronous support allows for straightforward setup of complex asynchronous agent flows, leading to faster initializations and real-time UI experimentation.
• Consistent APIs: The framework provides easy-to-use, consistent APIs that simplify the development process.

Comprehensive Configuration

• YAML Support: Complete agent definitions can be written in YAML files, ensuring easy customization and validation through an extensive JSON schema.
• Type Safety: Leveraging Pydantic's type-safe stack, LLMling-Agent enforces type validation throughout the agent interaction process, significantly enhancing reliability, especially for production systems.

Advanced Agent Functionality

• Multi-modal Support: Support for handling various data types, including images and PDFs, facilitating richer interactions with the agents.
• Command System: An integrated command system allows agents to be controlled through prompt-based interfaces, enhancing user interaction.
• Storage Providers: Agents can log interactions to SQL databases or local files, promoting substantial backtracking capabilities.

Human-AI Collaboration

• Flexible Interaction Modes: The framework offers a human-in-the-loop capability, allowing for various levels of human control in agent operations, ensuring a seamless collaboration between AI and human oversight.

Integration and Extensibility

• Extensive CLI Application: The CLI application supports interactive agent flow building with extensive slash command usage, ensuring enhanced developer friendliness.
• Event System: Built-in support for reactive behaviors based on events such as file changes or webhooks ensures the framework remains responsive to its environment.

Easy Multi-Agent Coordination

• Agent Pool: Facilitate the management of multiple agents for parallel processing of tasks, streamlining workflows in complex projects.

Seamless Documentation Generation

• MkDocs Integration: Easily generate static documentation for projects using LLMling-Agent in conjunction with MkNodes and the MkDocs plugin, simplifying the process of sharing knowledge.

Getting Started

Start creating powerful LLM-powered agents using LLMling-Agent. Here is a quick example to create a browser assistant:

from llmling_agent import AgentPool

async def main():
    async with AgentPool() as pool:
        browser = await pool.add_agent(
            'browser',
            system_prompt='Open Wikipedia pages matching the topics you receive.',
            model='openai:gpt-4o-mini',
            tools=['webbrowser.open']
        )
        agent = await pool.add_agent('assistant', model='openai:gpt-4o-mini')
        connection = agent >> browser  # Setup a permanent connection.
        await agent.run('Tell us a random major city! Just one word!')

Explore the flexibility and power of creating LLM-powered agents with LLMling-Agent and harness the capabilities of modern language models in your applications.