Tiny Crew

Tiny Crew is an innovative, flexible multi-agent AI system designed to tackle complex tasks through intelligent collaboration. It leverages the power of large language models to create a team of specialized AI agents that work together to achieve common goals.

Features

LLM-Driven Agent Selection: Utilizes a language model to intelligently assign tasks to the most suitable agent based on task requirements and agent capabilities.
Multi-Agent Collaboration: Create a crew of AI agents, each with their own specialization and goal.
Flexible Task Assignment: Dynamically assigns tasks to agents based on their skills and available tools.
Shared Knowledge Base: Agents can share information and build upon each other's work through a shared memory system.
Tool Integration: Equip agents with tools for interacting with external systems (e.g., file operations, database queries, web access).
Goal-Oriented Workflow: Define an overarching goal for the crew and let them work towards it collaboratively.
Extensible Architecture: Easily add new agents, tools, or modify existing ones to suit your specific needs.
Intelligent Summarization: Uses the LLM to generate comprehensive summaries of the crew's work, addressing the overall goal.

Prerequisites

Node.js or Bun
TypeScript
OpenAI API key or compatible API (e.g., Groq)

Installation

Clone the repository:

git clone https://github.com/skitsanos/tiny-crew.git
cd tiny-crew

Install dependencies:
```
npm install
```
or
```
bun install
```

Set up environment variables: Create a .env file in the root directory and add your API keys:

OPENAI_API_KEY=your_api_key_here
GROQ_API_KEY=your_groq_api_key_here
GROQ_API_URL=https://api.groq.com/openai/v1

Usage

Define your agents in the main script (e.g., src/main.ts):

const agent1 = new Agent({
    name: 'Alice',
    goal: 'Conduct research and provide concise summaries',
    expectedOutput: 'Bullet points or short paragraphs',
    model: BASE_MODEL
}, openai);

const agent2 = new Agent({
    name: 'Bob',
    goal: 'Perform code writing tasks, generate code examples, and save files',
    expectedOutput: 'Confirmation that the code snippet was saved to a file',
    model: BASE_MODEL
}, openai, [new FileWriteTool()]);

// Add more agents as needed

Create a crew and add agents:

const llm = new OpenAI({
    apiKey: process.env.GROQ_API_KEY,
    baseURL: process.env.GROQ_API_URL
});

const crew = new Crew('Develop and present a comprehensive overview of recent AI advancements and their implications', llm);
crew.addAgent(agent1);
crew.addAgent(agent2);

Define tasks and run the crew mission:

const tasks = [
    'Research recent advancements in AI and summarize them in 3 bullet points',
    'Create a Python file named "example.py" with a "Hello, World!" program',
    'Synthesize the research and analysis into a coherent overview'
];

async function runCrewMission()
{
    for (const task of tasks)
    {
        const result = await crew.assignTask(task);
        console.log(`Task result: ${result}`);
    }

    const crewSummary = await crew.achieveCrewGoal();
    console.log('Crew goal achievement summary:', crewSummary);
}

runCrewMission();

Run the script:

npx ts-node src/index.ts

bun run src/index.ts

Extending Tiny Crew

Adding New Agents

Create new agents by instantiating the Agent class with different goals and specializations.

Implementing Tools

Create new tools by implementing the Tool interface:

class MyNewTool implements Tool
{
    name = 'MyNewTool';
    description = 'Description of what the tool does';

    async use(args: any): Promise<any>
    {
        // Implement the tool's functionality
    }
}

Then, pass the tool to an agent when creating it:

const agentWithTool = new Agent({
    // Agent configuration
}, openai, [new MyNewTool()]);

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License