ThinkTank

ThinkTank is a modular, collaborative AI system designed to generate comprehensive and nuanced answers to complex questions. By leveraging multiple AI-powered experts, ThinkTank combines diverse perspectives to provide well-rounded insights, mimicking the dynamics of human expert panels.

Features

• Collaborative AI Models: Integrates multiple AI experts, each with unique prompts and areas of specialization.
• Modular Design: Easily add, remove, or customize experts to tailor ThinkTank to specific domains or use cases.
• Emergent Intelligence: Synthesizes insights from various experts, producing outputs that are greater than the sum of their parts.
• Interdisciplinary Reasoning: Handles complex, multifaceted queries requiring expertise in diverse fields.
• Customizable Outputs: Refines answers through iterative evaluation and context-aware synthesis.

How It Works

ThinkTank orchestrates the following workflow:

1. Input: A user provides a question or topic.
2. Expert Evaluation: Each AI expert independently evaluates the input based on its unique expertise and prompt.
3. Synthesis: ThinkTank consolidates expert responses, incorporating their insights into a new comprehensive prompt, providing a richer response.

Getting Started

Prerequisites

• Go 1.23 or later
• OpenAI API credentials
• Homebrew (for dependency management)

Installation

1. Clone the repository:

   git clone https://github.com/kmesiab/think-tank.git
   cd think-tank

2. Install dependencies:

   go mod tidy

3. Set up OpenAI API credentials: Ensure your OpenAI API key is set as an environment variable:

   export OPENAI_API_KEY="your-api-key-here"

4. Build the project:

   go build main.go

Usage

Run the ThinkTank application with the following command:

./think-tank

Customizing Experts

You can modify or add experts in the main.go file. Each expert is defined as follows:

var economicsExpert = &Expert{
    CallOptions: callOptions,
    LLM:         llm,
    Name:        "Economics Expert",
    Description: "A highly skilled economist with expertise in international
    trade, economic growth, and policy analysis.",
    Prompt:      "Given the following question: %s\nProvide an in depth...",
}

To add a new expert, copy this structure and update the Name, Description, and Prompt fields.

---

Architecture

Core Components

1. ThinkTank:

   • The main orchestrator that manages experts and synthesizes their outputs.

2. Expert:

   • Represents an AI model with a specific area of expertise, prompt, and evaluation logic.

3. LLMs:

   • Leverages the langchaingo library for seamless integration with OpenAI models.