Gemini Thinking

🚀 Model Context Protocol - Gemini Thinking Server

This is a server implemented based on the Model Context Protocol (MCP), integrating the Google Gemini API to provide analytical thinking capabilities without code generation.

🚀 Quick Start

The Gemini Thinking Server is a dedicated MCP server that leverages Google's Gemini model to offer serialized thinking and problem-solving capabilities. It supports:

• Breaking down complex problems into steps
• Planning and design with room for revision
• Analyzing situations where a change of direction may be needed
• Problems with an as-yet unclear scope

✨ Features

• Gemini-powered Thinking: Utilize Gemini's analytical capabilities to generate meaningful responses.
• Meta-commentary: Provide insights into the reasoning process.
• Confidence Level: Indicate Gemini's confidence in its analysis.
• Alternative Paths: Suggest different problem-solving approaches.
• Branch Thoughts: Allow exploration of different thinking paths.
• Revision Capability: Support the revision of previous thoughts.
• Session Persistence: Save and resume analysis sessions.

📦 Installation

# Clone the repository
git clone <repository-url>

# Install dependencies
npm install

# Build the project
npm run build

💻 Usage Examples

Basic Usage

Environment Configuration

Before running the server, you need to set the Gemini API key:

export GEMINI_API_KEY=your_api_key_here

Run the Server

node dist/gemini-index.js

Tool Parameters

The geminithinking tool accepts the following parameters:

• query (Required): The question or topic to be analyzed.
• context (Optional): Additional context information.
• approach (Optional): A suggested method for problem-solving.
• previousThoughts (Optional): An array of previous thinking steps.
• thought (Optional): The current thinking step (if empty, it will be generated by Gemini).
• nextThoughtNeeded (Required): Whether another thinking step is needed.
• thoughtNumber (Required): The current thinking number.
• totalThoughts (Required): The estimated total number of thinking steps required.
• isRevision (Optional): Whether you are revising a previous thought.
• revisesThought (Optional): The number of the thought being reconsidered.
• branchFromThought (Optional): The number of the thought at the branching point.
• branchNumber (Optional): The current branch number.
• thoughtFilter (Optional): A custom function for filtering the output.

Advanced Usage

Session Management

Save a Session

To save the current session, run:

node dist/save-session.js

Load a Session

To load a previously saved session, run:

node dist/load-session.js

Response Format

The server will return the following structured JSON response:

{
  "thought": "string", // The thinking content generated by Gemini
  "reasoning": "string", // The reasoning steps in the thinking process
  "confidence": number, // The confidence level in the current conclusion (between 0 and 1)
  "sources": ["string"] // The information sources or references used
}

📄 License

MIT