GEMINI

Gemini AI-powered assistant

This file provides instructions for the Gemini AI-powered assistant on how to work with the speedgrapher project.

Project Overview

Speedgrapher is a local MCP (Model Context Protocol) server written in Go. It's designed to assist writers, especially in the tech industry, by providing a suite of tools to streamline the writing process. It uses the official Go SDK for MCP and communicates over the stdio transport layer.

Development Workflow

The project uses a Makefile to manage common development tasks.

• Building the project: To build the server, run the following command: ```bash make build

  This will create an executable at `speedgrapher`.
  
  

• Running tests: To run the project's tests, use the following command:

  make test
  

• Cleaning the project: To remove the executable, run:

  make clean
  

Dependencies

The main dependency for this project is the github.com/modelcontextprotocol/go-sdk. The project's module path is github.com/danicat/speedgrapher.

Coding Style

The project follows standard Go coding conventions. Please maintain this style when adding or modifying code.

Prompts

The server's functionality is exposed through a series of prompts. New prompts can be added in the internal/prompts directory. Each new prompt requires two functions:

1. A function that defines the prompt's name, description, and arguments.
2. A handler function that implements the prompt's logic.

After creating these two functions, you must register the new prompt in the run function in cmd/speedgrapher/main.go.

Available Prompts:

• haiku: Creates a haiku about a given topic.
• interview: Interviews an author to produce a technical blog post.
• localize: Translates an article into a target language.
• readability: Analyzes the last generated text for readability and suggests improvements. It uses the fog tool.
• reflect: Analyzes the current session and proposes improvements to the development process.
• review: Reviews an article against the editorial guidelines.

Tools

The server can also expose tools to the user. New tools can be added in the internal/tools directory.

Available Tools:

• fog: Calculates the Gunning Fog Index for a given text. The implementation uses a heuristic based on vowel groups to count syllables.

General Development Principles

• Prioritize API/SDK Understanding: Before writing code that uses an external library or SDK, I must first use documentation-lookup tools (godoc, web_search) to understand the correct API usage. I will prioritize official documentation and user-provided links.
• Favor Consistent Heuristics: When implementing features based on heuristics (e.g., readability formulas), I will propose a simple, robust, and consistent algorithm first. I will explain its trade-offs and seek user approval before attempting more complex, exception-laden solutions.
• Trust but Verify: I must critically evaluate all information, including user-provided test cases and my own generated text. I will use the tools at my disposal to verify data and assumptions before acting on them.
• Maintain an Objective Tone: I will present facts and results objectively and avoid defensive justifications unless explicitly asked for them. My primary goal is to be a helpful, data-driven assistant.
• Use the Right Tool for the Job: Prioritize built-in tools over shell commands. If a specific tool is available for a task (e.g., gopretty, godoc), use it directly instead of trying to execute it through run_shell_command.
• Always Read the README: At the start of every session, read the README.md file to ensure you have the latest information about the project, especially the testing procedures.