lmeow.nvim

   ∧,,,∧
  (  • ω •)
  / づ💡 

🐱 AI-powered inline code editing for Neovim

✨ Features

• 🎯 Inline AI editing - Select code and get AI-powered modifications
• 🤖 Multiple LLM providers - OpenAI, Claude, Gemini, OpenRouter, Grok
• 🚀 Environment variable support - No hardcoded API keys needed
• 🎨 Custom models - Easy addition of your own AI models
• 📝 Smart prompts - Preserves existing content while making requested changes
• 🎹 Simple interface - Visual select → press shortcut → enter prompt → done
• 🔧 No configuration required - Works out of the box with sensible defaults

🚀 Installation

Add to your lazy.nvim config:

{
  "0xdilo/lmeow.nvim",
  dependencies = { "nvim-lua/plenary.nvim" },
  config = function()
    require("lmeow").setup({
      -- Configuration goes here (see below)
    })
  end
}

🔑 Setup API Keys

Set environment variables in your shell:

# OpenAI
export OPENAI_API_KEY="your-openai-key"

# Anthropic/Claude  
export ANTHROPIC_API_KEY="your-claude-key"

# Google Gemini
export GEMINI_API_KEY="your-gemini-key"

# OpenRouter
export OPENROUTER_API_KEY="your-openrouter-key"

# X.AI/Grok
export XAI_API_KEY="your-grok-key"

🎯 Quick Start

1. Select code in visual mode (v)
2. Press <leader>ac (default keymap)
3. Enter your prompt (e.g., "add error handling", "optimize this function", "translate to Python")
4. AI replaces your selection with the improved code

📖 Commands

• :Lmeow model <name> - Switch to a different model
• :Lmeow status - Show current model and available models

🤖 Built-in Models

The plugin includes these models by default:

Model Name	Provider	Description
gpt4	OpenAI	GPT-4
gpt4o	OpenAI	GPT-4o
gpt4turbo	OpenAI	GPT-4 Turbo
gpt35	OpenAI	GPT-3.5 Turbo
claude	Anthropic	Claude 3.5 Sonnet
claudeopus	Anthropic	Claude 3 Opus
claudehaiku	Anthropic	Claude 3 Haiku
gemini	Google	Gemini 2.5 Flash
geminipro	Google	Gemini 2.5 Pro
gpt4router	OpenRouter	GPT-4 via OpenRouter
llama	OpenRouter	Llama 3.1 70B
grok	X.AI	Grok Beta

⚙️ Configuration

Basic Setup

require("lmeow").setup({
  default_model = "gpt4",  -- Your preferred default model
  keymaps = {
    edit_selection = "<leader>ac"  -- Or your preferred keymap
  }
})

Adding Custom Models

require("lmeow").setup({
  default_model = "gpt5",
  
  models = {
    -- Add your custom models here
    gpt5 = {
      provider = "openai",
      model = "gpt-5",
      name = "GPT-5",
      -- Per-model params (override provider defaults)
      params = {
        temperature = 1,
        max_completion_tokens = 32000
      }
    },
    
    myclaude = {
      provider = "claude",
      model = "claude-3-5-sonnet-20241022", 
      name = "My Claude",
      params = { max_tokens = 5000 }
    },
    
    mixtral = {
      provider = "openrouter",
      model = "mistralai/mixtral-8x7b-instruct",
      name = "Mixtral 8x7B"
    },
    
    coding_assistant = {
      provider = "gemini",
      model = "gemini-2.5-flash",
      name = "Fast Coder"
    }
  },
  
  -- Override provider settings if needed
  providers = {
    openai = {
      defaultModelParams = {
        max_completion_tokens = 3000,
        temperature = 1
      }
    },
    claude = {
      defaultModelParams = {
        max_tokens = 4000,
        temperature = 0.7
      }
    }
  },
  
  keymaps = {
    edit_selection = "<leader>ac"
  }
})

Advanced Configuration

require("lmeow").setup({
  default_model = "gpt4",
  
  -- Add your own custom system prompt that gets combined with the default one
  custom_system_prompt = "Speak with enthusiasm and use emojis when appropriate 🚀",
  
  models = {
    -- Your custom models
  },
  
  providers = {
    openai = {
      defaultModelParams = {
        max_completion_tokens = 32000,
        temperature = 1
      },
      -- Override base URL for custom endpoints
      base_url = "https://api.openai.com/v1/chat/completions"
    }
  },
  
  keymaps = {
    edit_selection = "<leader>ac"
  }
})

Provider vs Model Params

• Provider config sets defaults for model parameters under providers.<name>.defaultModelParams (e.g., temperature, max_tokens).
• Model config can override these under models.<name>.params for that specific model.
• Non-payload settings like base_url, env_var, api_key stay out of the JSON payload.
• API keys resolve from providers.<name>.env_var automatically; a model-level api_key can still override if needed.

Example:

require("lmeow").setup({
  models = {
    precise_gpt4 = {
      provider = "openai",
      model = "gpt-4o",
      name = "Precise GPT-4o",
      params = {
        temperature = 0.1,   -- override provider default
        max_tokens = 4096    -- override provider default
      }
    }
  },
  providers = {
    openai = {
      defaultModelParams = {
        temperature = 0.7,   -- default for all OpenAI models
        max_tokens = 2000
      }
    }
  }
})

Custom System Prompt

You can add your own custom system prompt that will be combined with the default system prompt. This is perfect for:

• Setting a specific tone or style (e.g., "Always speak with emojis 😊")
• Adding domain-specific knowledge or requirements
• Customizing the AI's behavior for your specific needs

require("lmeow").setup({
  default_model = "gpt4",
  
  -- Examples of custom system prompts:
  custom_system_prompt = "Speak without using any emojis",  -- Disable emojis
  -- OR
  custom_system_prompt = "Always respond with a helpful tone and include relevant code examples",
  -- OR  
  custom_system_prompt = "You are a senior Python developer. Focus on clean, idiomatic Python code.",
  
  keymaps = {
    edit_selection = "<leader>ac"
  }
})

🎨 Usage Examples

Code Refactoring

-- Select this function and prompt: "add error handling"
function getData()
  return fetch("/api/data")
end

-- AI might return:
async function getData() {
  try {
    const response = await fetch("/api/data");
    if (!response.ok) {
      throw new Error(`HTTP error! status: ${response.status}`);
    }
    return await response.json();
  } catch (error) {
    console.error("Failed to fetch data:", error);
    throw error;
  }
}

Text Processing

Hello, how are you?

Prompt: "translate to spanish"

Result:

Hola, ¿cómo estás?

HTML Enhancement

<head>
  <meta charset="utf-8" />
</head>

Prompt: "add meta tags for SEO"

Result:

<head>
  <meta charset="utf-8" />
  <meta name="viewport" content="width=device-width, initial-scale=1" />
  <meta name="description" content="Your page description" />
  <meta name="keywords" content="keyword1, keyword2, keyword3" />
  <meta name="author" content="Your Name" />
</head>

🛠️ How It Works

1. Visual Selection - Select code/text in visual mode
2. Smart Prompting - Plugin creates context-aware prompts that preserve existing content
3. API Integration - Sends request to your chosen AI provider
4. Inline Replacement - Replaces selection with AI's response
5. Content Preservation - AI is instructed to keep unrelated content intact

🔍 Troubleshooting

Common Issues

"No text selected" error

• Make sure you've actually selected text in visual mode
• Try selecting again - sometimes visual marks need to be reset

"API key not set" error

• Check your environment variables are set correctly
• Restart Neovim after setting environment variables

Model not found

• Check the model name in your config
• Use :Lmeow status to see available models

Debug Mode

Enable debug logging:

require("lmeow").setup({
  -- Your config
})

🤝 Contributing

Contributions welcome! Please feel free to submit issues and pull requests.

📄 License

MIT License - see LICENSE file for details.

---

Made with 🐱💡 by 0xdilo

   ∧,,,∧
  (  • ω •)
  / ❤️ づ