opencommit

To use Ollama as your provider in OpenCommit, follow these steps:

1. Install Ollama: Download from https://ollama.ai/download.
2. Pull a model: Run ollama pull <model_name> (e.g., ollama pull llama3:8b).
3. Start the server: Run ollama serve.
4. Configure OpenCommit: Run the setup command and select Ollama as your provider. The wizard will attempt to detect your local models and ask for the Ollama URL (default is http://localhost:11434).

You can manage your OpenCommit configuration using the oco config command. The tool supports three modes: set to update configuration values, get to retrieve current settings, and describe to view available options.

Configuration is stored in a global file at ~/.opencommit (INI format) and can be overridden by a .env file in your current working directory.

To set a configuration value, use the set command with key-value pairs. For example, to set your API key and provider:

oco config set OCO_API_KEY=your_key OCO_AI_PROVIDER=openai

To use llama.cpp as your provider in OpenCommit:

1. Install llama.cpp: Follow the instructions at https://github.com/ggerganov/llama.cpp.
2. Start the server: Run llama-server -m <model.gguf> --port 8080.
3. Configure OpenCommit: Run the setup command and select llama.cpp. You will be prompted to provide the server URL (default is http://localhost:8080).

You can use custom message templates by providing an argument that contains the OCO_MESSAGE_TEMPLATE_PLACEHOLDER (defined in your configuration). The tool will replace this placeholder with the AI-generated commit message.

For example, if your configuration defines a placeholder, you can pass a string like "feat: ${OCO_MESSAGE_TEMPLATE_PLACEHOLDER}" as an extra argument to ensure all generated messages follow a specific prefix pattern.

The OpenCommit configuration is stored in a global config file. When running the setup wizard, the following configuration keys are populated. You can also use these keys to understand the structure of the configuration:

The prepareCommitMessageHook behavior is influenced by the following configuration settings:

• OCO_API_KEY: Required for generating commit messages. If missing, the hook will exit with an error message instructing you to set it via oco config set OCO_API_KEY=<value>.
• OCO_HOOK_AUTO_UNCOMMENT: Controls how the generated message is written to the commit file.
  • When true, the generated message is written without comment characters (#).
  • When false, the message is commented out with a header/footer block, allowing you to manually edit/uncomment it in your git editor.

Use fetchModelsForProvider to retrieve a list of available model IDs from a given AI provider. This function attempts to fetch live data from the provider's API, but falls back to a local cache or a hardcoded list if the request fails or if no API key is provided.

Supported providers include:

• openai
• ollama
• llamacpp
• anthropic
• gemini
• groq
• mistral
• deepseek
• aimlapi
• openrouter

Note: For ollama and llamacpp, you can provide a baseUrl. For providers requiring authentication, you must provide an apiKey to fetch live models instead of the fallback list.

OpenCommit provides utilities to transform raw errors into user-friendly, actionable structures and styled strings.

formatUserFriendlyError

Converts an error into a FormattedError object containing a title, message, helpUrl (if applicable), and a suggestion.

import { formatUserFriendlyError } from './utils/errors';

const formatted = formatUserFriendlyError(error, 'openai');
// Returns: { title, message, helpUrl, suggestion }

printFormattedError

Takes a FormattedError object and returns a string styled with chalk (red for errors, cyan for help, yellow for suggestions) suitable for CLI output.

import { formatUserFriendlyError, printFormattedError } from './utils/errors';

const formatted = formatUserFriendlyError(error, 'anthropic');
console.log(printFormattedError(formatted));