Configuration¶

GPTChangelog can be configured at both the global and project levels. This page explains all available configuration options and how to manage them.

Configuration Locations¶

GPTChangelog looks for configuration in the following locations, in order of precedence:

Project-specific configuration: ./.gptchangelog/config.ini
Global configuration: ~/.config/gptchangelog/config.ini

If a setting is defined in both places, the project-specific setting takes precedence.

Managing Configuration¶

Initializing Configuration¶

To create a new configuration file:

gptchangelog config init

You'll be prompted to choose between global or project-specific configuration, then select either the OpenAI API or a local Codex ChatGPT subscription.

Viewing Current Configuration¶

To see your current configuration:

gptchangelog config show

This will display both global and project-specific configurations if they exist.

Configuration File Format¶

The configuration file uses the INI format:

[openai]
provider = openai
api_key = your-api-key-here
model = gpt-5.5

Configuration Options¶

OpenAI Settings¶

Option	Description	Default
`provider`	`openai` for API keys or `codex` to reuse `~/.codex/auth.json`	Auto-detected
`api_key`	Your OpenAI API key when `provider = openai`	(Required for `openai`)
`model`	The model to use	`gpt-5.5` for `openai`, `gpt-5.4-mini` for `codex`

Environment Variables¶

You can also use environment variables to override configuration settings:

Variable	Corresponding Config Option
`GPTCHANGELOG_PROVIDER`	`[openai] provider`
`OPENAI_API_KEY`	`[openai] api_key`
`GPTCHANGELOG_MODEL`	`[openai] model`

Environment variables take precedence over configuration file settings.

Advanced Configuration¶

Using a Codex Subscription¶

If you already use codex with ChatGPT, GPTChangelog can reuse that login instead of requiring an API key:

codex login
gptchangelog generate --provider codex

Equivalent config:

[openai]
provider = codex
model = gpt-5.4-mini

Using Different Models¶

GPTChangelog works best with GPT-5 family models, but you can use other options:

[openai]
model = gpt-5.4-mini

Smaller models may produce less comprehensive results but can be more cost-efficient.

Multiple Configurations¶

You can maintain different configurations for different projects:

Set up a global configuration with default settings
Create project-specific configurations for projects with special requirements

Example Configurations¶

Minimal Configuration¶

[openai]
provider = openai
api_key = your-api-key-here

Full Configuration¶

[openai]
provider = openai
api_key = your-api-key-here
model = gpt-5.5

Codex Subscription Configuration¶

[openai]
provider = codex
model = gpt-5.4-mini

Configuration for Large Repositories¶

[openai]
api_key = your-api-key-here
model = gpt-5.5

Configuration for Lower Cost¶

[openai]
api_key = your-api-key-here
model = gpt-5.4-mini