Generator-Specific Configuration

This document provides a complete reference for configuring individual AI generators in ai-commit. Each generator has its own configuration block within the generators array, containing driver specifications, authentication credentials, model parameters, and execution options.

For information about selecting and using generators during commit workflow, see AI Generators Overview. For details on the generator architecture and implementation patterns, see AI Generator System.

Configuration Structure

Each generator configuration follows a common structure but contains generator-specific options based on its integration pattern (API-based or CLI-based).

Generator Configuration to Runtime Instantiation Flow

Sources: config/ai-commit.php78-219 app/Generators/OpenAIGenerator.php26-29 app/Generators/BitoCliGenerator.php20-32 app/Commands/CommitCommand.php39-76

Configuration Access at Runtime

Sources: app/Commands/CommitCommand.php74-76 app/Generators/OpenAIGenerator.php26-29 app/Generators/BitoCliGenerator.php20-28

Common Configuration Keys

All generator configurations share these required keys:

Key	Type	Description
driver	string	Generator driver identifier matching key in GeneratorManager::$drivers
parameters	array	Generator-specific parameters passed to AI API or CLI process

API-based generators additionally include:

Key	Type	Description
api_key	string	Authentication key for API access
http_options	array	Guzzle HTTP client options (proxy, timeout, etc.)

CLI-based generators additionally include:

Key	Type	Description
binary	string	Path or command name for CLI executable
options	array	Command-line flags and their values

Sources: config/ai-commit.php78-219

API-Based Generators

API-based generators make HTTP requests to cloud-based AI services using SDK wrappers.

OpenAI Generator

The openai generator uses OpenAI's legacy completions API endpoint.

Configuration Block:

Parameters Reference

Parameter	Type	Default	Description	Config Line
model	string	text-davinci-003	Model identifier. Options: text-davinci-003, text-davinci-002, text-curie-001, text-babbage-001, text-ada-001	182
suffix	string|null	null	Text appended after completion	184
max_tokens	integer	600	Maximum tokens to generate	185
temperature	float	0.0	Sampling temperature (0.0-2.0). Lower values = more deterministic	186
top_p	float	1.0	Nucleus sampling probability mass	187
n	integer	1	Number of completions to generate	188
stream	boolean	true	Enable streaming responses via buildWriter() closure	189
logprobs	integer|null	null	Include log probabilities in response	190
echo	boolean	false	Echo back the prompt in response	191
stop	string|array|null	null	Sequences where API stops generating	192
presence_penalty	float	0	Penalty for new tokens (-2.0 to 2.0)	193
frequency_penalty	float	0	Penalty for repeated tokens (-2.0 to 2.0)	194
best_of	integer	1	Generate N completions and return best	195

Environment Variables:

• OPENAI_API_KEY: API authentication key (line 180)

Implementation Details:

• Parameters merged with prompt and UUID user identifier at app/Generators/OpenAIGenerator.php58-61
• Streaming handled by buildWriter() closure at app/Generators/OpenAIGenerator.php77-87
• Response text extracted via getCompletionMessages() at app/Generators/OpenAIGenerator.php69-72

Sources: config/ai-commit.php178-198 app/Generators/OpenAIGenerator.php26-88

OpenAI Chat Generator

The openai_chat generator uses OpenAI's chat completions API endpoint, supporting GPT-3.5 and GPT-4 models.

Configuration Block:

Parameters Reference

Parameter	Type	Default	Description	Config Line
model	string	gpt-3.5-turbo	Model identifier. Options: gpt-4, gpt-4-0613, gpt-4-32k, gpt-4-32k-0613, gpt-3.5-turbo, gpt-3.5-turbo-0613, gpt-3.5-turbo-16k, gpt-3.5-turbo-16k-0613	206
max_tokens	integer	600	Maximum tokens to generate	208
temperature	float	0.0	Sampling temperature (0.0-2.0)	209
top_p	float	1.0	Nucleus sampling probability mass	210
n	integer	1	Number of chat completion choices	211
stream	boolean	true	Enable streaming responses	212
stop	string|array|null	null	Up to 4 sequences where API stops	213
presence_penalty	float	0	Penalty for new topics (-2.0 to 2.0)	214
frequency_penalty	float	0	Penalty for repetition (-2.0 to 2.0)	215

HTTP Options: Configured at lines 201-203. Supports all Guzzle RequestOptions constants:

• PROXY: Proxy server URL (example: 'https://proxy.com/v1')
• TIMEOUT: Request timeout in seconds (global default: 120 at line 35)
• VERIFY: SSL certificate verification (global default: false at line 33)
• CONNECT_TIMEOUT: Connection timeout (global default: 30 at line 34)

Environment Variables:

• OPENAI_API_KEY: API authentication key (line 204)

Sources: config/ai-commit.php32-218

Ernie Bot Generators

Baidu's Ernie Bot (文心一言) generators provide two model variants: ernie_bot and ernie_bot_turbo.

Configuration Block (ernie_bot):

Configuration Block (ernie_bot_turbo):

Parameters Reference

Parameter	Type	Default	Description	Config Line
temperature	float	0.95	Sampling temperature affecting randomness	110 / 123
top_p	float	0.8	Nucleus sampling threshold	111 / 124
penalty_score	float	1.0	Penalty for repeated content (1.0-2.0)	112 / 125
stream	boolean	true	Enable streaming responses	113 / 126

Authentication: Both generators require OAuth-based authentication with two credentials:

• api_key: Baidu AI Cloud application API Key
• secret_key: Used for access token generation via OAuth endpoint

The authentication flow exchanges API key and secret key for an access token before making generation requests.

Environment Variables:

• ERNIE_API_KEY: API key for Baidu AI Cloud (lines 106, 119)
• ERNIE_SECRET_KEY: Secret key for OAuth token generation (lines 107, 120)

Sources: config/ai-commit.php103-128

Moonshot Generator

The moonshot generator integrates with Moonshot AI (Kimi) services.

Configuration Block:

Parameters Reference

Parameter	Type	Default	Description	Config Line
model	string	moonshot-v1-8k	Model identifier. Options: moonshot-v1-8k (8K context), moonshot-v1-32k (32K context), moonshot-v1-128k (128K context)	166
max_tokens	integer	600	Maximum tokens to generate in response	168
temperature	float	0.0	Sampling temperature (0.0-1.0). 0.0 = deterministic	169
top_p	float	1.0	Nucleus sampling probability threshold	170
n	integer	1	Number of completion choices to generate	171
stream	boolean	true	Enable Server-Sent Events (SSE) streaming	172
stop	string|array|null	null	Stop sequences for generation termination	173
presence_penalty	float	0	Penalty for introducing new topics (-2.0 to 2.0)	174
frequency_penalty	float	0	Penalty for token repetition (-2.0 to 2.0)	175

HTTP Options: Configured at lines 161-163. Example proxy configuration commented out.

Environment Variables:

• MOONSHOT_API_KEY: Moonshot API authentication key (line 164)

Sources: config/ai-commit.php159-177

CLI-Based Generators

CLI-based generators execute external command-line tools using Symfony Process.

CLI Generator Process Execution Flow

Configuration to Process Parameter Mapping:

Config Key	Process Constructor Parameter	Default	Description
parameters.cwd	$cwd	null	Working directory
parameters.env	$env	null	Environment variables
parameters.timeout	$timeout	120	Timeout in seconds
binary	$command[0]	Varies	Executable path
options	$command[1...]	Varies	CLI flags

Sources: app/Generators/BitoCliGenerator.php20-32 app/Generators/GithubCopilotCliGenerator.php20-28 app/Generators/GithubModelsCliGenerator.php20-32 config/ai-commit.php97-157

Bito CLI Generator

The bito_cli generator integrates with the Bito CLI tool.

Configuration Block:

Configuration Keys

Key	Type	Default	Description
driver	string	bito_cli	Generator driver identifier for GeneratorManager
binary	string	bito	Bito CLI executable path or command name resolved via env('BITO_CLI_BINARY')
prompt_filename	string	bito.prompt	Temporary file name for prompt storage during execution
options	array	See above	CLI flags and their values passed to bito command
parameters	array	See below	Symfony Process constructor parameters

CLI Options

All options are nullable and defined at lines 83-95. When non-null, they are appended to the command via ensureWithOptions() method as --flag=value.

Option	Description	Config Line
--context	Context information for the request	84
--edit	Enable interactive editing mode	85
--file	Input file path	86
--help	Display help information	87
--ignore	Patterns to ignore during processing	88
--key	API key override (if supported by Bito CLI)	89
--list	List available options or models	90
--model	Model selection	91
--prompt	Prompt override	92
--temperature	Temperature setting for generation	93
--version	Display version information	94

Process Parameters

Configured at lines 97-101. Passed to Symfony\Component\Process\Process constructor.

Parameter	Type	Default	Description
cwd	string|null	null	Working directory for process execution
env	array|null	null	Environment variables array for subprocess
input	string|null	null	Initial stdin input (overridden by setInput($prompt) at runtime)
timeout	integer	120	Process timeout in seconds before termination

Command Construction: The generator constructs the command array by calling ensureWithOptions([$binary]) which appends non-null option values from the options array. The prompt is then passed via Process::setInput($prompt) at app/Generators/BitoCliGenerator.php29

Environment Variables:

• BITO_CLI_BINARY: Override default binary path (line 81, default: 'bito')

Sources: config/ai-commit.php79-102 app/Generators/BitoCliGenerator.php20-32

GitHub Copilot CLI Generator

The github_copilot_cli generator uses GitHub CLI's Copilot extension.

Configuration Block:

Configuration Keys

Key	Type	Default	Description
driver	string	github_copilot_cli	Generator driver identifier for GeneratorManager
binary	string	gh	GitHub CLI executable path resolved via env('GITHUB_COPILOT_CLI_BINARY')
options	array	See above	CLI flags for gh copilot explain command
parameters	array	See above	Symfony Process constructor parameters

Command Structure

The generator constructs the command array as:

Executed via Symfony\Component\Process\Process.

CLI Options

Configured at lines 132-134. Options are appended via ensureWithOptions() method.

Option	Description	Config Line
--hostname	GitHub Enterprise Server hostname for self-hosted instances	133

Command Structure: The generator builds the command array as [$binary, 'copilot', 'explain', ...$options, $prompt] at app/Generators/GithubCopilotCliGenerator.php25 The prompt is passed as the final command argument, not via stdin.

Process Parameters: Configured at lines 136-140. Same structure as Bito CLI generator.

Parameter	Type	Default	Description
cwd	string|null	null	Working directory for gh CLI execution
env	array|null	null	Environment variables (must include GH_TOKEN)
input	string|null	null	Stdin input (unused, prompt passed as argument)
timeout	integer	120	Process timeout in seconds

Environment Variables:

• GITHUB_COPILOT_CLI_BINARY: Override default binary path (line 131, default: 'gh')
• GH_TOKEN: GitHub authentication token required by gh CLI

Sources: config/ai-commit.php129-141 app/Generators/GithubCopilotCliGenerator.php20-28

GitHub Models CLI Generator

The github_models_cli generator uses GitHub CLI's models feature.

Configuration Block:

Configuration Keys

Key	Type	Default	Description
driver	string	github_models_cli	Generator driver identifier for GeneratorManager
binary	string	gh	GitHub CLI executable path resolved via env('GITHUB_MODELS_CLI_BINARY')
model	string	openai/gpt-4.1-mini	Model identifier resolved via env('GITHUB_MODELS_CLI_MODEL')
options	array	See above	CLI flags for gh models run command
parameters	array	See above	Symfony Process constructor parameters

Command Structure

The generator constructs the command array as:

The prompt is passed via stdin using Process::setInput($prompt).

CLI Options

Configured at lines 146-151. Options are appended via ensureWithOptions() method before stdin input.

Option	Type	Default	Description	Config Line
--max-tokens	integer|null	null	Maximum tokens to generate in response	147
--system-prompt	string	You are a git commit message generator.	System prompt providing model context	148
--temperature	float|null	null	Sampling temperature for generation	149
--top-p	float|null	null	Nucleus sampling probability threshold	150

Command Structure: The generator builds the command array as [$binary, 'models', 'run', $model, ...$options] at app/Generators/GithubModelsCliGenerator.php27 The prompt is passed via stdin using Process::setInput($prompt) at line 29.

Model Configuration: The model parameter (line 145) specifies the AI model identifier in format provider/model-name. Default is openai/gpt-4.1-mini.

Process Parameters: Configured at lines 152-157. Same structure as other CLI generators.

Parameter	Type	Default	Description
cwd	string|null	null	Working directory for gh CLI execution
env	array|null	null	Environment variables (must include GH_TOKEN)
input	string|null	null	Initial stdin (overridden by setInput($prompt))
timeout	integer	120	Process timeout in seconds

Environment Variables:

• GITHUB_MODELS_CLI_BINARY: Override default binary path (line 144, default: 'gh')
• GITHUB_MODELS_CLI_MODEL: Override default model (line 145, default: 'openai/gpt-4.1-mini')
• GH_TOKEN: GitHub authentication token required by gh CLI

Sources: config/ai-commit.php142-158 app/Generators/GithubModelsCliGenerator.php20-32

Configuration Hierarchy

Generator configurations are loaded and merged according to the four-tier priority system described in Configuration Management.

Configuration Loading and Merging Flow

Configuration Resolution Example:

For generator openai_chat:

1. PHP defaults: config/ai-commit.php:199-218 (base: temperature: 0.0)
2. Global JSON: ~/.ai-commit/.ai-commit.json (override: temperature: 0.5)
3. Local JSON: ./.ai-commit.json (override: model: "gpt-4")
4. CLI flag: --generator openai_chat (final selection)

Result: {model: "gpt-4", temperature: 0.5, ...} passed to OpenAIChatGenerator constructor.

Sources: app/Commands/CommitCommand.php39-215 app/Generators/OpenAIGenerator.php26-29

Environment Variable Support

All generator configurations support environment variable substitution using the env() helper function. This allows sensitive data like API keys to be stored outside the configuration file.

Common Environment Variables

Variable	Used By	Purpose
OPENAI_API_KEY	openai, openai_chat	OpenAI API authentication
ERNIE_API_KEY	ernie_bot, ernie_bot_turbo	Baidu AI Cloud API key
ERNIE_SECRET_KEY	ernie_bot, ernie_bot_turbo	Baidu AI Cloud secret key
MOONSHOT_API_KEY	moonshot	Moonshot AI API key
BITO_CLI_BINARY	bito_cli	Bito CLI executable path
GITHUB_COPILOT_CLI_BINARY	github_copilot_cli	GitHub CLI executable path
GITHUB_MODELS_CLI_BINARY	github_models_cli	GitHub CLI executable path
GITHUB_MODELS_CLI_MODEL	github_models_cli	Default model identifier
GH_TOKEN	github_copilot_cli, github_models_cli	GitHub authentication token

Setting Environment Variables:

Sources: config/ai-commit.php81-218

HTTP Options Configuration

API-based generators support HTTP client customization through the http_options array. These options are passed directly to the Guzzle HTTP client.

Global HTTP Options

Global HTTP options apply to all API-based generators unless overridden:

Generator-Specific HTTP Options

Individual generators can override global HTTP options:

Common HTTP Options

Option	Type	Description
VERIFY	boolean|string	SSL certificate verification. false disables, string provides CA bundle path
PROXY	string|array	Proxy server configuration
TIMEOUT	float	Request timeout in seconds
CONNECT_TIMEOUT	float	Connection timeout in seconds
HEADERS	array	Additional HTTP headers
DEBUG	boolean	Enable debug output

Example - Using a Proxy:

Sources: config/ai-commit.php32-36 config/ai-commit.php161-163 config/ai-commit.php201-203

Configuration Management Commands

Generator configurations can be managed using the config command. See Configuration Management for complete documentation.

Examples:

Sources: app/Commands/CommitCommand.php200-215