Summary

Allow llm-commit to use environment variables for setting default values for command-line flags.

Expected Benefits:

• Reduces the need to repeatedly specify options for frequent users.
• Provides a flexible mechanism for customizing llm-commit behavior.
• Ensures compatibility with existing workflows by preserving explicit command-line argument precedence.

Problem statement

Currently, every execution of llm commit requires specifying command-line arguments for options such as model selection, token limits, temperature settings, and truncation behavior. This can be cumbersome, especially for users who consistently use the same settings.

While a previous update introduced LLM_COMMIT_STYLE for defining a default commit style, a broader mechanism for setting defaults via environment variables is still missing.

Proposed solution

Introduce support for environment variables corresponding to existing command-line flags. If an environment variable is set, it will be used as the default value unless overridden by an explicit command-line argument.

Environment Variables to Support:

• LLM_COMMIT_MODEL → Default model (--model)
• LLM_COMMIT_MAX_TOKENS → Default max tokens (--max-tokens)
• LLM_COMMIT_TEMPERATURE → Default temperature (--temperature)
• LLM_COMMIT_TRUNCATION_LIMIT → Default truncation limit (--truncation-limit)
• LLM_COMMIT_NO_TRUNCATION → Disable truncation (--no-truncation)
• LLM_COMMIT_HINT → Default hint (--hint)

Priority Rule

• Command-line arguments always override environment variable values.
• If no environment variable is set, fall back to tool defaults.

Example Usage:

# Set defaults in shell profile (e.g., .bashrc, .zshrc) or in current directory with direnv (.env)
export LLM_COMMIT_MODEL="4o"
export LLM_COMMIT_MAX_TOKENS=1000
export LLM_COMMIT_TEMPERATURE=0.95
export LLM_COMMIT_HINT="We are working on data-science related projects."

# Running without command-line flags uses these defaults
llm commit

# Override behavior using flags
llm commit --model "gpt-3.5-turbo" --temperature 0.9