Skip to content

reword

regit reword [<commit-ish>…] [flags]

Regenerate commit message(s) with an LLM. Each targeted commit's diff (plus optional context such as the README or file states) is sent to the model, which proposes a new title and body; only the message changes — trees, dates, and authorship are untouched. Headless runs request exactly one proposal per commit and preview it until --execute.

Positional arguments

Argument Default Description
<commit-ish>… HEAD The commit(s) to reword.

Flags

Flag Type Default Description
--llm VALUE — provider/model, e.g. ollama/qwen2.5-coder.
--agent VALUE — Agent CLI to drive instead of HTTP.
--agent-args VALUE — Replace default agent args.
--agent-scan BOOL false Agent may browse repo files read-only for extra context (requires --agent).
--llm-base-url VALUE — Override the OpenAI-compatible endpoint.
--prompt VALUE — Short guidance passed to the LLM.
--context A,B — Context keys to include, e.g. readme,filestates.
--context-lines N 0 git diff -U depth (0 = config default).
--no-cache BOOL false Ignore cached proposals.
--clear-cache BOOL false Delete the reword cache and exit.
--repo VALUE . Path to the git repository.
--execute BOOL false Apply the reword (default: dry-run preview).
--force BOOL false Allow rewriting already-pushed commits.
--sign BOOL false Re-sign rewritten commits with your configured signing key.
Examples

Correct:

# preview a new message for HEAD
regit reword HEAD --llm=ollama/qwen2.5-coder

# reword two commits and apply
regit reword HEAD~2 HEAD --llm=openai/gpt-4o --execute

Incorrect:

# BAD: --agent-scan without --agent is an error — scan mode drives an agent CLI
regit reword --agent-scan HEAD

Limitations

  • Headless always requests exactly one variant (the TUI browses several).
  • --agent-scan requires --agent.