FAQ & troubleshooting

Do I need an LLM?

No. Template mode (the default) is deterministic and dependency-free — perfect for CI. The LLM (--model) is optional polish via a local Ollama model.

"No commits found for the given range"

Check the range is valid (git log <range> works). With no range, gitscribe uses <latest tag>..HEAD; if there are no tags, it uses all history.

My commits show up under "Other"

They don't follow Conventional Commits (feat:, fix:, feat(scope)!: …). That's fine — non-conforming commits are grouped under "Other". Adopt the convention for cleaner output.

The LLM invented a feature

Unlikely, because the model receives the already-categorized commit list, not raw diffs — it rewrites, it doesn't research. If it happens, use template mode, or a stronger local model.

Can I use it in CI?

Yes — use template mode (no --model) so it's deterministic and needs no Ollama. See How it works & CI usage.

Which Ollama model should I use?

Any pulled chat model works; smaller models are faster. Compare them with ollabench.