v0.1.1

Fixed

• Toponymy fitting now avoids a large fixed-width Unicode array allocation that could cause memory errors on large corpora.
• Notebook and session resume now load translated, tokenized, and author-disambiguated snapshots even when earlier-stage frames are already in memory.

Changed

• The OpenRouter notebook example now uses Gemini Flash 3 for translation and topic labeling, with Qwen3 embeddings.