mirror of
https://github.com/rybaier/obsidian-chat-splitter.git
synced 2026-07-22 06:09:04 +00:00
Remove MAX_USER_PARAGRAPH_LENGTH -- character limits are the wrong heuristic for distinguishing user from assistant text. Content-based checks (markdown, URLs, em-dashes, continuation patterns) are the correct approach. Remove overly broad "it is/this means" pattern that falsely matched user text. Add common abbreviation prefix handling (St, Dr, Mr, etc.) to normalizeAbbreviations() so "St. Kitts" stays together during entity extraction sentence splitting. Changes: - Remove MAX_USER_PARAGRAPH_LENGTH constant and length check - Remove (that|this|it) pattern from ASSISTANT_CONTINUATION_PATTERNS - Add abbreviated prefix period stripping in normalizeAbbreviations() BREAKING CHANGE: none |
||
|---|---|---|
| .. | ||
| decisions.md | ||
| GAP_ANALYSIS.md | ||
| IMPLEMENTATION.md | ||
| PLAN.md | ||