0.1.1

Changed

• The README now shows screenshots of ghost-text suggestions (continuing a sentence, a list item, the next word) and of the settings.
• The troubleshooting guide ends with a “Getting help” section that links to the issue tracker.

0.1.0

Added

• Initial scaffold: manifest, build, vendored kit modules, settings model.
• Ghost-text suggestion for the current line, shown after a typing pause; Tab, Right arrow and Escape accept, accept a word and dismiss.
• Two request paths: chat and fill-in-the-middle (FIM, Qwen2.5-Coder), chosen automatically per model or fixed in the settings.
• Endpoints, keys and model list from the LLM Endpoint Manager; sampling values from the kit table, mode "complete".
• Context window with redaction of private keys, e-mail addresses and tokens before the cut; frontmatter and query blocks are never sent.
• Exclusion of notes by .gitignore-style patterns and by ghostline: false in the properties.
• Commands for accept, accept word, dismiss, suggest now and on/off; status bar item with state, last latency and errors.
• Settings tab with help row, German and English interface.
• Requires Obsidian 1.11.4 or newer: the endpoint keys are read from the Obsidian keychain (app.secretStorage).
• npm run smoke:gui (GUI smoke against a running Obsidian) and npm run latency (latency measurement, only after agreement).

Fixed

• While a suggestion streams in, a repeated start of the sentence is no longer shown, and Tab never inserts text that the suggestion no longer shows.
• The "Last request" line in the Request section now shows the parameters of the last request (sampling values and stop list, never note text).
• Only the pane with the cursor asks for suggestions; a second pane on the same note stays quiet.
• The command "Turn Ghostline on or off" is now "Turn suggestions on or off" (Obsidian already shows the plugin name in front).