LLM Answer Grader v0.5.0 — accent keyboard

Optional lexilogos-style accent keyboard under the answer box, enabled per profile from a preset-layout dropdown in Settings (currently French: À Â Æ Ç É È Ê Ë Î Ï Ô Œ Ù Û Ü Ÿ over the matching lowercase row). Click a key to insert the letter at the cursor; works in the follow-up field too. Off by default.

LLM Answer Grader v0.4.2

Fixes the follow-up answer's "Answer" label rendering on the same line as the response text.

LLM Answer Grader v0.4.1

Fixes the settings dialog layout: the per-profile model override section no longer overlaps the grading instructions on smaller windows (profile editor is now scrollable with proper minimum sizes).

LLM Answer Grader v0.4.0 — settings UI, follow-ups, per-profile models

The AnkiWeb-readiness release:

• Visual settings dialog (Tools → LLM Answer Grader Settings…): pick note types and fields with checkboxes instead of editing JSON; per-profile model overrides; raw-JSON editor still available
• Ask follow-up questions about any grading ("why is that wrong?") — answered in context
• Per-profile provider/model: e.g. local Ollama for easy decks, Claude for hard ones
• Hardening: config errors can never break card display (fail-open); first-run setup dialog; keyboard focus released after grading so rating shortcuts work immediately; multi-template notes graded with template context

Install: download llm_answer_grader.ankiaddon → Anki → Tools → Add-ons → Install from file.

LLM Answer Grader v0.3.1

Configurable display labels for the suggested rating and verdict badge (rating_labels / verdict_labels config) — for non-English Anki UIs and relabeled answer buttons. The four suggestion categories still map to Anki's fixed ease values.

LLM Answer Grader v0.3.0 — bring your own model

Adds an OpenAI-compatible provider alongside Claude:

• provider: openai_compatible + openai_base_url works with OpenAI, OpenRouter, Groq, Gemini's compat endpoint, and local Ollama / LM Studio (no API key needed — fully private grading)
• Automatic capability negotiation (json_schema → json_object → prompt-enforced JSON) and tolerant parsing, so small local models work too
• Claude remains the default and recommended grader

Install: download llm_answer_grader.ankiaddon → Anki → Tools → Add-ons → Install from file.

LLM Answer Grader v0.2.0

First packaged release: type an answer on any configured card, get graded by Claude inline in the reviewer.

• Profile-based card targeting (any note type / language / task)
• Verdict, 0-100 score, corrected version, specific feedback, suggested rating
• Structured outputs + adaptive thinking via the Anthropic Messages API
• Bring your own API key (see README)

Install: download llm_answer_grader.ankiaddon → Anki → Tools → Add-ons → Install from file. Requires Anki 23.10+ desktop (tested on 25.09).