プラン特典

Addon T

追加機能利用者向けのオプションプラン

Addon T料金プラン案内

料金プラン案内

機能別追加料金比較 (-25%)

Review successful usage only. Math similar-question generation applies academy contract prices first; Questi uses actual user plan/generation prices.

{billableCount} billable items / {totalCount} total items

Billing by feature

Billing Invoice

View Book Chat web usage trends by day/week/month/custom range.

Book Chat Trend

View library/Book Chat OCR analysis usage trends by day/week/month/custom range.

Book OCR Analysis Trend

Admin Console

View cramming usage trend over day/week/month/custom range.

Cramming Trend

View generation/review trends and key operational metrics in one place.

Analyze similar-problem generation trends

内容を編集

適用を押すとページに反映されます。

数式を編集

Generation

マイDB

Review

Manage user list and feature restrictions

User Controls

Persisted activity in the selected period, using the measure shown on each card. Generation, editing and grading requests count separately.

Admin Dashboard

View 1:1 tutor explanation usage trend over day/week/month/custom range.

1:1 Tutor Trend

画像からテンプレートを作成するリクエストの処理結果と、実際のAPI呼び出し・トークン数・費用を確認します。1件のリクエストには最大5ページが含まれます。

画像→テンプレートの使用量

View document conversion usage trend over day/week/month/custom range.

Doc Conversion Trend

Review analysis times, requesters, indexing attempts and errors per document. Tokens and costs not reported by File Search remain unavailable.

Book OCR analysis

Review segmentation, generation and modification tokens, calls, agents, costs and errors per solution history.

Handwritten solutions

Historical usage statistics. Past token usage and costs are not backfilled.

User downloads

View essay correction usage trend over day/week/month/custom range.

Essay Correction Trend

View exam similarity analysis usage trends by day/week/month/custom range.

Exam Similarity Analysis Trend

View exam analysis usage trend over day/week/month/custom range.

Exam Analysis Trend

Review and collect school and mock exam inputs from document conversion, exam analysis, and similarity analysis.

Review and collect exam material

Subject

Exam material collection

日付別の成功・失敗

日付ごとにメニューの成功・失敗を確認し、失敗したアカウントと原因を調べ、修正後に解決通知を送ります。

{count} 名にメールを送信します。

解決通知

障害運用コンソール

View find question (source) usage trend over day/week/month/custom range.

Find Question Trend

Usage and request counts for the selected period.

AI mathematics problem generation

View per-run records and existing usage statistics in one place.

View historical usage and request counts for this subject. Legacy and V2 are different views, not a split by creation date.

View English similar problem generation usage by hour, day, week, month, or a custom period.

English Similar Problems Usage Trend

View Korean similar problem generation usage by hour, day, week, month, or a custom period.

Korean Similar Problems Usage Trend

View math similar problem generation usage by hour, day, week, month, or a custom period.

Math Similar Problems Usage Trend

Inspect each history's actor, LLM usage, cost, and errors from stored operational metadata.

Individual call records load only when details are expanded.

LLM call records

Run details

Errors across all matching runs, ordered by frequency.

Most frequent errors

Runs with stored observability metadata / all runs

Operations overview

Run history

Problem generation runs

Science problem generation

View STEAM problem generation usage trends by day/week/month/custom range.

STEAM Problem Trend

View similar-problem generation trends by day/week/month/custom period.

Generation Usage

View handwritten solution usage trend over day/week/month/custom range.

Handwritten Solution Trend

View Labeling API usage trends by day/week/month/custom range.

Labeling API Trend

View math manipulative usage trend over day/week/month/custom range.

Math Tool Trend

収集出典を管理し、パイプラインを実行し、ニュースを直接登録します。

タイトルとURLを入れるとAIが要約・重要度・分類の下書きを作ります。確認後に掲載してください。

掲載タイトル

原文タイトル

発表タイトルをそのまま貼り付けてください

直接登録

収集 → 重複除去 → AI選別 → 公式原文照合 → 掲載 → 通知を一度実行します。

収集パイプライン

出典

フィードがあり robots.txt が許可する出典のみ登録します。

収集出典

教育ニュース運用

日/週/月/任意期間でマイDB利用推移を確認します。

マイDB利用推移

Legacy shows historical exam analysis usage. V2 combines measured tokens, cost, agents, and errors from generation, modifications, and retries for each history. Earlier calls are not backfilled.

View review usage trends by day/week/month/custom period.

Review Usage

View item review and shortcut usage trend over day/week/month/custom range.

Review/Shortcut Trend

Combine segmentation, child video generation, and retry tokens, LLM/TTS calls, API cost, and errors per original request. API cost covers metered calls; rendering and storage costs are excluded.

Legacy shows historical video generation requests and completed videos. V2 shows newly measured API usage per history. Earlier calls are not backfilled.

Completed videos and original requests grouped by request time. Child histories do not duplicate the request count.

Video generation

View syntactic analysis usage trend over day/week/month/custom range.

Syntactic Analysis Trend

Inspect each conversion’s operator, results, tokens, API cost, and errors.

View existing document conversion usage and request counts over time. Legacy and V2 are different views.

特定ユーザー1名の生成/レビュー/マイDB利用推移を確認します。

ユーザー統計

クレジット変動履歴

Browse paginated users, grant credits, and manage per-user feature restrictions.

Add custom feature

許可された追加機能

Deduct User Credits

Grant User Credits

Change User Feature Restrictions

Change User Membership

過去の購読履歴

All users

Users

User usage summary

View lecture video analysis usage trends by day/week/month/custom range.

Lecture Video Analysis Trend

View English vocabulary builder usage trends by day/week/month/custom range.

Vocab Builder Trend

View educational webtoon usage trend over day/week/month/custom range.

Edu Webtoon Trend

Gemini 3.8 Flash plans SVG and GGB corrections; gpt-image-2.5-flare draws the corrected figures. GGB corrections reference the previous SVG correction plan and image. Figures are stored as images, without individually editable points or lines.

Option 2-1 · Gemini + GPT Image 2.5 Flare

Highest difficulty design, independent review and accurate figures · isolated prompts

Advanced: import killer configuration

Killer option

Keep the original image and OCR; skip original SVG reconstruction and cached SVG reuse. Generate and validate the transformed figure with the same option 1-1 settings.

Process argon1015

Previous record: QuestiGeo 1014_1023

Classify the selected crop as geometry, algebraic, physics, or other and cache the result. Classification no longer shows a warning.

Classify before upload

Save each figure crop. Show the figure count multiplied by 500 credits and start processing only after the user agrees to the deduction.

Crop each figure and confirm 500 credits

Convert SVG1 to GeoGebra XML and GGB1, and show GGB1 as soon as it is ready.

SVG1 to GGB1

Load GGB1 and inspect its quality. Create GGB2 and replace the active GGB only when the corrected result is selected; retain GGB1 in history.

Load and review GGB; optionally create GGB2

Store the original image's problem OCR results in a separate JSON artifact.

Save problem OCR separately

Save SVG and GGB candidates with validation history, then settle the credit reservation for completed figures.

Save results and settle credits

Generate SVG1 directly from the crop without a construction plan, then show its thumbnail immediately.

Source crop to SVG1 and thumbnail

Compare the source crop with rendered SVG1. Create SVG2 and replace the active SVG only when a correction is needed; retain SVG1 in history.

Compare source and optionally create SVG2

QuestiGeo Process

Previous record: Process argon1014

QuestiGeo 1014_1023

Reuse the preflight classification cache when present. Without a cache, or when an old other-type result needs checking, classify as geometry, algebraic, physics, or other.

Classify figure type

The direct-upload worker detects the figure region within the chosen crop before OCR. It stops if several figures are found, then normalizes the crop to PNG. A cached detection skips the model call.

Detect and normalize figure crop

Generate GeoGebra XML from the chosen SVG and source image. Production pipeline 091720 selects the SVG candidate before conversion and may call the model again for fidelity repair.

Convert SVG to GeoGebra XML

Static integrity checks and headless GeoGebra loading/capture use no LLM. Visual review of source, SVG, and GGB capture, plus one possible repair, use gemini-3.8-flash. The default gate mode warns and can select the best candidate.

Load, review, and repair GGB

Read problem text, figure flags, and an optional bounding box from the original image. When several figures are present, use the chosen crop for context.

OCR and figure hints from original

Send the cropped image to /figures/classify-sources to classify geometry, algebraic, physics, or other. Pass the result to /figures and save it as the worker classification cache.

Classify the crop before registration

When needed, segment the original page and match the chosen figure to its problem. The selected-crop path for multiple figures can skip this call.

Segment problems and select context

Method A creates an optional matplotlib reference and structure plan. It was enabled in production. Other-type figures skip it; a recitation failure may use the gpt-6.1-sol fallback.

Method A Python reference and plan

Save SVG, GGB, XML, capture, thumbnail, checkpoints, and validation history, then finalize the conversion and credits. Saving calls no LLM.

Save artifacts and settle credits

If another SVG candidate exists, compare both with the source and select the more faithful one. A single candidate needs no comparison call.

Compare SVG candidates

Generate the first SVG candidate from the source figure, problem text, and optional Python reference, then parse, normalize, and check its geometry.

Reconstruct source SVG

When review feedback or the candidate policy calls for another version, generate a new SVG candidate from the original. Cache and validation results can skip this call.

Regenerate an SVG candidate

Compare the rendered SVG with the original for missing objects, intersections, labels, and typography, and provide feedback when another candidate is needed.

Review SVG and regenerate if needed

The browser crops the selected image and uploads both crop and original to storage. Browser processing and file transfer use no LLM.

Select, crop, and upload

皆さんの学習を担当するQuestiの先生方をご紹介します！
どの先生があなたにぴったり合うでしょうか？