套餐优惠

Addon T

Addon T 套餐指南

套餐指南

功能附加费用比较 (-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

编辑内容

编辑公式

Generation

我的 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

查看从图片创建模板的请求处理结果，以及实际 API 调用、令牌数和费用。每个请求最多包含 5 页。

图片→模板用量

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

按日期的成功与失败

按日期查看各菜单的成功与失败，确认失败账号与原因，修复后向这些账号发送解决通知。

将向 {count} 个账号发送邮件。

解决通知

未保存原始错误。请查看显示的说明和该任务的工作进程日志。

故障运营控制台

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

管理采集来源、运行流水线、手动发布资讯。

填入标题和网址，AI 会生成摘要、重要度与分类草稿。确认后再发布。

发布标题

原文标题

请原样粘贴发布标题

手动录入

执行一次：采集 → 去重 → AI 筛选 → 官方原文核对 → 发布 → 通知。

采集流水线

来源

只登记提供订阅源且 robots.txt 允许的来源。

采集来源

教育资讯运营

按日/周/月/自定义区间查看我的 DB 使用趋势。

我的 DB 使用趋势

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

按名称

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

Syntactic Analysis Trend

Admin-only Claude document conversion process

Read the whole page, divide questions, exclude doodles and redraw figures as SVG.

Document OCR

Convert the generated SVG into a GeoGebra file.

SVG to GeoGebra (optional)

Create a solution and answer for each question, plus a solution SVG if needed.

Generate solutions (optional)

Tag curriculum, format and concepts; save question number and source in the private bank.

Tag and save each question

Document conversion process Claude 261020

The run log shows LLM calls and tokens by History ID. Its cost total includes only priced models; currently unpriced calls such as Gemini Nano Banana 2.1 image generation need a separate estimate.

Cost per run

Document conversion flow and stage models as of October 20, 2026

OCR and problem tagging may use GPT-5.5 on the final retry after Gemini failures. Figure detection or OCR may use GPT-6.1 Sol on the final retry after a Gemini 503 error. Retry calls also add cost.

Retries and fallback models

Standard conversion connects figure references by detection_id and marks problem boundaries with problem-break. Solution mode maps figures to problem crops by geometry.

Linking figures to problems

These calls run when enabled and a target figure exists. GeoGebra generation also requires USE_AUTOMATIC_GGB_GENERATION on the server.

Optional figure processing

Problems are split into crops, tagged, OCRed, and solved individually. Successful problems are stored separately in problem_banks.

Generated solution mode

Document type

Identify candidates for SVG and GeoGebra conversion.

Figure classification

Find figure regions and detection IDs in each source image.

Figure detection

Refine OCR with figure crops when figures exist.

Figure-aware OCR refinement

Convert SVG when both the user option and server setting are enabled.

GeoGebra conversion

When enabled, remove handwriting or doodles from detected figures.

Handwriting removal

Extract text, formulas, problem boundaries, and figure references.

Initial OCR

Create per-problem crops, including problems spanning pages.

Problem segmentation

Per-problem solution

Extract problem text and tags from each crop.

Per-problem tagging and OCR

Reconstruct selected figures as vector SVG.

SVG reconstruction

Each PDF page or uploaded image is processed as a source. Results are stored by source in typing_history; this path does not create one database record per problem.

Standard conversion

문서변환 프로세스 261020

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.

查看单个用户在生成/审阅/我的 DB 使用上的趋势。

用户统计

积分变动记录

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

Add custom feature

允许的附加功能

Deduct User Credits

Grant User Credits

Change User Feature Restrictions

Change User Membership

历史订阅记录

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 argon1014

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.