Web Progress, Results, and History

After a translation starts, this guide explains how to watch task progress, preview and download results in the browser, and manage the history records stored on the server. Progress is shown in real time through the log box rather than a percentage bar; the results list exists only in the current browser; history is stored on the server, so the same account can see it from any browser. Uploading, configuration, and starting a translation are covered in Upload, config, and translate, sessions and language switching in Login, language, and session, and administrator management of all history and tasks in Administrator interface.

UI and API scope

• The “results list” is a temporary view inside the current browser; entries are saved in localStorage.translationResults (as blob URLs). It is not server history and cannot be recovered after clearing browser data or switching browsers.
• “History records” are written automatically by the server after a successful translation and are isolated per user; a regular user sees only their own history, and whether it can be viewed, downloaded, or deleted depends on permissions. The corresponding endpoints return 403 without permission.
• Progress is presented through streaming progress frames, task-log polling every 500 ms, and the log box; there is no percentage progress bar.
• This guide covers web user operations only. The frame format and the history/download-ticket HTTP contracts live in Streaming protocol and History, files, and download tickets.

Use it in the Web UI

View translation progress

1. After clicking the “开始任务” (start task) button, the “日志输出” (log output) area on the right streams progress messages in real time.
2. Single-file normal translation uses the streaming endpoint; progress messages come from two sources:
   • progress frames in the stream response (status=1), for example “加载图片中...”, “初始化翻译器...”, “翻译中...”, and “Done!”;
   • after the task starts, the browser polls /api/logs?limit=200&task_id=... every 500 ms to fetch fuller task logs (detection, OCR, translator calls, and so on) filtered by timestamp.
3. Multi-file normal translation is chunked by cli.batch_size; the log shows “批次 N/M: X 个文件”, “批量翻译中...”, and finally “所有任务完成！”.
4. Queuing and concurrency: when the concurrency limit is reached, a progress frame sends “排队中... (前面还有 N 个任务)”, and after acquiring a slot it sends “获得翻译槽位，开始处理...”.
5. Failure and cancellation: a progress frame with status=2 carries the error into the log; an administrator-cancelled task ends with 499.
6. Expired session: when log polling receives 401, polling stops and the UI shows “登录状态已过期，已停止实时日志轮询。当前任务可能仍在继续，请重新登录后再查看日志。”.

Preview and manage results

1. Each finished image is added to the “results list” automatically, newest first; ZIP entries show 📦 and images show 🖼️.
2. Each entry has “查看” (view), “下载” (download), and “×” (delete) actions: view opens the image in a new tab, download saves it under its original filename, and delete releases the corresponding blob URL.
3. When image results exist, “🔍 展开图片查看器” appears; the toolbar also has “打包下载” (download all) and “清空” (clear). “打包下载” packs all results into translation_results_<timestamp>.zip with JSZip.
4. The image-viewer modal shows thumbnails on the left and the large image on the right, with a “下载” (download) action; the mobile layout supports pinch-to-zoom.
5. Clearing asks for confirmation (“确定要清空所有翻译结果吗？”), then releases all blob URLs and empties the list.

Open the history gallery

1. On page load, /api/history is called for the current user’s history; the “history” area shows only the 5 most recent entries (timestamp + “N 个文件”).
2. Clicking “📷 打开相册” or “📷 查看全部 (N)” opens the “翻译历史相册” modal, where history is grouped by date into cards.
3. Each card shows a thumbnail, time, and file count and can be checked; both thumbnail and large-image requests carry X-Session-Token to reach the protected image endpoints.
4. Clicking “查看” on a card opens a fullscreen image viewer with left/right arrow navigation and Esc to close.

Download history

1. Single record: fetch the session detail first; with only 1 file, request a single-file download ticket; with multiple files, request a ticket for the whole session’s ZIP.
2. After checking several records, “下载选中” (download selected) requests a batch ticket; without a selection, “下载全部” (download all) packs the entire history into one ZIP (filename prefixes history_selected / history_all).
3. Tickets are short-lived URLs (5 minutes by default), and the server cleans up the temporary ZIP after the download.

Delete history

1. Click 🗑 on a gallery card, confirm with “确定要删除这条翻译历史吗？”, and DELETE /api/history/{token} is called.
2. Deletion removes both the server session directory and the index record; the local gallery list refreshes.
3. Without delete permission the endpoint returns 403 and the UI shows “删除失败”.

Requests and data flow

Stream progress frames and log polling

Single-file normal translation calls POST /translate/with-form/image/stream; the response is a custom stream of “1-byte status + 4-byte length + data” frames: status=1 is a progress JSON (stages include task_id, start, image_loading, translator_init, translating, transforming, sending, complete, plus queued and slot_acquired while waiting), status=0 is the result image data, and status=2 is an error. The frontend parses the frames, writes each message into the log box, and uses the value of the task_id stage as the current task ID.

Multi-file normal translation uses POST /translate/batch/images: the body carries base64 images, config, batch_size, and filenames, and the response is a ZIP with the custom X-Content-Type: application/zip header; the frontend unpacks it with JSZip and adds each image to the results list. The batch request sets a 30-minute timeout in the frontend with AbortController.

Results list and browser storage

Every completion (single-file stream, batch unpack, or other paths that return a blob) calls addResult(), which appends {id, filename, imageData, type, timestamp} to resultsList and writes it to localStorage.translationResults. imageData is a blob URL created with URL.createObjectURL().

Blob URLs are valid only within the page session that created them: after a refresh or in another browser, preview/download of old entries usually no longer works. History is the durable, cross-browser, cross-session storage.

Server history and download tickets

Only requests that go through the while_streaming pipeline write server history automatically: in the web UI, single-file “normal translation” (/translate/with-form/image/stream) and batch translation (one record per image, token {task_id}_{i}) are saved; export original/translated, import-and-render, colorize-only, upscale-only, and inpaint-only use non-streaming endpoints in the web UI and do not write history (the /stream variants of those workflows do). Saving uses task_id as session_token, copies the result image into a session folder under the result directory, and writes metadata.json plus an index record; a failed save only logs a warning and does not interrupt the main flow.

History listing, thumbnails, large images, downloads, and deletion all require login (requests carry X-Session-Token). Downloads never expose file paths directly: first a short-lived ticket is requested — one ticket for a single file or a whole session, and batch-download-ticket for multiple records — then GET /api/history/downloads/t/{ticket} serves the file within 5 minutes by default, and the server cleans up the temporary ZIP afterward.

flowchart LR
    A["Click start task"] --> B{"Normal translation with multiple files?"}
    B -->|yes| C["POST /translate/batch/images"]
    B -->|no| D["POST /translate/with-form/image/stream"]
    C --> E["Log: batch N/M, batch translating"]
    C --> F["ZIP response<br/>JSZip unpack → add to results list"]
    D --> G["Progress frame status=1<br/>queued / task_id / translating / complete"]
    D --> H["Result frame status=0 → add to results list"]
    D --> I["Error frame status=2 → log error"]
    G --> J["Log box shows in real time"]
    G --> K["Poll /api/logs every 500 ms"]
    K --> J
    F --> L["Results list (localStorage.translationResults)"]
    L --> M["View / single download / download all / clear"]
    D --> N["Server saves history<br/>session_token = task_id"]
    C --> N
    N --> O["/api/history → sidebar 5 most recent"]
    O --> P["History gallery (grouped by date)"]
    P --> Q["View large image / download / delete"]
    P --> R["Short-lived download ticket → ZIP"]

The diagram describes the data flow and does not claim that every run has history: saving is best-effort and only warns on failure; export/import/colorize/upscale/inpaint go through non-streaming endpoints in the web UI and produce no history entry, and the results list always exists only in the current browser.

Permissions, security, and limits

• The results list and server history are two independent mechanisms: the former lives in localStorage.translationResults (blob URLs), the latter in the server result directory and translation_history.json. Do not mix them up.
• Progress visibility depends on the session: once session_token expires, streaming requests, history endpoints, and log polling all return 401; polling stops automatically and prompts a re-login.
• History is isolated per user: a regular user can only view, download, and delete their own history; view/delete permissions come from the account permissions. For the administrator view, see Administrator interface.
• Download tickets have a TTL (5 minutes by default) and their temporary ZIPs are cleaned up; request a new ticket after long idle time.
• The 30-minute frontend timeout for batch requests matches the server’s timeout_keep_alive=1800, but it does not mean every image in the batch succeeded; cancellation and failures are handled by the server task machinery, see Translation endpoints.
• Log content may contain business text and paths; remove request bodies, log messages, paths, and credentials before sharing, see Privacy, cleanup, and log sharing.

See the reference index: UI Options Reference.