Open-source project
shiyutim/tickets avatar
shiyutim/tickets

tickets: plaintext cookies, a frozen clock offset, and a monitor that does not order

A ticket-grabbing tool for concerts on Damai and Bilibili Member Purchase via their APIs, with remaining-ticket monitoring and WeChat notifications.

3,586 stars456 forksRustMIT

At a glance

What is it?
A desktop workbench that schedules ticket purchases on two Chinese ticketing platforms from a Rust backend, monitors for remaining stock, and sends alerts through a chat bot. The author states his own limits plainly: payment happens on the official site, the stock monitor never probes availability by attempting an order, and the interfaces may change at any time.
Who is it for?
Read this as a well-documented automation project rather than as a reliable way to get a ticket, because the README itself says the interfaces may change, that simulated testing cannot substitute for verification with a real account at the real sale time, and that no purchase is guaranteed. Three things to weigh before running it on a machine that matters.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 63 days ago.
What is it written in?
Mainly Rust, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • shiyutim/tickets3,586 ★ · +236
Star history
Repository2026-09-242026-11-15
shiyutim/tickets3.4k3.6k

236 Stars in the last 52 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

The scope statement and the support caveats are the honest part

The README opens with a line that most projects of this kind leave out: it is for technical study and exchange, asks for reasonable use, and says not to use it for commercial ticket buying on other people's behalf. Then there is a support section that is more useful than the feature list. Support for the first platform follows the activities its mobile interface exposes, and the second prefers its newer product detail interface while staying compatible with the older one. Only ticket tiers that need no seat selection are handled, and payment is completed on the official page. When a login expires or a human verification is demanded, the instruction is to handle it on the official site first, then reload the event and start the task. And the closing line is the one to quote: building and simulated testing cannot replace verification with a real account at the real sale time, and it does not guarantee a successful purchase.

Two platforms, one workbench, and the backend owns the timing

The application is built on a desktop shell with a Rust backend and a Vue front end, and the layout is a left-hand platform switcher with separate forms per platform, plus pages for tasks and records and for settings. Running locally needs Node 18 or newer, a recent stable Rust, and the desktop shell's system dependencies:

bash
npm ci
npm run tauri dev

Everything time-sensitive lives in the backend: waiting for a scheduled start, retrying a purchase, and cancelling are all backend responsibilities, and switching pages does not disturb a task that has already started. That split is the reason the front end is as thin as it is. The limitation follows from the same design: a task keeps running only while the application is running, so the machine has to stay awake and the process has to stay open, and nothing is resumed automatically after a restart. The README says so in the timing section rather than leaving you to discover it.

Clock calibration samples two interfaces and then freezes the answer

The most interesting piece of engineering is time. The app asks a public time interface on one commercial site first and falls back to one on the other, and it takes several samples and keeps the one with the shortest round trip. The correction it stores is defined precisely as server time minus local time, so a positive value means the machine is slow. The wait a backend task uses is then the target time minus that correction minus the current local time, and the front-end countdown is explicitly for display only. Two consequences are documented and both matter. A task uses the correction value from the moment it was created, so calibrating again later does not move an already scheduled task. And the fallback interface has only second-level precision, so the interface shows a corresponding error rather than promising millisecond accuracy. A failed calibration keeps the previous value and can be retried or typed in by hand.

Retries are bounded, and one purchase task runs per platform

The scheduling controls are deliberately finite: you set an attempt count and a retry interval, and the backend honours both. On top of that there is a concurrency rule stated twice, because it is the one that will surprise you. At most one purchasing task per platform runs at a time, while stock monitors may run several concurrently. And the application as a whole runs at most a hundred tasks at once. Order handling is also more careful than the rest of the project: after credentials expire the order is prepared again, but verification failures, purchase limits and duplicate-order responses stop the task and ask a human to deal with it. When the order response cannot be confirmed either way, the app tells you to check the official order rather than retrying blindly, which is the correct instinct for a non-idempotent operation.

The stock monitor reads a detail interface and stops before ordering

The monitoring feature is the part with the clearest ethical line in the whole repository, and it is enforced in the code path rather than in a disclaimer. It only reads the product detail interface and explicitly does not probe availability by attempting a purchase. You can monitor a session and tier that are sold out or not yet on sale, which means it works before a sale as well as after one. The query interval defaults to sixty seconds and can be set between five and an hour, and the number of queries can be unlimited. When a tier is explicitly purchasable it alerts once, in the app, optionally with a sound, optionally through the chat bot with the event, session, tier and the official link, and then the task ends without ordering. Missing fields, unknown states and a tier the interface has not returned yet all mean keep waiting, and five consecutive failures stop the monitor.

Cookies in plaintext, a bot token in plaintext, and file permissions

Local data handling is documented in more detail than the rest of the documentation, and not all of it is reassuring. Accounts live in a local table with cookies stored in plaintext on the device; the purchasing and monitoring forms keep only an account id and never copy the credentials. Attendee details are used for the current run only and are not written to logs or task history, and the operation log uses bound parameters. General settings live in their own table and only take effect after a successful save, with the older settings table migrated field by field and then kept. Two separate database files are used for development and production. The chat bot is the sharpest edge: the login credentials and session token are stored unencrypted in a binding file inside the application data directory, readable and writable only by the backend, with a directory mode and a file mode set on Unix systems.

The proxy layer handles common formats and refuses to convert the rest

Subscription proxy management comes from a single pinned external crate, and the supported formats are listed exactly. Clash configuration in either syntax, plain lines of proxy addresses, and base64 of either. The node protocols that actually work are three families, two of them secure and one of them a SOCKS variant. Everything else, a named list of six protocols, is skipped, and the README says plainly that no protocol conversion program will be started for them. So a subscription full of nodes you cannot use will still load and still list, with those nodes silently ignored. Refreshing uses conditional requests keyed on two headers, a failed refresh keeps the nodes already loaded for the run, and a subscription link may contain an access token, which is one more reason the file holding it is worth protecting.

A previous major of the desktop shell, three lockfiles, and two unpinned dependencies

The manifest is where the project's vintage shows. The desktop shell's JavaScript packages are on the first major version and the documentation links to that version's prerequisites page, while the front-end tooling is on a fourth major version of the bundler and the front-end library is on an early point release of its third major. There are two lockfiles in the repository and the instructions say to use one package manager, while also mentioning a second. Tests run on the runtime's own built-in test runner with a shell glob, so there is no test framework in the dependency list at all. Two dependencies deserve a look: one is named for an installation helper, and one plugin is installed from a repository address rather than from a registry with no version pinned. And the download section warns that published installers may not contain the newest source changes.

Editorial conclusion

Read this as a well-documented automation project rather than as a reliable way to get a ticket, because the README itself says the interfaces may change, that simulated testing cannot substitute for verification with a real account at the real sale time, and that no purchase is guaranteed. Three things to weigh before running it on a machine that matters. Your platform cookies are stored in plaintext and the chat bot token is stored unencrypted, so the machine is the trust boundary. A scheduled task freezes its clock correction at creation, so a task created from a bad calibration will fire at the wrong moment and later calibration will not fix it. And both the proxy support and the stock monitor have hard limits the documentation states plainly: some node protocols are skipped rather than converted, and an alert does not place an order.

Frequently asked questions

What is the tickets application?

A multi-platform ticket-purchasing workbench built on a desktop shell with a Rust backend and a Vue front end, supporting two Chinese ticketing platforms, with scheduled purchasing, remaining-ticket monitoring and chat bot alerts. Payment is completed on the official platform's own page.

Does tickets place an order automatically when tickets appear?

No. The monitoring feature only reads the product detail interface, does not probe stock by attempting an order, notifies you when a tier is explicitly purchasable, and then ends the task without ordering.

How does tickets keep accurate time for a sale start?

It samples two public time interfaces in preference order and keeps the sample with the shortest round trip, defining the correction as server time minus local time. A scheduled task fixes that correction when it is created, so later calibration does not move its trigger time.

Where does tickets store account credentials?

In a local account table with cookies stored in plaintext on the device; the forms keep only an account id. The chat bot credentials and session token are stored unencrypted in a binding file in the application data directory, with restrictive file permissions on Unix systems.

When was the last tickets release?

Version 0.6.1 on 2026-09-13, preceded by 0.6.0 eight days earlier and 0.5.0 in May 2024. The README also warns that published installers may not include the newest source changes.

Official sources

  1. Issues
  2. License: MIT
  3. README
  4. Releases
  5. shiyutim/tickets on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/shiyutim-tickets.svg)](https://hysenlabs.com/projects/shiyutim-tickets)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/shiyutim-tickets.svg)](https://hysenlabs.com/projects/shiyutim-tickets)

Was this useful?