A person with only a Windows PC used AI to build a flight-price-tracking tool: 52 days, 466 stars
What it is Flight-price tracking software you run on your own computer: it watches the same flights across Ctrip, Qunar, Fliggy and others, and when a price hits your threshold, it wakes you with DingTalk, pop-ups, or even a phone call.
Why it took off China had no cross-platform flight-tracking tool, a widely recognized gap. The author posted in the relevant LINUX DO community, then used AI for rapid iteration, shipping whatever users asked for the same day and bumping the thread with replies, turning a set of scripts into a complete product.
BHow you could follow it Don't do flights again. Use the same playbook on a different topic (multi-platform hotel price monitoring, which people are already asking for in the thread, plus award-ticket tracking and cashback price comparison), ship it as a product with a web-based config, no-install package and demo site, and launch it in LINUX DO's open-source promotion section, using the thread as a running changelog.
- Data (updated daily) · Repo created 2026-09-15 · 472 stars · 7 days +109
GitHub · stars Data 2026-10-15 → 2026-11-08 (25 days, 21 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
Anyone in China who buys flight tickets knows that for the same flight, Ctrip, Qunar, Fliggy and Tongcheng quote different prices, and the prices change daily. Yet no tool watched them together for you. An ordinary developer used AI to turn this into open-source software you can download and run, posted in one community, and then for three weeks granted nearly every request in the thread. Below: what it is, where the stars came from, why no competitor broke out, and whether you can build something along the same lines.
One post, and the first hour was all "real need"
On the morning of Oct 2026 13, LINUX DO user dengm posted in the "Open-source promotion" section with a plain title: "Personal flight monitoring open-source project." The body was just three links: a live demo, the GitHub repo, and a Windows no-install package.
7 minutes later came the first reply: "This really is a real need, bro." Over the next hour, some said "thanks, I'll try it first," some asked "can you add hotel monitoring?", and others asked whether it could search cheap flights by departure city and include the return leg. Three days later the author replied in a down-to-earth tone: "This project was written with AI, and I keep correcting the flight-scraping data." Asked whether he'd add cheap return flights, he said: "Sure, I still have tokens, I'll study that section with AI and integrate it."
By Nov 6 the thread had 83 replies, 2074 views and 130 likes. The repo's stars went from 10-15 to 10-20 in step with the thread, rising from 23 to 130, and the timing matches exactly. The author's GitHub account has only this one repo.
What it actually does
In one line: on your own computer, 7×24 it watches specified routes, puts the same flight from several platforms side by side, and wakes you when the price falls below your line.
- Collection: the browser visits Qunar, Fliggy, Ctrip, Tongcheng and Tuniu 5 sites (using Playwright to simulate mobile web pages; login required), with an optional 5 more App channels (Zhixing, Qunar App, Fliggy, Tongcheng, Ctrip App), read via an Android emulator.
- Judging: separate price thresholds for direct and connecting flights, with fine filters by airline, time window, connection transfer and baggage.
- Alerts: one DingTalk image-and-text push per cycle; when a target is hit, the title turns red, the group @s you, plus a Windows pop-up, ntfy, Bark, and a final layer of Alibaba Cloud phone calls and SMS; the README calls this "four-layer strong alerts."
- Interface: a "flight dispatcher style" web console with price trends, K-line charts, a price calendar, same-flight cross-channel comparison, and health status for each channel; all configuration is edited in the web UI, so you never touch a config file.
- Install: Windows / Linux / macOS packages that don't need Python installed, plus a Docker image, and a live demo site with fake data you can click through.
The README opens not with an introduction but with a sample push: "Shanghai→Urumqi 11/05, direct ¥2470, threshold ¥1900, ¥570 away (30%), China Eastern MU5678 12:30→17:05... 💡 Direct is 30% over the line, keep watching." A reader sees right away what will show up on their phone.
Why this one and not the earlier ones
Flight price tracking has been done before. Similar projects on GitHub (star counts are live GitHub values as of 11-06):
| Project | When it appeared | Stars | What it looks like |
|---|---|---|---|
| davidwushi1145/flightAlert | 2024-08 | 70 | Only uses Ctrip's API, WeChat push, a script |
| liuzhunai/flights_monitor | 2026-04 | 60 | Cheap-flight monitoring script |
| yangka1212/JiPiao | 2026-06 | 17 | Advertised "five platforms + WeChat push", no web config, no install package, stopped updating after June |
| zhanglong-ustc/flight-price-monitor | 2026-08 | 16 | Five platforms, runs free on GitHub Actions, no local console |
| dengmeiluan/ticket-monitoring | 2026-09 | 466 | Web console, no-install package, Docker, demo site, four-layer alerts |
JiPiao wrote nearly the same one-line description three months earlier and got only 17 stars. The difference isn't the idea, it's completeness: others shipped "a script that runs," while this one shipped "software you can download and use," and the README screenshots do the selling.
The big companies didn't hold this spot either. Ctrip, Fliggy and Travelsky all offer low-price alerts, but each only covers its own quotes and won't tell you a rival has the same flight cheaper; Google Flights and Tianxun don't cover domestic agent quotes, and in the thread someone literally called it the "Tianxun community edition." On V2EX someone posted a thread titled "No flight monitor on the market?", and replies mentioned an earlier app that did this, "Dragonfly," which is now gone. The spot was empty.
Where the stars came from
Lining up the daily star counts shows three phases:
| Timeline | Stars | What happened |
|---|---|---|
| 10-13 → 10-20 | 23 → 130 | The LINUX DO thread went up; 10-15 the author also published a first-person introduction on CSDN |
| 10-21 → 10-27 | 136 → 152 | Plateau, only a dozen or so stars per week |
| 10-28 | 152 → 273 | Single day +121 |
| 10-29 → 11-06 | 273 → 460 | +10 to +35 per day; +128 over the last 7 days, +186 over the first 7 days, slowing down |
(GitHub daily stars, 11-06)
The first phase is clear: that post. The second driver was the author's "grant nearly every request" approach in the thread: 10-16 added a Mac package and all cities, 10-19 added international routes and multi-return-day trips, responded to people asking for Bark and Docker, 10-27 replied "let AI try the MuMu emulator," and 11-02 announced "now supports the Zhixing App and Qunar App." Each time, the thread was bumped back to "latest," and these responses are also where each feature in the README came from: the thread served as its requirements pool and its changelog.
The source of the +121 on 10-28 couldn't be traced: it wasn't in that issue of HelloGitHub, not in Ruan Yifeng's weekly, and not in GitHub Trending's daily archive, and no one mentioned it on X. The timing follows the round of thread replies from 10-27, so the inference is that a WeChat official account or group reposted it, or it briefly appeared on the Python category of Trending; unverified.
Beyond that, the only visible traffic came from automatic listings on a few open-source aggregator sites; searching "flight price monitoring github" puts it first. No big-name accounts reposted it, and no media covered it.
Two numbers need a cool look: the only formal release, v1.0, has a combined download count of 156 across its three platform packages, so real installs are in the hundreds; there are 57 forks, a fork rate of about 12%, which is high. In the thread, some said they'd "git clone it locally and let cc tidy it up," so many people are using it as a base to modify themselves.
Its weak spots
- Scraping OTA with logged-in sessions. When Qunar shut down its PC version 10-10, the scraper switched to the mobile web. Any platform redesign or new anti-bot check can break it at any time, so you have to keep chasing the changes.
- A high bar to install. You need to log into 5 platform accounts and install a browser engine, and App channels also need an emulator, so this only suits people who enjoy tinkering.
- A visible ceiling. The topic is domestic flight tickets, the readers are domestic developers who self-host, and growth beyond four or five hundred stars is slowing; climbing to thousands isn't realistic.
If it were you
One-line verdict: the approach is clear and replicable, but the pool is small. It suits someone who wants to build a few hundred stars with a useful Chinese-language tool and practice community operations along the way; it doesn't suit someone aiming for big star counts or international reach.
How to do it: don't build another flight monitor, since that spot is already taken. Use the same playbook on a different topic: multi-platform hotel price-comparison monitoring, which people are already asking for in the thread, plus individual needs where big platforms each cover only their own inventory and nobody aggregates, such as award-ticket redemption, credit-card/cashback platform price comparison, and concert-ticket availability. Hold the deliverable to the same standard: web-based config, no-install packages for three platforms, Docker, a demo site with fake data, and a real sample push notification on the first screen of the README.
How to get users: launch in LINUX DO's "Open-source promotion" section (note its rules: AI-generated introductions must be posted as screenshots), and on the same day publish your own introduction on CSDN to catch search traffic. After that, treat the thread as a changelog: when a requested feature is done, reply and @ the requester, two or three bumps a week. V2EX's "why is there no flight monitor?" style question threads are a good place to find the next topic.
How early you need to be: there's no hard window, and topic choice doesn't depend on a one-off event. But the approach of "one person using AI to pile scripts into a product" is becoming more common, and whoever ships first on a given topic takes the top search spot, so the earlier the better; for the hotel-monitoring topic, it's worth starting within one to two months.
What would show you misjudged it: if the thread gets over a thousand views in a week but stars stay under 50, the topic isn't urgent enough or installation is too hard; or if the platforms you chose add anti-bot defenses and you can't collect data, forcing weekly crawler fixes, then the maintenance cost of this topic exceeds what you can earn from it.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one Ctrip / Qunar / Fliggy / Tongcheng built-in low-price alerts Big-company features Years Hundreds of millions of users It only compares prices on its own platform and its alerts are coarse-grained. This one compares across 5-10 channels for the same flight, with carrier/time-slot/connection/baggage filters and phone & SMS alerts. VariFlight / Flight Butler / Zhixing low-price and seat-availability alerts Big-company features Years tens of millions Same as above: single platform, rules can't be customized Google Flights price tracking / Skyscanner / Hopper Web Years Mainstream overseas Doesn't cover domestic OTA agent quotes; users on LINUX DO call it the "community edition of Skyscanner" Dragonfly (a domestic flight-monitoring app) Other Discontinued Not verified V2EX says it has disappeared, leaving the niche open davidwushi1145/flightAlert Open source 2024-08 70 stars Only uses Ctrip's API, WeChat push via ServerChan, script-only, no console liuzhunai/flights_monitor Open source 2026-04 60 stars A bargain-monitoring script with no multi-platform breakdown and no web config yangka1212/JiPiao Open source 2026-06 17 stars Also claims five platforms + WeChat push, but no web config or release builds, and stopped updating after June zhanglong-ustc/flight-price-monitor Open source 2026-08 16 stars Five platforms, runs on GitHub Actions, no local console or strong alerts punitarani/fli Open source 2025-01 3482 stars Google Flights library/MCP: overseas data sources, different audience AI-assistant flight-price-monitoring skills such as Fliggy FlyAI / QClaw and various flight MCP/Skill packages Big-company features 2026 Not verified Mostly query and Agent-call oriented, not local price watching with strong alerts like 7x24 - Where users come from
- Verified LINUX DO open-source promotion thread: https://linux.do/t/topic/3018961 posted 2026-10-13, 83 replies / 2074 views / 130 likes; 10-15→10-20 stars, 23→130
- Verified Thread kept alive by responding to every request and repeatedly bumping: the author 10-16, 10-19, 10-20, 10-27, 11-02 replied many times announcing features added on user request (Mac build, international routes, round trips, Bark, Docker, Zhixing/Qunar apps)
- Verified Self-written CSDN article: blog.csdn.net/2604_97231789/article/details/168030752, first-person introduction, 2026-10-15
- Verified Passive indexing by SEO/open-source aggregator sites: reporank.net, kaiyuanbang.cn, repo.tinix.ai, PanelAlpha and locdd.com mirrors; ranks #1 for "flight price monitoring github"
- Inferred Source of the single-day +121 stars on 10-28: not found in HelloGitHub issue 127, Ruan Yifeng's weekly, X, V2EX or GitHub Trending daily archives; timing follows a round of thread bumps by 10-27, possibly a WeChat account/group share or a brief Trending appearance
dengmeiluan/ticket-monitoring: self-hosted multi-platform flight price monitoring for China
Research date 2026-11-06 · channel github · data card: /home/ljr/fanout/why/cards/gh_dengmeiluan__ticket-monitoring.json
Bottom line
A regular domestic developer (self-described as having only a Windows PC and a "project written with AI") posted an open-source promo thread on LINUX DO that hit a real pain point: there's no reliable cross-platform price-tracking tool for domestic OTA sites. The thread reached floor 83, got 2074 views and 130 likes, and brought in the first wave of about 110 stars. After that, the author kept replying to every request in the thread (round trips, international routes, Bark, Docker, Mac builds, app stores), which kept bumping it back up. Add organic traffic from SEO aggregator sites and forks, and it went from 52 to 466 stars in days. What made it take off was a well-chosen topic + one community post + extremely fast iteration, not any big external account resharing it.
Key facts
- Repo created 2026-09-15; GitHub API 2026-11-06 shows 466 stars / 57 forks / 1 watchers / 1 open issues; Apache-2.0; Python + Playwright.
- The author account dengmeiluan has only this one repo (data card owner_repos=1); the LINUX DO username is dengm.
- Star curve (data card daily, from our github database): 10-15 23 stars → 10-20 130 stars (LINUX DO thread, posted 10-13) → plateau 10-21~10-27 136→152 → 10-28 single-day +121 (273) → afterwards +10~35 per day, 11-06 460. Last 7 days +128, first 7 days +186, so it's slowing.
- Only one formal release, v1.0 (2026-10-23), with 156 cumulative downloads across the three platform builds (GitHub releases API, 2026-11-06). Real installs are in the hundreds, not thousands. The repo also has many v1.5.x tag pages (Brave search shows v1.5.16 through v1.5.63), which indicates very fast iteration.
- Star trend: GitHub's stargazer timestamp endpoint is closed to our token (403), so the source of the 10-28 jump couldn't be pinned down precisely. See "Unconfirmed".
What it is
自托管的机票价格监控：浏览器 5 家（去哪儿 / 飞猪 / 携程 / 同程 / 途牛，用 Playwright 走 H5 或 PC 页面、需登录）+ App 5 家（智行 / 去哪儿 / 飞猪 / 同程 / 携程 App，经 Android 模拟器 + adb，默认关）采集航班明细；按直飞 / 中转两条「心理价」判达标；每轮聚合一条钉钉图文推送，达标再叠加钉钉 @、Windows 弹窗、ntfy、Bark、阿里云电话短信（"四层强提醒"）；带一个"飞行签派风格"的网页控制台（走势 / K 线 / 价格日历 / 同班跨渠道比价 / 渠道健康），配置全在网页、热加载；提供 Windows/Linux/macOS 免装 Python 的包和 Docker（ghcr）镜像；有在线演示站 dengmeiluan.github.io/ticket-monitoring/site/。（出处：仓库 README）
Competitors
| Product | Where | Timeline | Scale | How it differs |
|---|---|---|---|---|
| "Low-price / price-drop alerts" inside the Ctrip / Qunar / Fliggy / Tongcheng apps | Big-company features | Years | Hundreds of millions of users | Each runs only its own platform, with no cross-platform comparison. Alerts are coarse (mostly lowest-price calendars) and can't be filtered by carrier, time slot, connection, or baggage. ticket-monitoring is exactly the cross-channel comparison of the same flight across 5~10 channels, plus fine-grained filters. |
| VariFlight / Flight Butler / Zhixing seat-availability and low-price alerts | Big-company features | Years | tens of millions | Same as above; articles like the one from Sina (2026-10-24) list them as the "mainstream monitoring methods" |
| Google Flights price tracking / Skyscanner price alerts / Hopper | Web/app | Years | Mainstream overseas | Doesn't cover domestic OTA agent quotes; LINUX DO users call it the "community edition of Skyscanner" |
| Dragonfly (a domestic flight-monitoring app) | iOS/Android | Discontinued | — | V2EX thread "No flight monitor on the market?" (t/1200310) mentions "there used to be Dragonfly in this space, not sure if it was acquired or shut down". The slot is empty. |
| davidwushi1145/flightAlert (and the same-named omegatao fork) | Open source | 2024-08 | 70 stars | Only uses Ctrip's API, WeChat push via ServerChan, script-only |
| liuzhunai/flights_monitor | Open source | 2026-04 | 60 stars | Cheap-flight monitoring script |
| yangka1212/JiPiao | Open source | 2026-06 | 17 stars | Also claims five platforms + WeChat push, but no web config and no release builds, and stopped updating after June |
| zhanglong-ustc/flight-price-monitor | Open source | 2026-08 | 16 stars | Five platforms, runs free on GitHub Actions, no local console |
| tangtaizong666/ticket-price-sentinel, xanthichi/ctrip-flight-alter | Open source | 2026-03/05 | 2~4 stars | Ctrip only |
| punitarani/fli | Open source | 2025-01 | 3482 stars | Google Flights library/MCP: overseas data sources, not the same audience |
| Fliggy FlyAI open platform, "price monitoring" skills in AI assistants like QClaw, and various flight MCP/Skill packages | Big-tech features and others | 2026 | Not verified | Built on official APIs or Agents, leaning toward querying, not local price watching with strong alerts like 7×24 |
(Open-source repo star counts are live values from the GitHub API, 2026-11-06)
Conclusion: there are plenty of similar open-source projects, but they all stop at the "script that runs" level, with the top one at 70 stars. ticket-monitoring is the first to turn it into a product people can download and use.
Growth tactics (evidence)
- LINUX DO "open-source promotion" thread (confirmed, main channel): linux.do/t/topic/3018961, posted 2026-10-13 11:04 UTC, 83 replies, 2074 views, 130 likes (Discourse JSON API). Early replies in the first hour were things like "this is a real need, bro" and "I'll try it first". Stars went from 10-15 to 23, then 10-20 to 130, and the timing matches exactly. The thread was later reposted to mirror sites like locdd.com's "大佬说".
- Responding to every request in the thread and bumping it repeatedly (confirmed): 10-16 added a Mac build and all cities, 10-19 added international filters, multi-leg round trips and configurable dates, and responded to requests for Bark and Docker/NAS. 10-27 replied "let AI try the MuMu emulator", and 11-02 announced "Zhixing and Qunar app support is live." Each reply bumped the thread back to "latest", and each README feature traces back to one of these replies.
- Self-written CSDN article (confirmed): "Self-hosted flight price monitoring open-source project", posted 2026-10-15 (blog.csdn.net/2604_97231789/article/details/168030752), first-person by the author, covering search traffic.
- Passive indexing by SEO / aggregator sites (confirmed, small volume): reporank.net, kaiyuanbang.cn (open-source rankings), repo.tinix.ai, the PanelAlpha Engine one-click deploy page and RepoMasters all index it; it ranks #1 for searches like "flight price monitoring github / open source DingTalk".
- Live demo site + no-install package + Docker (confirmed): lowers the bar for "look once and star"; the release package has been downloaded 156 times.
- Single-day +121 stars on 10-28 (source not pinned down): checked HelloGitHub issue 127 (posted 10-28, not included), Ruan Yifeng's weekly, X (60 posts in the x-radar corpus over 0 days), V2EX, and the GitHub Trending daily archives (Rain1er/trending #419, raiscui/gh-trends 10-28 both absent). Nothing found. The timing follows a fresh round of replies and bumps in the thread by 10-27. It could also be a WeChat account/group share or a brief appearance on GitHub Trending (Python); not verified.
- Not seen: big X accounts, Xiaohongshu/Douyin influencers, media coverage, paid promotion.
Why it took off
- The topic is a widely acknowledged unmet need: domestic flight prices swing a lot and quotes differ across OTAs, while the cross-platform tracker (Dragonfly) is gone. A V2EX user posted specifically asking "No flight monitor on the market?", and a LINUX DO user said "Everything has a price history, why not flight tickets?"
- Posted in the most fitting community: LINUX DO users are domestic developers who have tokens, like to tinker and hunt deals, which is exactly the crowd that can install Playwright and is willing to log into 5 OTA accounts. The community's "open-source promotion" tag is itself a traffic source.
- Product polish far beyond similar scripts: web config, hot reloading, DingTalk card messages, four alert layers, trend/candlestick charts, same-flight cross-channel comparison, no-install packages for three platforms, Docker, and a live demo. The README screenshots alone sell it. Similar open-source projects are a few hundred lines of script, topping out at 70 stars.
- Very fast iteration powered by AI: the author admits it was "written with AI" and "I still have tokens", shipping a version every few days and launching user-requested features the same day, creating a loop of "request → ship → bump → newcomers star".
- Fork/star ≈ 12%: many people fork it to modify (one user: "git it locally and let cc tidy it up"). A "foundation for secondary development with AI" naturally has a high fork rate.
Differentiation
- Compared with the OTAs' built-in alerts: cross-platform, same-flight price comparison with fine-grained filters (carrier, time slot, connection, direct baggage) + strong alerts (phone and SMS), with data kept on your own machine.
- Compared with similar open-source scripts: it's a product, not a script. It has a web console, no install packages, Docker, a live demo site, and notification channels, and it keeps up with platform changes (after Qunar's PC version 2026-10-10 went offline, it switched to H5; the App channel runs on an emulator).
- This differentiation is why it took off. The same "five-platform monitoring" pitch was written by yangka1212/JiPiao back in June, and it only got 17 stars. The gap is in finish and in ongoing community work.
How visible it is
The buzz in the small circle: one LINUX DO post (2074 views) + the author's CSDN sponsored article + a few SEO aggregator sites. On X, 0 posts, no media coverage, not featured in HelloGitHub or Ruan Yifeng's weekly. Open source and free, no revenue involved.
Copycats
No standalone copycat project yet. There are 57 forks. JuWei231/ticket-monitoring changed the description and added English ("Self-hosted flight price monitoring with a dispatch-board console"), and LLL151/ticket renamed the project and keeps pushing updates (1 stars). Both are fork-and-tweak jobs, nothing that has gained traction.
Takeaways
- Pick a personal pain point that big platforms ignore and nobody aggregates (flight → hotel, mileage tickets, cashback; commenters on the post were already asking for it), and build it as a self-hosted tool.
- Post in the open-source promotion section of a relevant community like LINUX DO, then use the thread as a demand pool and changelog: every time you ship a feature a commenter asked for, reply and @ them, and keep bumping the thread.
- Use AI to turn "scripts" into a "product": web-based config, no install packages, Docker, a live demo site, and a README full of push-notification samples and plenty of screenshots.
- Risks to keep in view: scraping OTA sites with logged-in sessions means breakage whenever platforms redesign or tighten anti-bot measures. Install base is still only in the hundreds.
Unconfirmed
- The specific source of 10-28's +121 stars in a single day (the stargazer timestamp API was unavailable, and searches turned up no matching recommendation placement).
- Real usage: only the 156 downloads of the v1.0 release package are a reference point. The number of source and Docker users is unknown.
- How long the scraper can run stably under OTA anti-bot measures, and whether it will attract legal pressure from platforms, is unknown.