SignalWatch

Maker of SignalWatch. Happy to answer questions about change monitoring and webhooks.

页面变更监控 + webhook 告警的组合很实用，尤其对盯竞品定价页和 changelog 这种场景。现在不少页面是动态渲染的，想了解一下：监控是基于渲染后快照还是原始 HTML？另外内容差异的粒度如何（整页 diff 还是区块级 diff），能否忽略广告位之类的噪声区域？免费档 3 个监控位对小型团队挺友好了。

Webhook-on-change is the right primitive here — polling dashboards rot, but a POST you can route into Slack or a script fits how ops teams already work. Does the 15-minute interval also apply to downtime checks, or is there a faster probe tier for status pages?

Does the webhook payload include the changed text itself, or only the URL and a fingerprint? If it sends the actual diff, I could route the alerts straight into a Slack channel without opening every page.

The maker's reply already covers the parts I would have asked about: raw HTTP plus extracted text (no headless render), a CSS selector to scope the watch, ignore_regex and min_change_ratio for noise, and a unified diff inside an HMAC-signed webhook so it can go straight to Slack. That is a solid setup for a changelog or a server-rendered pricing table. The constraint I would plan around is the no-JS fetch, since a lot of competitor prices only exist after client render. On the free tier of 3 monitors at 15 minutes, how is a failed fetch treated — 403, bot wall, or an empty body? Is that downtime, content drift, or a separate error so it does not look like the price disappeared?

The webhook-first angle is the right call for competitor pricing pages; email alerts for that tend to get ignored. One thing that usually makes or breaks change monitors is noise: rotating testimonials, cookie banners, timestamps and A/B variants trigger false 'drift'. Do you let users scope a monitor to a CSS selector or ignore regions, and does the webhook payload include a text diff or just a changed/unchanged flag? A diff in the payload would make it easy to pipe straight into Slack.

Thanks for the sharp questions — answering them in one place: 1) Rendering: we fetch the raw HTTP response and extract visible text (BeautifulSoup). No headless browser / JS render. If the price only exists after client-side render, we will not see it (and the site says so). 2) Diff granularity: unified text diff of the extracted lines, not a screenshot. Scope with a CSS selector (e.g. the pricing table), strip noise with ignore_regex (timestamps, "3 minutes ago"), ignore_order for list reshuffles, and a min_change_ratio so tiny A/B flicker does not page you. 3) Interval: the same schedule covers content + downtime + recovery (15 / 5 / 2 min by plan). There is not a separate faster probe tier today. 4) Webhook payload: yes — content.changed includes the unified diff (and Slack/Discord get a formatted ```diff block). You can pipe straight into Slack/n8n without opening the page. Payload is HMAC-signed. Live sample with real diffs: https://signalwatch.litework.me/live.html — happy to dig into any edge case.

Using a CSS selector plus ignore_regex to cut noise is exactly what most page monitors are missing, and a signed webhook with a text diff makes it easy to plug into n8n or Make. Two questions: is there a Slack-ready format for the webhook, or is that left to the receiver? And does the dashboard keep a history of past diffs so you can see how a pricing page changed over a few months?

Raw HTML scoped by a CSS selector makes sense for pricing pages. If a selector stops matching after a site redesign, does that send its own alert, or does it look like nothing changed? That silent miss is what I'd worry about most when watching competitors.

For competitor pricing pages, a lot of noise comes from rotating banners, timestamps and cookie notices. Can I scope a monitor to a CSS selector or exclude regions, so the fingerprint only changes when the actual price block does? That would make the webhook far more useful for alerting.

Can I set the Accept-Language header or pick where the request comes from? A lot of pricing pages change by country, so it would be handy to watch the EU and the US version of the same page as two separate monitors.

Webhooks for competitor pricing pages are a good fit for the free tier's 3 monitors. If a page flaps, for example a CDN hiccup that returns an error once and then recovers, do you send an alert for each event or hold back until the change persists across checks? I'd want to avoid a noisy Slack channel on a flaky page.