repository·master·Indexed 18 days ago
https://github.com/lumenresearch/uasurferA lightweight Go package for parsing and abstracting HTTP User-Agent strings. It identifies browser names, operating system versions, and device types (mobile, tablet, desktop, TV, console, wearable). The library includes specialized logic for bot and crawler detection, high-performance parsing for request loops via ParseUserAgent, and a CLI tool called uastats for generating User-Agent statistics.
The DeviceType field identifies the hardware category (e.g., DeviceComputer, DeviceTablet, DeviceTV).
Android Phone vs. Tablet Caveat:
It is not always possible to distinguish between an Android phone and a tablet because some vendors ship tablets that announce Mobile in their User-Agent. In these cases, an Android agent with no specific tablet indicator defaults to being classified as a phone.
DeviceTV. However, Fire tablets are categorized separately and will report DeviceTablet.Versions in uasurfer are represented by a Version struct with Major, Minor, and Patch fields.
To check if a browser meets a minimum requirement, compare the Major field directly or use the Less method for more complex comparisons.
Example: Checking Chrome version
if ua.Browser.Version.Major > 23 {
// Handle modern Chrome
}Example: Using Less()
if ver1.Less(ver2) {
// ver1 is older than ver2
}DeviceTV. Because many TV platforms are based on Linux but do not explicitly state a version, you should use DeviceTV to identify the device type and then rely on the reported browser and its version to determine rendering capabilities.Instead of relying solely on a static list of names, uasurfer uses convention-based detection to identify bots. It catches:
…bot, …spider, or …crawler.+http….curl, python-requests, Go-http-client, and Scrapy.High-volume crawlers are identified by specific constants. If a crawler is not one of these specific named constants, it is reported as BrowserBot.
Specific constants include:
BrowserGoogleBot (covers AdsBot, Mediapartners, GoogleOther, Google-InspectionTool, and Googlebot)BrowserBingBotBrowserOpenAIBotBrowserAnthropicBotBrowserPerplexityBotBrowserAmazonBotBrowserBytedanceBotBrowserAhrefsBotBrowserSemrushBotuasurfer does not attempt to maintain an exhaustive list of every crawler in existence. Consequently, it may not catch crawlers that:
Datanyze, Rigor, Scope3/2.0, binlar).Detection accuracy is approximately 83% based on measured crawler agents, with zero false positives reported against real device agents.
DeviceTV.To extend the library with support for new devices, systems, or browsers, follow these steps:
browser.go, device.go, or system.go (e.g., using strings.Contains(ua, "identifier")).testdata/ that currently fail without your changes.gofmt, go vet, golangci-lint, and go test. For parsing logic, run benchmarks using go test -run=XXX -bench=Parse -benchmem -count=6 to ensure no performance regressions.If you are parsing User-Agents within a high-frequency request loop, use ParseUserAgent to avoid repeated allocations. This function fills a UserAgent struct that you own.
Important: You must call .Reset() on your UserAgent instance before reusing it for the next string to ensure data from the previous parse does not persist.
// Assuming 'ua' is a pre-allocated *uasurfer.UserAgent
ua.Reset()
uasurfer.ParseUserAgent(rawUAString, ua)The library includes specialized logic to detect Amazon Fire tablets and phones. It looks for specific model tokens within the User Agent string, specifically:
kabcde).sd<4 digits>ur (e.g., sd1234ur).When detected, these are typically categorized under the Linux platform with the OSKindle name.
The uasurfer package classifies a User Agent string into a specific DeviceType based on OS platform, browser, and specific hardware markers. The classification logic follows a priority hierarchy to ensure accuracy (e.g., checking for TV markers before defaulting to mobile/phone).
Supported DeviceType categories include:
DeviceComputer: Windows, Mac, or Linux (if not otherwise classified).DeviceTV: Identified via specific markers like roku, chromecast, stb, or Amazon Fire TV (aft).DeviceTablet: Identified via iPad, kindle/, playbook, or specific Android tablet identifiers (e.g., nexus 7, sm-t).DevicePhone: Identified via iPhone, Blackberry, mobile, or mobi strings.DeviceConsole: Identified via Playstation, Xbox, or Nintendo platforms.DeviceWearable: Identified via glass, watch, or sm-v markers.DeviceUnknown: The fallback when no patterns match.Note: The classification logic is internal to the Parse process and updates the DeviceType field on the UserAgent struct.
The platformGroup function extracts the substring located between the first opening parenthesis ( and the first closing parenthesis ).
If the closing parenthesis is missing or appears before the opening one, the function returns everything from the opening parenthesis to the end of the User Agent string. This extracted group is often used as the primary source for identifying the OS and version before falling back to a full-string search.
aft prefix within the platform group of the User Agent. To avoid false positives (like the word "after"), the parser ensures aft is at a field edge (preceded by a space, ;, or ,) and is followed by alphanumeric characters representing the specific model (e.g., aftb, aftmm).