@apideck/agent-analytics 0.5.0 → 0.7.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (9) hide show
  1. package/dist/index.cjs +79 -2
  2. package/dist/index.cjs.map +1 -1
  3. package/dist/index.d.cts +42 -11
  4. package/dist/index.d.ts +42 -11
  5. package/dist/index.js +78 -3
  6. package/dist/index.js.map +1 -1
  7. package/dist/markdown.cjs.map +1 -1
  8. package/dist/markdown.js.map +1 -1
  9. package/package.json +1 -1
package/dist/index.cjs CHANGED
@@ -34,6 +34,40 @@ function parseBotName(userAgent) {
34
34
  if (s.includes("cursor")) return "Cursor";
35
35
  if (s.includes("windsurf")) return "Windsurf";
36
36
  if (s.includes("petalbot")) return "PetalBot";
37
+ if (s.includes("ahrefsbot")) return "Ahrefs";
38
+ if (s.includes("semrushbot")) return "Semrush";
39
+ if (s.includes("mj12bot")) return "Majestic";
40
+ if (s.includes("dotbot")) return "Moz";
41
+ if (s.includes("rogerbot")) return "Moz";
42
+ if (s.includes("screaming frog")) return "Screaming Frog";
43
+ if (s.includes("sitebulb")) return "Sitebulb";
44
+ if (s.includes("linkfluence")) return "Linkfluence";
45
+ if (s.includes("dataforseo")) return "DataForSEO";
46
+ if (s.includes("serpstatbot")) return "Serpstat";
47
+ if (s.includes("uptimerobot")) return "UptimeRobot";
48
+ if (s.includes("pingdom")) return "Pingdom";
49
+ if (s.includes("statuscake")) return "StatusCake";
50
+ if (s.includes("newrelicpinger")) return "New Relic";
51
+ if (s.includes("datadogagent") || s.includes("datadog")) return "Datadog";
52
+ if (s.includes("slackbot")) return "Slack";
53
+ if (s.includes("twitterbot")) return "Twitter";
54
+ if (s.includes("linkedinbot")) return "LinkedIn";
55
+ if (s.includes("discordbot")) return "Discord";
56
+ if (s.includes("telegrambot")) return "Telegram";
57
+ if (s.includes("whatsapp")) return "WhatsApp";
58
+ if (s.includes("linkupbot")) return "Linkup";
59
+ if (s.includes("sogou")) return "Sogou";
60
+ if (s.includes("yandexbot")) return "Yandex";
61
+ if (s.includes("baiduspider")) return "Baidu";
62
+ if (s.includes("facebookexternalhit")) return "Facebook";
63
+ if (s.includes("com.apple.webkit")) return "Apple URL Preview";
64
+ if (s.includes("ohdear")) return "Oh Dear";
65
+ if (s.includes("scrapy")) return "Scrapy";
66
+ if (s.includes("headlesschrome")) return "Headless Chrome";
67
+ if (s.includes("phantomjs")) return "PhantomJS";
68
+ if (s.includes("wget")) return "wget";
69
+ if (s.includes("httpie")) return "HTTPie";
70
+ if (s.includes("guzzlehttp")) return "Guzzle";
37
71
  if (s.includes("electron/")) return "Electron";
38
72
  if (/curl\//.test(s)) return "curl";
39
73
  if (/axios\//.test(s)) return "axios";
@@ -56,6 +90,33 @@ function firstUserAgentProduct(userAgent) {
56
90
  const first = userAgent.trim().split("/")[0]?.trim().split(/\s+/)[0]?.trim();
57
91
  return first || "Other";
58
92
  }
93
+ function detectHeadless(req) {
94
+ const signals = [];
95
+ const ua = (req.headers.get("user-agent") || "").toLowerCase();
96
+ const isBrowserUA = ua.includes("mozilla") || ua.includes("chrome") || ua.includes("safari") || ua.includes("firefox");
97
+ if (!isBrowserUA) return { score: 0, signals: [], likely: false };
98
+ if (!req.headers.get("accept-language")) {
99
+ signals.push("missing-accept-language");
100
+ }
101
+ if (!req.headers.get("sec-fetch-mode")) {
102
+ signals.push("missing-sec-fetch-mode");
103
+ }
104
+ const secChUa = req.headers.get("sec-ch-ua");
105
+ if (!secChUa) {
106
+ signals.push("missing-sec-ch-ua");
107
+ } else if (secChUa.toLowerCase().includes("headlesschrome")) {
108
+ signals.push("headless-chrome-hint");
109
+ }
110
+ const accept = req.headers.get("accept") || "";
111
+ if (!accept || accept === "*/*") {
112
+ signals.push("missing-or-bare-accept");
113
+ }
114
+ if ((req.headers.get("connection") || "").toLowerCase() === "close") {
115
+ signals.push("connection-close");
116
+ }
117
+ const score = signals.length;
118
+ return { score, signals, likely: score >= 2 };
119
+ }
59
120
  function classifyAgent(userAgent) {
60
121
  const label = parseBotName(userAgent);
61
122
  const aiBot = isAiBot(userAgent);
@@ -67,6 +128,16 @@ function classifyAgent(userAgent) {
67
128
  else kind = "other";
68
129
  return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient };
69
130
  }
131
+ function classifyRequest(req) {
132
+ const userAgent = req.headers.get("user-agent") || "";
133
+ const base = classifyAgent(userAgent);
134
+ const headless = detectHeadless(req);
135
+ let kind = base.kind;
136
+ if (kind === "browser" && headless.likely) {
137
+ kind = "headless-likely";
138
+ }
139
+ return { ...base, kind, headless };
140
+ }
70
141
 
71
142
  // src/hash.ts
72
143
  function hashId(input) {
@@ -83,7 +154,9 @@ async function trackVisit(req, opts) {
83
154
  const onlyBots = opts.onlyBots ?? false;
84
155
  const skipBrowsers = opts.skipBrowsers ?? false;
85
156
  if (onlyBots && !isAiBot(userAgent)) return;
86
- if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) return;
157
+ if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) {
158
+ if (!detectHeadless(req).likely) return;
159
+ }
87
160
  let pathname = "/";
88
161
  let originFromUrl = "";
89
162
  try {
@@ -97,7 +170,7 @@ async function trackVisit(req, opts) {
97
170
  const forwardedFor = req.headers.get("x-forwarded-for") || "";
98
171
  const ip = forwardedFor.split(",")[0]?.trim() ?? "";
99
172
  const referer = req.headers.get("referer");
100
- const classification = classifyAgent(userAgent);
173
+ const classification = classifyRequest(req);
101
174
  const event = {
102
175
  event: opts.eventName ?? "agent_visit",
103
176
  distinctId: hashId(`${ip}:${userAgent}`),
@@ -111,6 +184,8 @@ async function trackVisit(req, opts) {
111
184
  bot_name: classification.label,
112
185
  ua_category: classification.kind,
113
186
  coding_agent_hint: classification.codingAgentHint,
187
+ headless_score: classification.headless?.score ?? 0,
188
+ headless_likely: classification.headless?.likely ?? false,
114
189
  referer,
115
190
  source: opts.source ?? null,
116
191
  ...opts.properties
@@ -175,7 +250,9 @@ function customAnalytics(capture) {
175
250
  exports.AI_BOT_PATTERN = AI_BOT_PATTERN;
176
251
  exports.HTTP_CLIENT_PATTERN = HTTP_CLIENT_PATTERN;
177
252
  exports.classifyAgent = classifyAgent;
253
+ exports.classifyRequest = classifyRequest;
178
254
  exports.customAnalytics = customAnalytics;
255
+ exports.detectHeadless = detectHeadless;
179
256
  exports.firstUserAgentProduct = firstUserAgentProduct;
180
257
  exports.hashId = hashId;
181
258
  exports.isAiBot = isAiBot;
package/dist/index.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/bots.ts","../src/hash.ts","../src/track.ts","../src/adapters/posthog.ts","../src/adapters/webhook.ts","../src/adapters/custom.ts"],"names":[],"mappings":";;;AAgBO,IAAM,cAAA,GACX;AAmBK,IAAM,mBAAA,GACX;AAEK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;AAEO,SAAS,aAAa,SAAA,EAA+C;AAC1E,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,mBAAA,CAAoB,KAAK,SAAS,CAAA;AAC3C;AAeO,SAAS,aAAa,SAAA,EAA8C;AACzE,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,CAAA,GAAI,UAAU,WAAA,EAAY;AAGhC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,IAAK,CAAA,CAAE,SAAS,QAAQ,CAAA;AAC1G,IAAA,OAAO,SAAA;AACT,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AAC5F,EAAA,IAAI,CAAA,CAAE,SAAS,eAAe,CAAA,IAAK,EAAE,QAAA,CAAS,iBAAiB,GAAG,OAAO,YAAA;AACzE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,cAAA;AAChC,EAAA,IAAI,CAAA,CAAE,SAAS,iBAAiB,CAAA,IAAK,EAAE,QAAA,CAAS,WAAW,GAAG,OAAO,QAAA;AACrE,EAAA,IAAI,CAAA,CAAE,SAAS,mBAAmB,CAAA,IAAK,EAAE,QAAA,CAAS,UAAU,GAAG,OAAO,OAAA;AACtE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,MAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,SAAS,oBAAoB,CAAA,IAAK,EAAE,QAAA,CAAS,aAAa,GAAG,OAAO,MAAA;AAC1E,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,SAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,EAAG,OAAO,YAAA;AACxC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAInC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,UAAA;AACpC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAC7B,EAAA,IAAI,SAAA,CAAU,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAC9B,EAAA,IAAI,4BAAA,CAA6B,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,KAAA;AACjD,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,cAAA,CAAe,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,YAAA;AACnC,EAAA,IAAI,mBAAA,CAAoB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,iBAAA;AACxC,EAAA,IAAI,kBAAA,CAAmB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,gBAAA;AACvC,EAAA,IAAI,UAAA,CAAW,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,QAAA;AAC/B,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,SAAA;AAChC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAG7B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,SAAS,SAAS,CAAA;AAC/F,IAAA,OAAO,SAAA;AAET,EAAA,OAAO,OAAA;AACT;AAOO,SAAS,sBAAsB,SAAA,EAA8C;AAClF,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,eAAA,GAAkB,SAAA,CAAU,KAAA,CAAM,yCAAyC,CAAA;AACjF,EAAA,IAAI,eAAA,IAAmB,gBAAgB,CAAC,CAAA,SAAU,eAAA,CAAgB,CAAC,EAAE,IAAA,EAAK;AAC1E,EAAA,MAAM,QAAQ,SAAA,CAAU,IAAA,EAAK,CAAE,KAAA,CAAM,GAAG,CAAA,CAAE,CAAC,CAAA,EAAG,IAAA,GAAO,KAAA,CAAM,KAAK,CAAA,CAAE,CAAC,GAAG,IAAA,EAAK;AAC3E,EAAA,OAAO,KAAA,IAAS,OAAA;AAClB;AAmCO,SAAS,cAAc,SAAA,EAA2D;AACvF,EAAA,MAAM,KAAA,GAAQ,aAAa,SAAS,CAAA;AACpC,EAAA,MAAM,KAAA,GAAQ,QAAQ,SAAS,CAAA;AAC/B,EAAA,MAAM,UAAA,GAAa,aAAa,SAAS,CAAA;AAEzC,EAAA,IAAI,IAAA;AACJ,EAAA,IAAI,OAAO,IAAA,GAAO,kBAAA;AAAA,OAAA,IACT,YAAY,IAAA,GAAO,mBAAA;AAAA,OAAA,IACnB,KAAA,KAAU,WAAW,IAAA,GAAO,SAAA;AAAA,OAChC,IAAA,GAAO,OAAA;AAEZ,EAAA,OAAO,EAAE,IAAA,EAAM,KAAA,EAAO,OAAA,EAAS,KAAA,EAAO,iBAAiB,UAAA,EAAW;AACpE;;;ACjKO,SAAS,OAAO,KAAA,EAAuB;AAC5C,EAAA,IAAI,CAAA,GAAI,IAAA;AACR,EAAA,KAAA,IAAS,CAAA,GAAI,CAAA,EAAG,CAAA,GAAI,KAAA,CAAM,QAAQ,CAAA,EAAA,EAAK;AACrC,IAAA,CAAA,GAAA,CAAM,KAAK,CAAA,IAAK,CAAA,GAAI,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA,GAAK,UAAA;AAAA,EAC7C;AACA,EAAA,OAAO,OAAA,GAAA,CAAW,CAAA,KAAM,CAAA,EAAG,QAAA,CAAS,EAAE,CAAA;AACxC;;;ACCA,eAAsB,UAAA,CACpB,KACA,IAAA,EACe;AACf,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAEnD,EAAA,MAAM,QAAA,GAAW,KAAK,QAAA,IAAY,KAAA;AAClC,EAAA,MAAM,YAAA,GAAe,KAAK,YAAA,IAAgB,KAAA;AAC1C,EAAA,IAAI,QAAA,IAAY,CAAC,OAAA,CAAQ,SAAS,CAAA,EAAG;AACrC,EAAA,IAAI,YAAA,IAAgB,CAAC,OAAA,CAAQ,SAAS,KAAK,CAAC,YAAA,CAAa,SAAS,CAAA,EAAG;AAErE,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI,aAAA,GAAgB,EAAA;AACpB,EAAA,IAAI;AACF,IAAA,MAAM,GAAA,GAAM,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA;AAC3B,IAAA,QAAA,GAAW,GAAA,CAAI,QAAA;AACf,IAAA,aAAA,GAAgB,GAAA,CAAI,MAAA;AAAA,EACtB,CAAA,CAAA,MAAQ;AAEN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AACA,EAAA,MAAM,MAAA,GAAS,KAAK,MAAA,IAAU,aAAA;AAE9B,EAAA,MAAM,YAAA,GAAe,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,IAAK,EAAA;AAC3D,EAAA,MAAM,EAAA,GAAK,aAAa,KAAA,CAAM,GAAG,EAAE,CAAC,CAAA,EAAG,MAAK,IAAK,EAAA;AACjD,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,SAAS,CAAA;AACzC,EAAA,MAAM,cAAA,GAAiB,cAAc,SAAS,CAAA;AAE9C,EAAA,MAAM,KAAA,GAAQ;AAAA,IACZ,KAAA,EAAO,KAAK,SAAA,IAAa,aAAA;AAAA,IACzB,YAAY,MAAA,CAAO,CAAA,EAAG,EAAE,CAAA,CAAA,EAAI,SAAS,CAAA,CAAE,CAAA;AAAA,IACvC,SAAA,EAAA,iBAAW,IAAI,IAAA,EAAK,EAAE,WAAA,EAAY;AAAA,IAClC,UAAA,EAAY;AAAA,MACV,uBAAA,EAAyB,KAAA;AAAA,MACzB,cAAc,MAAA,GAAS,CAAA,EAAG,MAAM,CAAA,EAAG,QAAQ,CAAA,CAAA,GAAK,QAAA;AAAA,MAChD,IAAA,EAAM,QAAA;AAAA,MACN,UAAA,EAAY,SAAA;AAAA,MACZ,WAAW,cAAA,CAAe,OAAA;AAAA,MAC1B,UAAU,cAAA,CAAe,KAAA;AAAA,MACzB,aAAa,cAAA,CAAe,IAAA;AAAA,MAC5B,mBAAmB,cAAA,CAAe,eAAA;AAAA,MAClC,OAAA;AAAA,MACA,MAAA,EAAQ,KAAK,MAAA,IAAU,IAAA;AAAA,MACvB,GAAG,IAAA,CAAK;AAAA;AACV,GACF;AAEA,EAAA,IAAI;AACF,IAAA,MAAM,IAAA,CAAK,SAAA,CAAU,OAAA,CAAQ,KAAK,CAAA;AAAA,EACpC,CAAA,CAAA,MAAQ;AAAA,EAER;AACF;;;ACpCO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,OAAA,GAAU,OAAO,IAAA,IAAQ,0BAAA;AAC/B,EAAA,MAAM,IAAA,GAAA,CAAQ,cAAA,CAAe,IAAA,CAAK,OAAO,CAAA,GAAI,OAAA,GAAU,CAAA,QAAA,EAAW,OAAO,CAAA,CAAA,EAAI,OAAA,CAAQ,KAAA,EAAO,EAAE,CAAA;AAC9F,EAAA,MAAM,QAAQ,MAAA,CAAO,IAAA,IAAQ,UAAA,EAAY,OAAA,CAAQ,WAAW,GAAG,CAAA;AAC/D,EAAA,MAAM,QAAA,GAAW,CAAA,EAAG,IAAI,CAAA,EAAG,IAAI,CAAA,CAAA;AAC/B,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AAEtC,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,OAAA,GAAU;AAAA,QACd,SAAS,MAAA,CAAO,MAAA;AAAA,QAChB,OAAO,KAAA,CAAM,KAAA;AAAA,QACb,aAAa,KAAA,CAAM,UAAA;AAAA,QACnB,WAAW,KAAA,CAAM,SAAA;AAAA,QACjB,YAAY,KAAA,CAAM;AAAA,OACpB;AACA,MAAA,MAAM,UAAU,QAAA,EAAU;AAAA,QACxB,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS,EAAE,cAAA,EAAgB,kBAAA,EAAmB;AAAA,QAC9C,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,OAAO,CAAA;AAAA,QAC5B,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC/BO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AACtC,EAAA,MAAM,SAAA,GAAY,MAAA,CAAO,SAAA,KAAc,CAAC,CAAA,KAA6B,CAAA,CAAA;AAErE,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,SAAA,CAAU,OAAO,GAAA,EAAK;AAAA,QAC1B,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS;AAAA,UACP,cAAA,EAAgB,kBAAA;AAAA,UAChB,GAAI,MAAA,CAAO,OAAA,IAAW;AAAC,SACzB;AAAA,QACA,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,SAAA,CAAU,KAAK,CAAC,CAAA;AAAA,QACrC,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC3BO,SAAS,gBACd,OAAA,EACkB;AAClB,EAAA,OAAO,EAAE,OAAA,EAAQ;AACnB","file":"index.cjs","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the UA:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'browser'` — looks like a real browser. Could be a genuine user or\n * a Playwright-based agent (Aider, OpenCode) that can't be distinguished\n * at the UA layer.\n * - `'other'` — unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n}\n\n/**\n * One-stop classification of a user-agent. Combines {@link isAiBot},\n * {@link isHttpClient}, and {@link parseBotName} into a single structured\n * result. Used internally by `trackVisit` to populate event properties;\n * useful in consumer code when you need all signals at once.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n","/**\n * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to\n * build stable anonymous distinct-ids from `ip:ua:...` tuples without\n * collecting identifying data. Not cryptographic — collisions are fine for\n * analytics segmentation.\n */\nexport function hashId(input: string): string {\n let h = 5381\n for (let i = 0; i < input.length; i++) {\n h = ((h << 5) + h + input.charCodeAt(i)) & 0xffffffff\n }\n return 'anon_' + (h >>> 0).toString(16)\n}\n","import { classifyAgent, isAiBot, isHttpClient } from './bots.js'\nimport { hashId } from './hash.js'\nimport type { TrackVisitOptions } from './types.js'\n\n/**\n * Capture an event describing the incoming request. Fire-and-forget: awaits\n * the adapter but swallows errors so a downed analytics backend never breaks\n * the response path. Callers typically don't await the returned promise.\n *\n * By default, captures every request so coding-agent traffic (axios, curl,\n * Electron, …) shows up alongside branded crawlers. Set `onlyBots: true` to\n * restrict capture to UAs matching {@link AI_BOT_PATTERN}.\n */\nexport async function trackVisit(\n req: Request,\n opts: TrackVisitOptions\n): Promise<void> {\n const userAgent = req.headers.get('user-agent') || ''\n\n const onlyBots = opts.onlyBots ?? false\n const skipBrowsers = opts.skipBrowsers ?? false\n if (onlyBots && !isAiBot(userAgent)) return\n if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) return\n\n let pathname = '/'\n let originFromUrl = ''\n try {\n const url = new URL(req.url)\n pathname = url.pathname\n originFromUrl = url.origin\n } catch {\n // Some runtimes hand us a relative URL; fall back to the raw string.\n pathname = req.url || '/'\n }\n const origin = opts.origin ?? originFromUrl\n\n const forwardedFor = req.headers.get('x-forwarded-for') || ''\n const ip = forwardedFor.split(',')[0]?.trim() ?? ''\n const referer = req.headers.get('referer')\n const classification = classifyAgent(userAgent)\n\n const event = {\n event: opts.eventName ?? 'agent_visit',\n distinctId: hashId(`${ip}:${userAgent}`),\n timestamp: new Date().toISOString(),\n properties: {\n $process_person_profile: false,\n $current_url: origin ? `${origin}${pathname}` : pathname,\n path: pathname,\n user_agent: userAgent,\n is_ai_bot: classification.isAiBot,\n bot_name: classification.label,\n ua_category: classification.kind,\n coding_agent_hint: classification.codingAgentHint,\n referer,\n source: opts.source ?? null,\n ...opts.properties\n }\n }\n\n try {\n await opts.analytics.capture(event)\n } catch {\n // Intentional swallow — analytics failures must not affect the response.\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface PostHogAdapterConfig {\n /** PostHog project API key (the public one used by the JS SDK). */\n apiKey: string\n /**\n * PostHog host, with or without scheme. Defaults to `https://us.i.posthog.com`.\n * Use `https://eu.i.posthog.com` for EU cloud, or your own reverse-proxy\n * domain (e.g. `https://svc.example.com`).\n */\n host?: string\n /**\n * Path on the host that accepts single-event captures. Defaults to\n * `/i/v0/e/` which is PostHog's current endpoint for this.\n */\n path?: string\n /**\n * Override the `fetch` implementation (useful for tests or custom runtimes\n * that need a pinned fetch).\n */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that posts each event to the PostHog capture endpoint. Uses\n * `keepalive: true` so the request survives after a serverless response\n * returns — events aren't guaranteed (fire-and-forget), but that's the\n * trade we want to keep the hot path fast.\n */\nexport function posthogAnalytics(config: PostHogAdapterConfig): AnalyticsAdapter {\n const hostRaw = config.host ?? 'https://us.i.posthog.com'\n const base = (/^https?:\\/\\//.test(hostRaw) ? hostRaw : `https://${hostRaw}`).replace(/\\/$/, '')\n const path = (config.path ?? '/i/v0/e/').replace(/^(?!\\/)/, '/')\n const endpoint = `${base}${path}`\n const fetchImpl = config.fetchImpl ?? fetch\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n const payload = {\n api_key: config.apiKey,\n event: event.event,\n distinct_id: event.distinctId,\n timestamp: event.timestamp,\n properties: event.properties\n }\n await fetchImpl(endpoint, {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(payload),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface WebhookAdapterConfig {\n /** Destination URL that receives a POST for each event. */\n url: string\n /** Extra headers merged onto the POST (useful for shared-secret auth). */\n headers?: Record<string, string>\n /**\n * Transform the event into the exact JSON body the destination expects.\n * Defaults to sending the {@link CaptureEvent} as-is.\n */\n transform?: (event: CaptureEvent) => unknown\n /** Override the `fetch` implementation. */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that POSTs each event to an arbitrary webhook URL. Keeps the\n * library analytics-backend-agnostic — use this when PostHog isn't your\n * analytics of record, or when you want to multiplex events through your\n * own ingestion layer.\n */\nexport function webhookAnalytics(config: WebhookAdapterConfig): AnalyticsAdapter {\n const fetchImpl = config.fetchImpl ?? fetch\n const transform = config.transform ?? ((e: CaptureEvent): unknown => e)\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n await fetchImpl(config.url, {\n method: 'POST',\n headers: {\n 'Content-Type': 'application/json',\n ...(config.headers ?? {})\n },\n body: JSON.stringify(transform(event)),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\n/**\n * Escape hatch for wiring a callback directly as an analytics adapter.\n * Useful when you want to log events, pipe them through your own SDK, or\n * compose multiple adapters.\n *\n * @example\n * ```ts\n * const devAnalytics = customAnalytics((e) => console.log('[doc_view]', e))\n * ```\n */\nexport function customAnalytics(\n capture: (event: CaptureEvent) => Promise<void> | void\n): AnalyticsAdapter {\n return { capture }\n}\n"]}
1
+ {"version":3,"sources":["../src/bots.ts","../src/hash.ts","../src/track.ts","../src/adapters/posthog.ts","../src/adapters/webhook.ts","../src/adapters/custom.ts"],"names":[],"mappings":";;;AAgBO,IAAM,cAAA,GACX;AAmBK,IAAM,mBAAA,GACX;AAEK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;AAEO,SAAS,aAAa,SAAA,EAA+C;AAC1E,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,mBAAA,CAAoB,KAAK,SAAS,CAAA;AAC3C;AAeO,SAAS,aAAa,SAAA,EAA8C;AACzE,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,CAAA,GAAI,UAAU,WAAA,EAAY;AAGhC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,IAAK,CAAA,CAAE,SAAS,QAAQ,CAAA;AAC1G,IAAA,OAAO,SAAA;AACT,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AAC5F,EAAA,IAAI,CAAA,CAAE,SAAS,eAAe,CAAA,IAAK,EAAE,QAAA,CAAS,iBAAiB,GAAG,OAAO,YAAA;AACzE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,cAAA;AAChC,EAAA,IAAI,CAAA,CAAE,SAAS,iBAAiB,CAAA,IAAK,EAAE,QAAA,CAAS,WAAW,GAAG,OAAO,QAAA;AACrE,EAAA,IAAI,CAAA,CAAE,SAAS,mBAAmB,CAAA,IAAK,EAAE,QAAA,CAAS,UAAU,GAAG,OAAO,OAAA;AACtE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,MAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,SAAS,oBAAoB,CAAA,IAAK,EAAE,QAAA,CAAS,aAAa,GAAG,OAAO,MAAA;AAC1E,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,SAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,EAAG,OAAO,YAAA;AACxC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAGnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,UAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,KAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,gBAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,aAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AAGtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,aAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,WAAA;AACzC,EAAA,IAAI,CAAA,CAAE,SAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,SAAS,GAAG,OAAO,SAAA;AAChE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,OAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAGnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,OAAA;AAGtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,qBAAqB,CAAA,EAAG,OAAO,UAAA;AAC9C,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,kBAAkB,CAAA,EAAG,OAAO,mBAAA;AAG3C,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AAGjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,iBAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,WAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,MAAM,CAAA,EAAG,OAAO,MAAA;AAC/B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,QAAA;AAIrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,UAAA;AACpC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAC7B,EAAA,IAAI,SAAA,CAAU,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAC9B,EAAA,IAAI,4BAAA,CAA6B,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,KAAA;AACjD,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,cAAA,CAAe,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,YAAA;AACnC,EAAA,IAAI,mBAAA,CAAoB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,iBAAA;AACxC,EAAA,IAAI,kBAAA,CAAmB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,gBAAA;AACvC,EAAA,IAAI,UAAA,CAAW,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,QAAA;AAC/B,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,SAAA;AAChC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAG7B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,SAAS,SAAS,CAAA;AAC/F,IAAA,OAAO,SAAA;AAET,EAAA,OAAO,OAAA;AACT;AAOO,SAAS,sBAAsB,SAAA,EAA8C;AAClF,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,eAAA,GAAkB,SAAA,CAAU,KAAA,CAAM,yCAAyC,CAAA;AACjF,EAAA,IAAI,eAAA,IAAmB,gBAAgB,CAAC,CAAA,SAAU,eAAA,CAAgB,CAAC,EAAE,IAAA,EAAK;AAC1E,EAAA,MAAM,QAAQ,SAAA,CAAU,IAAA,EAAK,CAAE,KAAA,CAAM,GAAG,CAAA,CAAE,CAAC,CAAA,EAAG,IAAA,GAAO,KAAA,CAAM,KAAK,CAAA,CAAE,CAAC,GAAG,IAAA,EAAK;AAC3E,EAAA,OAAO,KAAA,IAAS,OAAA;AAClB;AAkBO,SAAS,eAAe,GAAA,EAAiC;AAC9D,EAAA,MAAM,UAAoB,EAAC;AAC3B,EAAA,MAAM,MAAM,GAAA,CAAI,OAAA,CAAQ,IAAI,YAAY,CAAA,IAAK,IAAI,WAAA,EAAY;AAC7D,EAAA,MAAM,WAAA,GACJ,EAAA,CAAG,QAAA,CAAS,SAAS,KAAK,EAAA,CAAG,QAAA,CAAS,QAAQ,CAAA,IAAK,GAAG,QAAA,CAAS,QAAQ,CAAA,IAAK,EAAA,CAAG,SAAS,SAAS,CAAA;AAEnG,EAAA,IAAI,CAAC,WAAA,EAAa,OAAO,EAAE,KAAA,EAAO,GAAG,OAAA,EAAS,EAAC,EAAG,MAAA,EAAQ,KAAA,EAAM;AAEhE,EAAA,IAAI,CAAC,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,EAAG;AACvC,IAAA,OAAA,CAAQ,KAAK,yBAAyB,CAAA;AAAA,EACxC;AACA,EAAA,IAAI,CAAC,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,gBAAgB,CAAA,EAAG;AACtC,IAAA,OAAA,CAAQ,KAAK,wBAAwB,CAAA;AAAA,EACvC;AACA,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,WAAW,CAAA;AAC3C,EAAA,IAAI,CAAC,OAAA,EAAS;AACZ,IAAA,OAAA,CAAQ,KAAK,mBAAmB,CAAA;AAAA,EAClC,WAAW,OAAA,CAAQ,WAAA,EAAY,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG;AAC3D,IAAA,OAAA,CAAQ,KAAK,sBAAsB,CAAA;AAAA,EACrC;AACA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,CAAC,MAAA,IAAU,MAAA,KAAW,KAAA,EAAO;AAC/B,IAAA,OAAA,CAAQ,KAAK,wBAAwB,CAAA;AAAA,EACvC;AACA,EAAA,IAAA,CAAK,GAAA,CAAI,QAAQ,GAAA,CAAI,YAAY,KAAK,EAAA,EAAI,WAAA,OAAkB,OAAA,EAAS;AACnE,IAAA,OAAA,CAAQ,KAAK,kBAAkB,CAAA;AAAA,EACjC;AAEA,EAAA,MAAM,QAAQ,OAAA,CAAQ,MAAA;AACtB,EAAA,OAAO,EAAE,KAAA,EAAO,OAAA,EAAS,MAAA,EAAQ,SAAS,CAAA,EAAE;AAC9C;AA6CO,SAAS,cAAc,SAAA,EAA2D;AACvF,EAAA,MAAM,KAAA,GAAQ,aAAa,SAAS,CAAA;AACpC,EAAA,MAAM,KAAA,GAAQ,QAAQ,SAAS,CAAA;AAC/B,EAAA,MAAM,UAAA,GAAa,aAAa,SAAS,CAAA;AAEzC,EAAA,IAAI,IAAA;AACJ,EAAA,IAAI,OAAO,IAAA,GAAO,kBAAA;AAAA,OAAA,IACT,YAAY,IAAA,GAAO,mBAAA;AAAA,OAAA,IACnB,KAAA,KAAU,WAAW,IAAA,GAAO,SAAA;AAAA,OAChC,IAAA,GAAO,OAAA;AAEZ,EAAA,OAAO,EAAE,IAAA,EAAM,KAAA,EAAO,OAAA,EAAS,KAAA,EAAO,iBAAiB,UAAA,EAAW;AACpE;AAOO,SAAS,gBAAgB,GAAA,EAAmC;AACjE,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AACnD,EAAA,MAAM,IAAA,GAAO,cAAc,SAAS,CAAA;AACpC,EAAA,MAAM,QAAA,GAAW,eAAe,GAAG,CAAA;AAEnC,EAAA,IAAI,OAAO,IAAA,CAAK,IAAA;AAChB,EAAA,IAAI,IAAA,KAAS,SAAA,IAAa,QAAA,CAAS,MAAA,EAAQ;AACzC,IAAA,IAAA,GAAO,iBAAA;AAAA,EACT;AAEA,EAAA,OAAO,EAAE,GAAG,IAAA,EAAM,IAAA,EAAM,QAAA,EAAS;AACnC;;;AC3RO,SAAS,OAAO,KAAA,EAAuB;AAC5C,EAAA,IAAI,CAAA,GAAI,IAAA;AACR,EAAA,KAAA,IAAS,CAAA,GAAI,CAAA,EAAG,CAAA,GAAI,KAAA,CAAM,QAAQ,CAAA,EAAA,EAAK;AACrC,IAAA,CAAA,GAAA,CAAM,KAAK,CAAA,IAAK,CAAA,GAAI,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA,GAAK,UAAA;AAAA,EAC7C;AACA,EAAA,OAAO,OAAA,GAAA,CAAW,CAAA,KAAM,CAAA,EAAG,QAAA,CAAS,EAAE,CAAA;AACxC;;;ACCA,eAAsB,UAAA,CACpB,KACA,IAAA,EACe;AACf,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAEnD,EAAA,MAAM,QAAA,GAAW,KAAK,QAAA,IAAY,KAAA;AAClC,EAAA,MAAM,YAAA,GAAe,KAAK,YAAA,IAAgB,KAAA;AAC1C,EAAA,IAAI,QAAA,IAAY,CAAC,OAAA,CAAQ,SAAS,CAAA,EAAG;AACrC,EAAA,IAAI,YAAA,IAAgB,CAAC,OAAA,CAAQ,SAAS,KAAK,CAAC,YAAA,CAAa,SAAS,CAAA,EAAG;AAInE,IAAA,IAAI,CAAC,cAAA,CAAe,GAAG,CAAA,CAAE,MAAA,EAAQ;AAAA,EACnC;AAEA,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI,aAAA,GAAgB,EAAA;AACpB,EAAA,IAAI;AACF,IAAA,MAAM,GAAA,GAAM,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA;AAC3B,IAAA,QAAA,GAAW,GAAA,CAAI,QAAA;AACf,IAAA,aAAA,GAAgB,GAAA,CAAI,MAAA;AAAA,EACtB,CAAA,CAAA,MAAQ;AAEN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AACA,EAAA,MAAM,MAAA,GAAS,KAAK,MAAA,IAAU,aAAA;AAE9B,EAAA,MAAM,YAAA,GAAe,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,IAAK,EAAA;AAC3D,EAAA,MAAM,EAAA,GAAK,aAAa,KAAA,CAAM,GAAG,EAAE,CAAC,CAAA,EAAG,MAAK,IAAK,EAAA;AACjD,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,SAAS,CAAA;AACzC,EAAA,MAAM,cAAA,GAAiB,gBAAgB,GAAG,CAAA;AAE1C,EAAA,MAAM,KAAA,GAAQ;AAAA,IACZ,KAAA,EAAO,KAAK,SAAA,IAAa,aAAA;AAAA,IACzB,YAAY,MAAA,CAAO,CAAA,EAAG,EAAE,CAAA,CAAA,EAAI,SAAS,CAAA,CAAE,CAAA;AAAA,IACvC,SAAA,EAAA,iBAAW,IAAI,IAAA,EAAK,EAAE,WAAA,EAAY;AAAA,IAClC,UAAA,EAAY;AAAA,MACV,uBAAA,EAAyB,KAAA;AAAA,MACzB,cAAc,MAAA,GAAS,CAAA,EAAG,MAAM,CAAA,EAAG,QAAQ,CAAA,CAAA,GAAK,QAAA;AAAA,MAChD,IAAA,EAAM,QAAA;AAAA,MACN,UAAA,EAAY,SAAA;AAAA,MACZ,WAAW,cAAA,CAAe,OAAA;AAAA,MAC1B,UAAU,cAAA,CAAe,KAAA;AAAA,MACzB,aAAa,cAAA,CAAe,IAAA;AAAA,MAC5B,mBAAmB,cAAA,CAAe,eAAA;AAAA,MAClC,cAAA,EAAgB,cAAA,CAAe,QAAA,EAAU,KAAA,IAAS,CAAA;AAAA,MAClD,eAAA,EAAiB,cAAA,CAAe,QAAA,EAAU,MAAA,IAAU,KAAA;AAAA,MACpD,OAAA;AAAA,MACA,MAAA,EAAQ,KAAK,MAAA,IAAU,IAAA;AAAA,MACvB,GAAG,IAAA,CAAK;AAAA;AACV,GACF;AAEA,EAAA,IAAI;AACF,IAAA,MAAM,IAAA,CAAK,SAAA,CAAU,OAAA,CAAQ,KAAK,CAAA;AAAA,EACpC,CAAA,CAAA,MAAQ;AAAA,EAER;AACF;;;AC3CO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,OAAA,GAAU,OAAO,IAAA,IAAQ,0BAAA;AAC/B,EAAA,MAAM,IAAA,GAAA,CAAQ,cAAA,CAAe,IAAA,CAAK,OAAO,CAAA,GAAI,OAAA,GAAU,CAAA,QAAA,EAAW,OAAO,CAAA,CAAA,EAAI,OAAA,CAAQ,KAAA,EAAO,EAAE,CAAA;AAC9F,EAAA,MAAM,QAAQ,MAAA,CAAO,IAAA,IAAQ,UAAA,EAAY,OAAA,CAAQ,WAAW,GAAG,CAAA;AAC/D,EAAA,MAAM,QAAA,GAAW,CAAA,EAAG,IAAI,CAAA,EAAG,IAAI,CAAA,CAAA;AAC/B,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AAEtC,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,OAAA,GAAU;AAAA,QACd,SAAS,MAAA,CAAO,MAAA;AAAA,QAChB,OAAO,KAAA,CAAM,KAAA;AAAA,QACb,aAAa,KAAA,CAAM,UAAA;AAAA,QACnB,WAAW,KAAA,CAAM,SAAA;AAAA,QACjB,YAAY,KAAA,CAAM;AAAA,OACpB;AACA,MAAA,MAAM,UAAU,QAAA,EAAU;AAAA,QACxB,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS,EAAE,cAAA,EAAgB,kBAAA,EAAmB;AAAA,QAC9C,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,OAAO,CAAA;AAAA,QAC5B,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC/BO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AACtC,EAAA,MAAM,SAAA,GAAY,MAAA,CAAO,SAAA,KAAc,CAAC,CAAA,KAA6B,CAAA,CAAA;AAErE,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,SAAA,CAAU,OAAO,GAAA,EAAK;AAAA,QAC1B,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS;AAAA,UACP,cAAA,EAAgB,kBAAA;AAAA,UAChB,GAAI,MAAA,CAAO,OAAA,IAAW;AAAC,SACzB;AAAA,QACA,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,SAAA,CAAU,KAAK,CAAC,CAAA;AAAA,QACrC,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC3BO,SAAS,gBACd,OAAA,EACkB;AAClB,EAAA,OAAO,EAAE,OAAA,EAAQ;AACnB","file":"index.cjs","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // SEO crawlers and monitoring bots.\n if (s.includes('ahrefsbot')) return 'Ahrefs'\n if (s.includes('semrushbot')) return 'Semrush'\n if (s.includes('mj12bot')) return 'Majestic'\n if (s.includes('dotbot')) return 'Moz'\n if (s.includes('rogerbot')) return 'Moz'\n if (s.includes('screaming frog')) return 'Screaming Frog'\n if (s.includes('sitebulb')) return 'Sitebulb'\n if (s.includes('linkfluence')) return 'Linkfluence'\n if (s.includes('dataforseo')) return 'DataForSEO'\n if (s.includes('serpstatbot')) return 'Serpstat'\n\n // Monitoring and feed bots.\n if (s.includes('uptimerobot')) return 'UptimeRobot'\n if (s.includes('pingdom')) return 'Pingdom'\n if (s.includes('statuscake')) return 'StatusCake'\n if (s.includes('newrelicpinger')) return 'New Relic'\n if (s.includes('datadogagent') || s.includes('datadog')) return 'Datadog'\n if (s.includes('slackbot')) return 'Slack'\n if (s.includes('twitterbot')) return 'Twitter'\n if (s.includes('linkedinbot')) return 'LinkedIn'\n if (s.includes('discordbot')) return 'Discord'\n if (s.includes('telegrambot')) return 'Telegram'\n if (s.includes('whatsapp')) return 'WhatsApp'\n\n // AI search and indexing bots.\n if (s.includes('linkupbot')) return 'Linkup'\n if (s.includes('sogou')) return 'Sogou'\n if (s.includes('yandexbot')) return 'Yandex'\n if (s.includes('baiduspider')) return 'Baidu'\n\n // Link preview fetchers.\n if (s.includes('facebookexternalhit')) return 'Facebook'\n if (s.includes('com.apple.webkit')) return 'Apple URL Preview'\n\n // Uptime and monitoring.\n if (s.includes('ohdear')) return 'Oh Dear'\n\n // Generic scrapers.\n if (s.includes('scrapy')) return 'Scrapy'\n if (s.includes('headlesschrome')) return 'Headless Chrome'\n if (s.includes('phantomjs')) return 'PhantomJS'\n if (s.includes('wget')) return 'wget'\n if (s.includes('httpie')) return 'HTTPie'\n if (s.includes('guzzlehttp')) return 'Guzzle'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\n/**\n * Detect likely headless/automated browsers by checking for missing headers\n * that real browsers always send. Playwright, Puppeteer, and similar tools\n * spoof the UA but often omit standard browser headers.\n *\n * Signals checked (each scores 1 point):\n * - Missing `Accept-Language` — every real browser sends this\n * - Missing `Sec-Fetch-Mode` — sent by all modern browsers\n * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+\n * - `Sec-CH-UA` contains \"HeadlessChrome\"\n * - Missing or bare Accept header — browsers send detailed accept lists\n * - `Connection: close` with browser UA — browsers use keep-alive\n *\n * Returns a score (0-6), the signals that fired, and a boolean `likely`\n * flag (score >= 2 with a browser-like UA).\n */\nexport function detectHeadless(req: Request): HeadlessDetection {\n const signals: string[] = []\n const ua = (req.headers.get('user-agent') || '').toLowerCase()\n const isBrowserUA =\n ua.includes('mozilla') || ua.includes('chrome') || ua.includes('safari') || ua.includes('firefox')\n\n if (!isBrowserUA) return { score: 0, signals: [], likely: false }\n\n if (!req.headers.get('accept-language')) {\n signals.push('missing-accept-language')\n }\n if (!req.headers.get('sec-fetch-mode')) {\n signals.push('missing-sec-fetch-mode')\n }\n const secChUa = req.headers.get('sec-ch-ua')\n if (!secChUa) {\n signals.push('missing-sec-ch-ua')\n } else if (secChUa.toLowerCase().includes('headlesschrome')) {\n signals.push('headless-chrome-hint')\n }\n const accept = req.headers.get('accept') || ''\n if (!accept || accept === '*/*') {\n signals.push('missing-or-bare-accept')\n }\n if ((req.headers.get('connection') || '').toLowerCase() === 'close') {\n signals.push('connection-close')\n }\n\n const score = signals.length\n return { score, signals, likely: score >= 2 }\n}\n\nexport interface HeadlessDetection {\n /** Number of suspicious signals found (0-6). */\n score: number\n /** Names of the specific signals that fired. */\n signals: string[]\n /** True when score >= 2 — strong headless indication. */\n likely: boolean\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'headless-likely'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the request:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'headless-likely'` — Browser-like UA but missing standard headers.\n * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).\n * - `'browser'` — Looks like a real browser with expected headers present.\n * - `'other'` — Unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n /** Headless browser detection result. Only populated when `req` is passed. */\n headless?: HeadlessDetection\n}\n\n/**\n * UA-only classification. Use {@link classifyRequest} for full detection\n * including headless browser heuristics.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n\n/**\n * Full request classification — combines UA parsing with header-based\n * headless detection. When a browser-like UA is missing standard headers,\n * the kind is promoted from `'browser'` to `'headless-likely'`.\n */\nexport function classifyRequest(req: Request): AgentClassification {\n const userAgent = req.headers.get('user-agent') || ''\n const base = classifyAgent(userAgent)\n const headless = detectHeadless(req)\n\n let kind = base.kind\n if (kind === 'browser' && headless.likely) {\n kind = 'headless-likely'\n }\n\n return { ...base, kind, headless }\n}\n","/**\n * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to\n * build stable anonymous distinct-ids from `ip:ua:...` tuples without\n * collecting identifying data. Not cryptographic — collisions are fine for\n * analytics segmentation.\n */\nexport function hashId(input: string): string {\n let h = 5381\n for (let i = 0; i < input.length; i++) {\n h = ((h << 5) + h + input.charCodeAt(i)) & 0xffffffff\n }\n return 'anon_' + (h >>> 0).toString(16)\n}\n","import { classifyRequest, detectHeadless, isAiBot, isHttpClient } from './bots.js'\nimport { hashId } from './hash.js'\nimport type { TrackVisitOptions } from './types.js'\n\n/**\n * Capture an event describing the incoming request. Fire-and-forget: awaits\n * the adapter but swallows errors so a downed analytics backend never breaks\n * the response path. Callers typically don't await the returned promise.\n *\n * By default, captures every request so coding-agent traffic (axios, curl,\n * Electron, …) shows up alongside branded crawlers. Set `onlyBots: true` to\n * restrict capture to UAs matching {@link AI_BOT_PATTERN}.\n */\nexport async function trackVisit(\n req: Request,\n opts: TrackVisitOptions\n): Promise<void> {\n const userAgent = req.headers.get('user-agent') || ''\n\n const onlyBots = opts.onlyBots ?? false\n const skipBrowsers = opts.skipBrowsers ?? false\n if (onlyBots && !isAiBot(userAgent)) return\n if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) {\n // Not a declared bot or HTTP client — check headless heuristics.\n // Playwright-based agents (Aider, OpenCode) will pass if they're missing\n // standard browser headers. Real browsers get skipped.\n if (!detectHeadless(req).likely) return\n }\n\n let pathname = '/'\n let originFromUrl = ''\n try {\n const url = new URL(req.url)\n pathname = url.pathname\n originFromUrl = url.origin\n } catch {\n // Some runtimes hand us a relative URL; fall back to the raw string.\n pathname = req.url || '/'\n }\n const origin = opts.origin ?? originFromUrl\n\n const forwardedFor = req.headers.get('x-forwarded-for') || ''\n const ip = forwardedFor.split(',')[0]?.trim() ?? ''\n const referer = req.headers.get('referer')\n const classification = classifyRequest(req)\n\n const event = {\n event: opts.eventName ?? 'agent_visit',\n distinctId: hashId(`${ip}:${userAgent}`),\n timestamp: new Date().toISOString(),\n properties: {\n $process_person_profile: false,\n $current_url: origin ? `${origin}${pathname}` : pathname,\n path: pathname,\n user_agent: userAgent,\n is_ai_bot: classification.isAiBot,\n bot_name: classification.label,\n ua_category: classification.kind,\n coding_agent_hint: classification.codingAgentHint,\n headless_score: classification.headless?.score ?? 0,\n headless_likely: classification.headless?.likely ?? false,\n referer,\n source: opts.source ?? null,\n ...opts.properties\n }\n }\n\n try {\n await opts.analytics.capture(event)\n } catch {\n // Intentional swallow — analytics failures must not affect the response.\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface PostHogAdapterConfig {\n /** PostHog project API key (the public one used by the JS SDK). */\n apiKey: string\n /**\n * PostHog host, with or without scheme. Defaults to `https://us.i.posthog.com`.\n * Use `https://eu.i.posthog.com` for EU cloud, or your own reverse-proxy\n * domain (e.g. `https://svc.example.com`).\n */\n host?: string\n /**\n * Path on the host that accepts single-event captures. Defaults to\n * `/i/v0/e/` which is PostHog's current endpoint for this.\n */\n path?: string\n /**\n * Override the `fetch` implementation (useful for tests or custom runtimes\n * that need a pinned fetch).\n */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that posts each event to the PostHog capture endpoint. Uses\n * `keepalive: true` so the request survives after a serverless response\n * returns — events aren't guaranteed (fire-and-forget), but that's the\n * trade we want to keep the hot path fast.\n */\nexport function posthogAnalytics(config: PostHogAdapterConfig): AnalyticsAdapter {\n const hostRaw = config.host ?? 'https://us.i.posthog.com'\n const base = (/^https?:\\/\\//.test(hostRaw) ? hostRaw : `https://${hostRaw}`).replace(/\\/$/, '')\n const path = (config.path ?? '/i/v0/e/').replace(/^(?!\\/)/, '/')\n const endpoint = `${base}${path}`\n const fetchImpl = config.fetchImpl ?? fetch\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n const payload = {\n api_key: config.apiKey,\n event: event.event,\n distinct_id: event.distinctId,\n timestamp: event.timestamp,\n properties: event.properties\n }\n await fetchImpl(endpoint, {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(payload),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface WebhookAdapterConfig {\n /** Destination URL that receives a POST for each event. */\n url: string\n /** Extra headers merged onto the POST (useful for shared-secret auth). */\n headers?: Record<string, string>\n /**\n * Transform the event into the exact JSON body the destination expects.\n * Defaults to sending the {@link CaptureEvent} as-is.\n */\n transform?: (event: CaptureEvent) => unknown\n /** Override the `fetch` implementation. */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that POSTs each event to an arbitrary webhook URL. Keeps the\n * library analytics-backend-agnostic — use this when PostHog isn't your\n * analytics of record, or when you want to multiplex events through your\n * own ingestion layer.\n */\nexport function webhookAnalytics(config: WebhookAdapterConfig): AnalyticsAdapter {\n const fetchImpl = config.fetchImpl ?? fetch\n const transform = config.transform ?? ((e: CaptureEvent): unknown => e)\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n await fetchImpl(config.url, {\n method: 'POST',\n headers: {\n 'Content-Type': 'application/json',\n ...(config.headers ?? {})\n },\n body: JSON.stringify(transform(event)),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\n/**\n * Escape hatch for wiring a callback directly as an analytics adapter.\n * Useful when you want to log events, pipe them through your own SDK, or\n * compose multiple adapters.\n *\n * @example\n * ```ts\n * const devAnalytics = customAnalytics((e) => console.log('[doc_view]', e))\n * ```\n */\nexport function customAnalytics(\n capture: (event: CaptureEvent) => Promise<void> | void\n): AnalyticsAdapter {\n return { capture }\n}\n"]}
package/dist/index.d.cts CHANGED
@@ -70,18 +70,43 @@ declare function parseBotName(userAgent: string | null | undefined): string;
70
70
  * input.
71
71
  */
72
72
  declare function firstUserAgentProduct(userAgent: string | null | undefined): string;
73
- type AgentKind = 'declared-crawler' | 'coding-agent-hint' | 'browser' | 'other';
73
+ /**
74
+ * Detect likely headless/automated browsers by checking for missing headers
75
+ * that real browsers always send. Playwright, Puppeteer, and similar tools
76
+ * spoof the UA but often omit standard browser headers.
77
+ *
78
+ * Signals checked (each scores 1 point):
79
+ * - Missing `Accept-Language` — every real browser sends this
80
+ * - Missing `Sec-Fetch-Mode` — sent by all modern browsers
81
+ * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+
82
+ * - `Sec-CH-UA` contains "HeadlessChrome"
83
+ * - Missing or bare Accept header — browsers send detailed accept lists
84
+ * - `Connection: close` with browser UA — browsers use keep-alive
85
+ *
86
+ * Returns a score (0-6), the signals that fired, and a boolean `likely`
87
+ * flag (score >= 2 with a browser-like UA).
88
+ */
89
+ declare function detectHeadless(req: Request): HeadlessDetection;
90
+ interface HeadlessDetection {
91
+ /** Number of suspicious signals found (0-6). */
92
+ score: number;
93
+ /** Names of the specific signals that fired. */
94
+ signals: string[];
95
+ /** True when score >= 2 — strong headless indication. */
96
+ likely: boolean;
97
+ }
98
+ type AgentKind = 'declared-crawler' | 'coding-agent-hint' | 'headless-likely' | 'browser' | 'other';
74
99
  interface AgentClassification {
75
100
  /**
76
- * Categorical tag for the UA:
101
+ * Categorical tag for the request:
77
102
  *
78
103
  * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.
79
104
  * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose
80
105
  * signal; could be a coding agent, a curl script, or any automation.
81
- * - `'browser'` — looks like a real browser. Could be a genuine user or
82
- * a Playwright-based agent (Aider, OpenCode) that can't be distinguished
83
- * at the UA layer.
84
- * - `'other'` — unrecognised or empty.
106
+ * - `'headless-likely'` — Browser-like UA but missing standard headers.
107
+ * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).
108
+ * - `'browser'` — Looks like a real browser with expected headers present.
109
+ * - `'other'` — Unrecognised or empty.
85
110
  */
86
111
  kind: AgentKind;
87
112
  /** Human-readable label, same string {@link parseBotName} returns. */
@@ -90,14 +115,20 @@ interface AgentClassification {
90
115
  isAiBot: boolean;
91
116
  /** Loose: `true` for known HTTP-library / automation UAs. */
92
117
  codingAgentHint: boolean;
118
+ /** Headless browser detection result. Only populated when `req` is passed. */
119
+ headless?: HeadlessDetection;
93
120
  }
94
121
  /**
95
- * One-stop classification of a user-agent. Combines {@link isAiBot},
96
- * {@link isHttpClient}, and {@link parseBotName} into a single structured
97
- * result. Used internally by `trackVisit` to populate event properties;
98
- * useful in consumer code when you need all signals at once.
122
+ * UA-only classification. Use {@link classifyRequest} for full detection
123
+ * including headless browser heuristics.
99
124
  */
100
125
  declare function classifyAgent(userAgent: string | null | undefined): AgentClassification;
126
+ /**
127
+ * Full request classification — combines UA parsing with header-based
128
+ * headless detection. When a browser-like UA is missing standard headers,
129
+ * the kind is promoted from `'browser'` to `'headless-likely'`.
130
+ */
131
+ declare function classifyRequest(req: Request): AgentClassification;
101
132
 
102
133
  /**
103
134
  * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to
@@ -119,4 +150,4 @@ declare function hashId(input: string): string;
119
150
  */
120
151
  declare function customAnalytics(capture: (event: CaptureEvent) => Promise<void> | void): AnalyticsAdapter;
121
152
 
122
- export { AI_BOT_PATTERN, type AgentClassification, type AgentKind, AnalyticsAdapter, CaptureEvent, HTTP_CLIENT_PATTERN, TrackVisitOptions, classifyAgent, customAnalytics, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, trackVisit };
153
+ export { AI_BOT_PATTERN, type AgentClassification, type AgentKind, AnalyticsAdapter, CaptureEvent, HTTP_CLIENT_PATTERN, type HeadlessDetection, TrackVisitOptions, classifyAgent, classifyRequest, customAnalytics, detectHeadless, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, trackVisit };
package/dist/index.d.ts CHANGED
@@ -70,18 +70,43 @@ declare function parseBotName(userAgent: string | null | undefined): string;
70
70
  * input.
71
71
  */
72
72
  declare function firstUserAgentProduct(userAgent: string | null | undefined): string;
73
- type AgentKind = 'declared-crawler' | 'coding-agent-hint' | 'browser' | 'other';
73
+ /**
74
+ * Detect likely headless/automated browsers by checking for missing headers
75
+ * that real browsers always send. Playwright, Puppeteer, and similar tools
76
+ * spoof the UA but often omit standard browser headers.
77
+ *
78
+ * Signals checked (each scores 1 point):
79
+ * - Missing `Accept-Language` — every real browser sends this
80
+ * - Missing `Sec-Fetch-Mode` — sent by all modern browsers
81
+ * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+
82
+ * - `Sec-CH-UA` contains "HeadlessChrome"
83
+ * - Missing or bare Accept header — browsers send detailed accept lists
84
+ * - `Connection: close` with browser UA — browsers use keep-alive
85
+ *
86
+ * Returns a score (0-6), the signals that fired, and a boolean `likely`
87
+ * flag (score >= 2 with a browser-like UA).
88
+ */
89
+ declare function detectHeadless(req: Request): HeadlessDetection;
90
+ interface HeadlessDetection {
91
+ /** Number of suspicious signals found (0-6). */
92
+ score: number;
93
+ /** Names of the specific signals that fired. */
94
+ signals: string[];
95
+ /** True when score >= 2 — strong headless indication. */
96
+ likely: boolean;
97
+ }
98
+ type AgentKind = 'declared-crawler' | 'coding-agent-hint' | 'headless-likely' | 'browser' | 'other';
74
99
  interface AgentClassification {
75
100
  /**
76
- * Categorical tag for the UA:
101
+ * Categorical tag for the request:
77
102
  *
78
103
  * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.
79
104
  * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose
80
105
  * signal; could be a coding agent, a curl script, or any automation.
81
- * - `'browser'` — looks like a real browser. Could be a genuine user or
82
- * a Playwright-based agent (Aider, OpenCode) that can't be distinguished
83
- * at the UA layer.
84
- * - `'other'` — unrecognised or empty.
106
+ * - `'headless-likely'` — Browser-like UA but missing standard headers.
107
+ * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).
108
+ * - `'browser'` — Looks like a real browser with expected headers present.
109
+ * - `'other'` — Unrecognised or empty.
85
110
  */
86
111
  kind: AgentKind;
87
112
  /** Human-readable label, same string {@link parseBotName} returns. */
@@ -90,14 +115,20 @@ interface AgentClassification {
90
115
  isAiBot: boolean;
91
116
  /** Loose: `true` for known HTTP-library / automation UAs. */
92
117
  codingAgentHint: boolean;
118
+ /** Headless browser detection result. Only populated when `req` is passed. */
119
+ headless?: HeadlessDetection;
93
120
  }
94
121
  /**
95
- * One-stop classification of a user-agent. Combines {@link isAiBot},
96
- * {@link isHttpClient}, and {@link parseBotName} into a single structured
97
- * result. Used internally by `trackVisit` to populate event properties;
98
- * useful in consumer code when you need all signals at once.
122
+ * UA-only classification. Use {@link classifyRequest} for full detection
123
+ * including headless browser heuristics.
99
124
  */
100
125
  declare function classifyAgent(userAgent: string | null | undefined): AgentClassification;
126
+ /**
127
+ * Full request classification — combines UA parsing with header-based
128
+ * headless detection. When a browser-like UA is missing standard headers,
129
+ * the kind is promoted from `'browser'` to `'headless-likely'`.
130
+ */
131
+ declare function classifyRequest(req: Request): AgentClassification;
101
132
 
102
133
  /**
103
134
  * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to
@@ -119,4 +150,4 @@ declare function hashId(input: string): string;
119
150
  */
120
151
  declare function customAnalytics(capture: (event: CaptureEvent) => Promise<void> | void): AnalyticsAdapter;
121
152
 
122
- export { AI_BOT_PATTERN, type AgentClassification, type AgentKind, AnalyticsAdapter, CaptureEvent, HTTP_CLIENT_PATTERN, TrackVisitOptions, classifyAgent, customAnalytics, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, trackVisit };
153
+ export { AI_BOT_PATTERN, type AgentClassification, type AgentKind, AnalyticsAdapter, CaptureEvent, HTTP_CLIENT_PATTERN, type HeadlessDetection, TrackVisitOptions, classifyAgent, classifyRequest, customAnalytics, detectHeadless, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, trackVisit };
package/dist/index.js CHANGED
@@ -32,6 +32,40 @@ function parseBotName(userAgent) {
32
32
  if (s.includes("cursor")) return "Cursor";
33
33
  if (s.includes("windsurf")) return "Windsurf";
34
34
  if (s.includes("petalbot")) return "PetalBot";
35
+ if (s.includes("ahrefsbot")) return "Ahrefs";
36
+ if (s.includes("semrushbot")) return "Semrush";
37
+ if (s.includes("mj12bot")) return "Majestic";
38
+ if (s.includes("dotbot")) return "Moz";
39
+ if (s.includes("rogerbot")) return "Moz";
40
+ if (s.includes("screaming frog")) return "Screaming Frog";
41
+ if (s.includes("sitebulb")) return "Sitebulb";
42
+ if (s.includes("linkfluence")) return "Linkfluence";
43
+ if (s.includes("dataforseo")) return "DataForSEO";
44
+ if (s.includes("serpstatbot")) return "Serpstat";
45
+ if (s.includes("uptimerobot")) return "UptimeRobot";
46
+ if (s.includes("pingdom")) return "Pingdom";
47
+ if (s.includes("statuscake")) return "StatusCake";
48
+ if (s.includes("newrelicpinger")) return "New Relic";
49
+ if (s.includes("datadogagent") || s.includes("datadog")) return "Datadog";
50
+ if (s.includes("slackbot")) return "Slack";
51
+ if (s.includes("twitterbot")) return "Twitter";
52
+ if (s.includes("linkedinbot")) return "LinkedIn";
53
+ if (s.includes("discordbot")) return "Discord";
54
+ if (s.includes("telegrambot")) return "Telegram";
55
+ if (s.includes("whatsapp")) return "WhatsApp";
56
+ if (s.includes("linkupbot")) return "Linkup";
57
+ if (s.includes("sogou")) return "Sogou";
58
+ if (s.includes("yandexbot")) return "Yandex";
59
+ if (s.includes("baiduspider")) return "Baidu";
60
+ if (s.includes("facebookexternalhit")) return "Facebook";
61
+ if (s.includes("com.apple.webkit")) return "Apple URL Preview";
62
+ if (s.includes("ohdear")) return "Oh Dear";
63
+ if (s.includes("scrapy")) return "Scrapy";
64
+ if (s.includes("headlesschrome")) return "Headless Chrome";
65
+ if (s.includes("phantomjs")) return "PhantomJS";
66
+ if (s.includes("wget")) return "wget";
67
+ if (s.includes("httpie")) return "HTTPie";
68
+ if (s.includes("guzzlehttp")) return "Guzzle";
35
69
  if (s.includes("electron/")) return "Electron";
36
70
  if (/curl\//.test(s)) return "curl";
37
71
  if (/axios\//.test(s)) return "axios";
@@ -54,6 +88,33 @@ function firstUserAgentProduct(userAgent) {
54
88
  const first = userAgent.trim().split("/")[0]?.trim().split(/\s+/)[0]?.trim();
55
89
  return first || "Other";
56
90
  }
91
+ function detectHeadless(req) {
92
+ const signals = [];
93
+ const ua = (req.headers.get("user-agent") || "").toLowerCase();
94
+ const isBrowserUA = ua.includes("mozilla") || ua.includes("chrome") || ua.includes("safari") || ua.includes("firefox");
95
+ if (!isBrowserUA) return { score: 0, signals: [], likely: false };
96
+ if (!req.headers.get("accept-language")) {
97
+ signals.push("missing-accept-language");
98
+ }
99
+ if (!req.headers.get("sec-fetch-mode")) {
100
+ signals.push("missing-sec-fetch-mode");
101
+ }
102
+ const secChUa = req.headers.get("sec-ch-ua");
103
+ if (!secChUa) {
104
+ signals.push("missing-sec-ch-ua");
105
+ } else if (secChUa.toLowerCase().includes("headlesschrome")) {
106
+ signals.push("headless-chrome-hint");
107
+ }
108
+ const accept = req.headers.get("accept") || "";
109
+ if (!accept || accept === "*/*") {
110
+ signals.push("missing-or-bare-accept");
111
+ }
112
+ if ((req.headers.get("connection") || "").toLowerCase() === "close") {
113
+ signals.push("connection-close");
114
+ }
115
+ const score = signals.length;
116
+ return { score, signals, likely: score >= 2 };
117
+ }
57
118
  function classifyAgent(userAgent) {
58
119
  const label = parseBotName(userAgent);
59
120
  const aiBot = isAiBot(userAgent);
@@ -65,6 +126,16 @@ function classifyAgent(userAgent) {
65
126
  else kind = "other";
66
127
  return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient };
67
128
  }
129
+ function classifyRequest(req) {
130
+ const userAgent = req.headers.get("user-agent") || "";
131
+ const base = classifyAgent(userAgent);
132
+ const headless = detectHeadless(req);
133
+ let kind = base.kind;
134
+ if (kind === "browser" && headless.likely) {
135
+ kind = "headless-likely";
136
+ }
137
+ return { ...base, kind, headless };
138
+ }
68
139
 
69
140
  // src/hash.ts
70
141
  function hashId(input) {
@@ -81,7 +152,9 @@ async function trackVisit(req, opts) {
81
152
  const onlyBots = opts.onlyBots ?? false;
82
153
  const skipBrowsers = opts.skipBrowsers ?? false;
83
154
  if (onlyBots && !isAiBot(userAgent)) return;
84
- if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) return;
155
+ if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) {
156
+ if (!detectHeadless(req).likely) return;
157
+ }
85
158
  let pathname = "/";
86
159
  let originFromUrl = "";
87
160
  try {
@@ -95,7 +168,7 @@ async function trackVisit(req, opts) {
95
168
  const forwardedFor = req.headers.get("x-forwarded-for") || "";
96
169
  const ip = forwardedFor.split(",")[0]?.trim() ?? "";
97
170
  const referer = req.headers.get("referer");
98
- const classification = classifyAgent(userAgent);
171
+ const classification = classifyRequest(req);
99
172
  const event = {
100
173
  event: opts.eventName ?? "agent_visit",
101
174
  distinctId: hashId(`${ip}:${userAgent}`),
@@ -109,6 +182,8 @@ async function trackVisit(req, opts) {
109
182
  bot_name: classification.label,
110
183
  ua_category: classification.kind,
111
184
  coding_agent_hint: classification.codingAgentHint,
185
+ headless_score: classification.headless?.score ?? 0,
186
+ headless_likely: classification.headless?.likely ?? false,
112
187
  referer,
113
188
  source: opts.source ?? null,
114
189
  ...opts.properties
@@ -170,6 +245,6 @@ function customAnalytics(capture) {
170
245
  return { capture };
171
246
  }
172
247
 
173
- export { AI_BOT_PATTERN, HTTP_CLIENT_PATTERN, classifyAgent, customAnalytics, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, posthogAnalytics, trackVisit, webhookAnalytics };
248
+ export { AI_BOT_PATTERN, HTTP_CLIENT_PATTERN, classifyAgent, classifyRequest, customAnalytics, detectHeadless, firstUserAgentProduct, hashId, isAiBot, isHttpClient, parseBotName, posthogAnalytics, trackVisit, webhookAnalytics };
174
249
  //# sourceMappingURL=index.js.map
175
250
  //# sourceMappingURL=index.js.map
package/dist/index.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/bots.ts","../src/hash.ts","../src/track.ts","../src/adapters/posthog.ts","../src/adapters/webhook.ts","../src/adapters/custom.ts"],"names":[],"mappings":";AAgBO,IAAM,cAAA,GACX;AAmBK,IAAM,mBAAA,GACX;AAEK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;AAEO,SAAS,aAAa,SAAA,EAA+C;AAC1E,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,mBAAA,CAAoB,KAAK,SAAS,CAAA;AAC3C;AAeO,SAAS,aAAa,SAAA,EAA8C;AACzE,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,CAAA,GAAI,UAAU,WAAA,EAAY;AAGhC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,IAAK,CAAA,CAAE,SAAS,QAAQ,CAAA;AAC1G,IAAA,OAAO,SAAA;AACT,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AAC5F,EAAA,IAAI,CAAA,CAAE,SAAS,eAAe,CAAA,IAAK,EAAE,QAAA,CAAS,iBAAiB,GAAG,OAAO,YAAA;AACzE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,cAAA;AAChC,EAAA,IAAI,CAAA,CAAE,SAAS,iBAAiB,CAAA,IAAK,EAAE,QAAA,CAAS,WAAW,GAAG,OAAO,QAAA;AACrE,EAAA,IAAI,CAAA,CAAE,SAAS,mBAAmB,CAAA,IAAK,EAAE,QAAA,CAAS,UAAU,GAAG,OAAO,OAAA;AACtE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,MAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,SAAS,oBAAoB,CAAA,IAAK,EAAE,QAAA,CAAS,aAAa,GAAG,OAAO,MAAA;AAC1E,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,SAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,EAAG,OAAO,YAAA;AACxC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAInC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,UAAA;AACpC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAC7B,EAAA,IAAI,SAAA,CAAU,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAC9B,EAAA,IAAI,4BAAA,CAA6B,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,KAAA;AACjD,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,cAAA,CAAe,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,YAAA;AACnC,EAAA,IAAI,mBAAA,CAAoB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,iBAAA;AACxC,EAAA,IAAI,kBAAA,CAAmB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,gBAAA;AACvC,EAAA,IAAI,UAAA,CAAW,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,QAAA;AAC/B,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,SAAA;AAChC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAG7B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,SAAS,SAAS,CAAA;AAC/F,IAAA,OAAO,SAAA;AAET,EAAA,OAAO,OAAA;AACT;AAOO,SAAS,sBAAsB,SAAA,EAA8C;AAClF,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,eAAA,GAAkB,SAAA,CAAU,KAAA,CAAM,yCAAyC,CAAA;AACjF,EAAA,IAAI,eAAA,IAAmB,gBAAgB,CAAC,CAAA,SAAU,eAAA,CAAgB,CAAC,EAAE,IAAA,EAAK;AAC1E,EAAA,MAAM,QAAQ,SAAA,CAAU,IAAA,EAAK,CAAE,KAAA,CAAM,GAAG,CAAA,CAAE,CAAC,CAAA,EAAG,IAAA,GAAO,KAAA,CAAM,KAAK,CAAA,CAAE,CAAC,GAAG,IAAA,EAAK;AAC3E,EAAA,OAAO,KAAA,IAAS,OAAA;AAClB;AAmCO,SAAS,cAAc,SAAA,EAA2D;AACvF,EAAA,MAAM,KAAA,GAAQ,aAAa,SAAS,CAAA;AACpC,EAAA,MAAM,KAAA,GAAQ,QAAQ,SAAS,CAAA;AAC/B,EAAA,MAAM,UAAA,GAAa,aAAa,SAAS,CAAA;AAEzC,EAAA,IAAI,IAAA;AACJ,EAAA,IAAI,OAAO,IAAA,GAAO,kBAAA;AAAA,OAAA,IACT,YAAY,IAAA,GAAO,mBAAA;AAAA,OAAA,IACnB,KAAA,KAAU,WAAW,IAAA,GAAO,SAAA;AAAA,OAChC,IAAA,GAAO,OAAA;AAEZ,EAAA,OAAO,EAAE,IAAA,EAAM,KAAA,EAAO,OAAA,EAAS,KAAA,EAAO,iBAAiB,UAAA,EAAW;AACpE;;;ACjKO,SAAS,OAAO,KAAA,EAAuB;AAC5C,EAAA,IAAI,CAAA,GAAI,IAAA;AACR,EAAA,KAAA,IAAS,CAAA,GAAI,CAAA,EAAG,CAAA,GAAI,KAAA,CAAM,QAAQ,CAAA,EAAA,EAAK;AACrC,IAAA,CAAA,GAAA,CAAM,KAAK,CAAA,IAAK,CAAA,GAAI,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA,GAAK,UAAA;AAAA,EAC7C;AACA,EAAA,OAAO,OAAA,GAAA,CAAW,CAAA,KAAM,CAAA,EAAG,QAAA,CAAS,EAAE,CAAA;AACxC;;;ACCA,eAAsB,UAAA,CACpB,KACA,IAAA,EACe;AACf,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAEnD,EAAA,MAAM,QAAA,GAAW,KAAK,QAAA,IAAY,KAAA;AAClC,EAAA,MAAM,YAAA,GAAe,KAAK,YAAA,IAAgB,KAAA;AAC1C,EAAA,IAAI,QAAA,IAAY,CAAC,OAAA,CAAQ,SAAS,CAAA,EAAG;AACrC,EAAA,IAAI,YAAA,IAAgB,CAAC,OAAA,CAAQ,SAAS,KAAK,CAAC,YAAA,CAAa,SAAS,CAAA,EAAG;AAErE,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI,aAAA,GAAgB,EAAA;AACpB,EAAA,IAAI;AACF,IAAA,MAAM,GAAA,GAAM,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA;AAC3B,IAAA,QAAA,GAAW,GAAA,CAAI,QAAA;AACf,IAAA,aAAA,GAAgB,GAAA,CAAI,MAAA;AAAA,EACtB,CAAA,CAAA,MAAQ;AAEN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AACA,EAAA,MAAM,MAAA,GAAS,KAAK,MAAA,IAAU,aAAA;AAE9B,EAAA,MAAM,YAAA,GAAe,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,IAAK,EAAA;AAC3D,EAAA,MAAM,EAAA,GAAK,aAAa,KAAA,CAAM,GAAG,EAAE,CAAC,CAAA,EAAG,MAAK,IAAK,EAAA;AACjD,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,SAAS,CAAA;AACzC,EAAA,MAAM,cAAA,GAAiB,cAAc,SAAS,CAAA;AAE9C,EAAA,MAAM,KAAA,GAAQ;AAAA,IACZ,KAAA,EAAO,KAAK,SAAA,IAAa,aAAA;AAAA,IACzB,YAAY,MAAA,CAAO,CAAA,EAAG,EAAE,CAAA,CAAA,EAAI,SAAS,CAAA,CAAE,CAAA;AAAA,IACvC,SAAA,EAAA,iBAAW,IAAI,IAAA,EAAK,EAAE,WAAA,EAAY;AAAA,IAClC,UAAA,EAAY;AAAA,MACV,uBAAA,EAAyB,KAAA;AAAA,MACzB,cAAc,MAAA,GAAS,CAAA,EAAG,MAAM,CAAA,EAAG,QAAQ,CAAA,CAAA,GAAK,QAAA;AAAA,MAChD,IAAA,EAAM,QAAA;AAAA,MACN,UAAA,EAAY,SAAA;AAAA,MACZ,WAAW,cAAA,CAAe,OAAA;AAAA,MAC1B,UAAU,cAAA,CAAe,KAAA;AAAA,MACzB,aAAa,cAAA,CAAe,IAAA;AAAA,MAC5B,mBAAmB,cAAA,CAAe,eAAA;AAAA,MAClC,OAAA;AAAA,MACA,MAAA,EAAQ,KAAK,MAAA,IAAU,IAAA;AAAA,MACvB,GAAG,IAAA,CAAK;AAAA;AACV,GACF;AAEA,EAAA,IAAI;AACF,IAAA,MAAM,IAAA,CAAK,SAAA,CAAU,OAAA,CAAQ,KAAK,CAAA;AAAA,EACpC,CAAA,CAAA,MAAQ;AAAA,EAER;AACF;;;ACpCO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,OAAA,GAAU,OAAO,IAAA,IAAQ,0BAAA;AAC/B,EAAA,MAAM,IAAA,GAAA,CAAQ,cAAA,CAAe,IAAA,CAAK,OAAO,CAAA,GAAI,OAAA,GAAU,CAAA,QAAA,EAAW,OAAO,CAAA,CAAA,EAAI,OAAA,CAAQ,KAAA,EAAO,EAAE,CAAA;AAC9F,EAAA,MAAM,QAAQ,MAAA,CAAO,IAAA,IAAQ,UAAA,EAAY,OAAA,CAAQ,WAAW,GAAG,CAAA;AAC/D,EAAA,MAAM,QAAA,GAAW,CAAA,EAAG,IAAI,CAAA,EAAG,IAAI,CAAA,CAAA;AAC/B,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AAEtC,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,OAAA,GAAU;AAAA,QACd,SAAS,MAAA,CAAO,MAAA;AAAA,QAChB,OAAO,KAAA,CAAM,KAAA;AAAA,QACb,aAAa,KAAA,CAAM,UAAA;AAAA,QACnB,WAAW,KAAA,CAAM,SAAA;AAAA,QACjB,YAAY,KAAA,CAAM;AAAA,OACpB;AACA,MAAA,MAAM,UAAU,QAAA,EAAU;AAAA,QACxB,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS,EAAE,cAAA,EAAgB,kBAAA,EAAmB;AAAA,QAC9C,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,OAAO,CAAA;AAAA,QAC5B,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC/BO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AACtC,EAAA,MAAM,SAAA,GAAY,MAAA,CAAO,SAAA,KAAc,CAAC,CAAA,KAA6B,CAAA,CAAA;AAErE,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,SAAA,CAAU,OAAO,GAAA,EAAK;AAAA,QAC1B,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS;AAAA,UACP,cAAA,EAAgB,kBAAA;AAAA,UAChB,GAAI,MAAA,CAAO,OAAA,IAAW;AAAC,SACzB;AAAA,QACA,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,SAAA,CAAU,KAAK,CAAC,CAAA;AAAA,QACrC,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC3BO,SAAS,gBACd,OAAA,EACkB;AAClB,EAAA,OAAO,EAAE,OAAA,EAAQ;AACnB","file":"index.js","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the UA:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'browser'` — looks like a real browser. Could be a genuine user or\n * a Playwright-based agent (Aider, OpenCode) that can't be distinguished\n * at the UA layer.\n * - `'other'` — unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n}\n\n/**\n * One-stop classification of a user-agent. Combines {@link isAiBot},\n * {@link isHttpClient}, and {@link parseBotName} into a single structured\n * result. Used internally by `trackVisit` to populate event properties;\n * useful in consumer code when you need all signals at once.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n","/**\n * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to\n * build stable anonymous distinct-ids from `ip:ua:...` tuples without\n * collecting identifying data. Not cryptographic — collisions are fine for\n * analytics segmentation.\n */\nexport function hashId(input: string): string {\n let h = 5381\n for (let i = 0; i < input.length; i++) {\n h = ((h << 5) + h + input.charCodeAt(i)) & 0xffffffff\n }\n return 'anon_' + (h >>> 0).toString(16)\n}\n","import { classifyAgent, isAiBot, isHttpClient } from './bots.js'\nimport { hashId } from './hash.js'\nimport type { TrackVisitOptions } from './types.js'\n\n/**\n * Capture an event describing the incoming request. Fire-and-forget: awaits\n * the adapter but swallows errors so a downed analytics backend never breaks\n * the response path. Callers typically don't await the returned promise.\n *\n * By default, captures every request so coding-agent traffic (axios, curl,\n * Electron, …) shows up alongside branded crawlers. Set `onlyBots: true` to\n * restrict capture to UAs matching {@link AI_BOT_PATTERN}.\n */\nexport async function trackVisit(\n req: Request,\n opts: TrackVisitOptions\n): Promise<void> {\n const userAgent = req.headers.get('user-agent') || ''\n\n const onlyBots = opts.onlyBots ?? false\n const skipBrowsers = opts.skipBrowsers ?? false\n if (onlyBots && !isAiBot(userAgent)) return\n if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) return\n\n let pathname = '/'\n let originFromUrl = ''\n try {\n const url = new URL(req.url)\n pathname = url.pathname\n originFromUrl = url.origin\n } catch {\n // Some runtimes hand us a relative URL; fall back to the raw string.\n pathname = req.url || '/'\n }\n const origin = opts.origin ?? originFromUrl\n\n const forwardedFor = req.headers.get('x-forwarded-for') || ''\n const ip = forwardedFor.split(',')[0]?.trim() ?? ''\n const referer = req.headers.get('referer')\n const classification = classifyAgent(userAgent)\n\n const event = {\n event: opts.eventName ?? 'agent_visit',\n distinctId: hashId(`${ip}:${userAgent}`),\n timestamp: new Date().toISOString(),\n properties: {\n $process_person_profile: false,\n $current_url: origin ? `${origin}${pathname}` : pathname,\n path: pathname,\n user_agent: userAgent,\n is_ai_bot: classification.isAiBot,\n bot_name: classification.label,\n ua_category: classification.kind,\n coding_agent_hint: classification.codingAgentHint,\n referer,\n source: opts.source ?? null,\n ...opts.properties\n }\n }\n\n try {\n await opts.analytics.capture(event)\n } catch {\n // Intentional swallow — analytics failures must not affect the response.\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface PostHogAdapterConfig {\n /** PostHog project API key (the public one used by the JS SDK). */\n apiKey: string\n /**\n * PostHog host, with or without scheme. Defaults to `https://us.i.posthog.com`.\n * Use `https://eu.i.posthog.com` for EU cloud, or your own reverse-proxy\n * domain (e.g. `https://svc.example.com`).\n */\n host?: string\n /**\n * Path on the host that accepts single-event captures. Defaults to\n * `/i/v0/e/` which is PostHog's current endpoint for this.\n */\n path?: string\n /**\n * Override the `fetch` implementation (useful for tests or custom runtimes\n * that need a pinned fetch).\n */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that posts each event to the PostHog capture endpoint. Uses\n * `keepalive: true` so the request survives after a serverless response\n * returns — events aren't guaranteed (fire-and-forget), but that's the\n * trade we want to keep the hot path fast.\n */\nexport function posthogAnalytics(config: PostHogAdapterConfig): AnalyticsAdapter {\n const hostRaw = config.host ?? 'https://us.i.posthog.com'\n const base = (/^https?:\\/\\//.test(hostRaw) ? hostRaw : `https://${hostRaw}`).replace(/\\/$/, '')\n const path = (config.path ?? '/i/v0/e/').replace(/^(?!\\/)/, '/')\n const endpoint = `${base}${path}`\n const fetchImpl = config.fetchImpl ?? fetch\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n const payload = {\n api_key: config.apiKey,\n event: event.event,\n distinct_id: event.distinctId,\n timestamp: event.timestamp,\n properties: event.properties\n }\n await fetchImpl(endpoint, {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(payload),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface WebhookAdapterConfig {\n /** Destination URL that receives a POST for each event. */\n url: string\n /** Extra headers merged onto the POST (useful for shared-secret auth). */\n headers?: Record<string, string>\n /**\n * Transform the event into the exact JSON body the destination expects.\n * Defaults to sending the {@link CaptureEvent} as-is.\n */\n transform?: (event: CaptureEvent) => unknown\n /** Override the `fetch` implementation. */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that POSTs each event to an arbitrary webhook URL. Keeps the\n * library analytics-backend-agnostic — use this when PostHog isn't your\n * analytics of record, or when you want to multiplex events through your\n * own ingestion layer.\n */\nexport function webhookAnalytics(config: WebhookAdapterConfig): AnalyticsAdapter {\n const fetchImpl = config.fetchImpl ?? fetch\n const transform = config.transform ?? ((e: CaptureEvent): unknown => e)\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n await fetchImpl(config.url, {\n method: 'POST',\n headers: {\n 'Content-Type': 'application/json',\n ...(config.headers ?? {})\n },\n body: JSON.stringify(transform(event)),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\n/**\n * Escape hatch for wiring a callback directly as an analytics adapter.\n * Useful when you want to log events, pipe them through your own SDK, or\n * compose multiple adapters.\n *\n * @example\n * ```ts\n * const devAnalytics = customAnalytics((e) => console.log('[doc_view]', e))\n * ```\n */\nexport function customAnalytics(\n capture: (event: CaptureEvent) => Promise<void> | void\n): AnalyticsAdapter {\n return { capture }\n}\n"]}
1
+ {"version":3,"sources":["../src/bots.ts","../src/hash.ts","../src/track.ts","../src/adapters/posthog.ts","../src/adapters/webhook.ts","../src/adapters/custom.ts"],"names":[],"mappings":";AAgBO,IAAM,cAAA,GACX;AAmBK,IAAM,mBAAA,GACX;AAEK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;AAEO,SAAS,aAAa,SAAA,EAA+C;AAC1E,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,mBAAA,CAAoB,KAAK,SAAS,CAAA;AAC3C;AAeO,SAAS,aAAa,SAAA,EAA8C;AACzE,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,CAAA,GAAI,UAAU,WAAA,EAAY;AAGhC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,IAAK,CAAA,CAAE,SAAS,QAAQ,CAAA;AAC1G,IAAA,OAAO,SAAA;AACT,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AAC5F,EAAA,IAAI,CAAA,CAAE,SAAS,eAAe,CAAA,IAAK,EAAE,QAAA,CAAS,iBAAiB,GAAG,OAAO,YAAA;AACzE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,cAAA;AAChC,EAAA,IAAI,CAAA,CAAE,SAAS,iBAAiB,CAAA,IAAK,EAAE,QAAA,CAAS,WAAW,GAAG,OAAO,QAAA;AACrE,EAAA,IAAI,CAAA,CAAE,SAAS,mBAAmB,CAAA,IAAK,EAAE,QAAA,CAAS,UAAU,GAAG,OAAO,OAAA;AACtE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,MAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,SAAS,oBAAoB,CAAA,IAAK,EAAE,QAAA,CAAS,aAAa,GAAG,OAAO,MAAA;AAC1E,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,SAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,eAAe,CAAA,EAAG,OAAO,YAAA;AACxC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAGnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,UAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,KAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,KAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,gBAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,aAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AAGtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,aAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,EAAG,OAAO,SAAA;AAClC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,YAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,WAAA;AACzC,EAAA,IAAI,CAAA,CAAE,SAAS,cAAc,CAAA,IAAK,EAAE,QAAA,CAAS,SAAS,GAAG,OAAO,SAAA;AAChE,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,OAAA;AACnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,SAAA;AACrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,UAAA;AACtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,UAAU,CAAA,EAAG,OAAO,UAAA;AAGnC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,OAAO,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,QAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,aAAa,CAAA,EAAG,OAAO,OAAA;AAGtC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,qBAAqB,CAAA,EAAG,OAAO,UAAA;AAC9C,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,kBAAkB,CAAA,EAAG,OAAO,mBAAA;AAG3C,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,SAAA;AAGjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG,OAAO,iBAAA;AACzC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,WAAA;AACpC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,MAAM,CAAA,EAAG,OAAO,MAAA;AAC/B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,EAAG,OAAO,QAAA;AACjC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,YAAY,CAAA,EAAG,OAAO,QAAA;AAIrC,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,WAAW,CAAA,EAAG,OAAO,UAAA;AACpC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAC7B,EAAA,IAAI,SAAA,CAAU,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAC9B,EAAA,IAAI,4BAAA,CAA6B,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,KAAA;AACjD,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,OAAA;AAChC,EAAA,IAAI,cAAA,CAAe,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,YAAA;AACnC,EAAA,IAAI,mBAAA,CAAoB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,iBAAA;AACxC,EAAA,IAAI,kBAAA,CAAmB,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,gBAAA;AACvC,EAAA,IAAI,UAAA,CAAW,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,QAAA;AAC/B,EAAA,IAAI,WAAA,CAAY,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,SAAA;AAChC,EAAA,IAAI,QAAA,CAAS,IAAA,CAAK,CAAC,CAAA,EAAG,OAAO,MAAA;AAG7B,EAAA,IAAI,CAAA,CAAE,QAAA,CAAS,SAAS,CAAA,IAAK,EAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,QAAA,CAAS,QAAQ,CAAA,IAAK,CAAA,CAAE,SAAS,SAAS,CAAA;AAC/F,IAAA,OAAO,SAAA;AAET,EAAA,OAAO,OAAA;AACT;AAOO,SAAS,sBAAsB,SAAA,EAA8C;AAClF,EAAA,IAAI,CAAC,SAAA,IAAa,OAAO,SAAA,KAAc,UAAU,OAAO,OAAA;AACxD,EAAA,MAAM,eAAA,GAAkB,SAAA,CAAU,KAAA,CAAM,yCAAyC,CAAA;AACjF,EAAA,IAAI,eAAA,IAAmB,gBAAgB,CAAC,CAAA,SAAU,eAAA,CAAgB,CAAC,EAAE,IAAA,EAAK;AAC1E,EAAA,MAAM,QAAQ,SAAA,CAAU,IAAA,EAAK,CAAE,KAAA,CAAM,GAAG,CAAA,CAAE,CAAC,CAAA,EAAG,IAAA,GAAO,KAAA,CAAM,KAAK,CAAA,CAAE,CAAC,GAAG,IAAA,EAAK;AAC3E,EAAA,OAAO,KAAA,IAAS,OAAA;AAClB;AAkBO,SAAS,eAAe,GAAA,EAAiC;AAC9D,EAAA,MAAM,UAAoB,EAAC;AAC3B,EAAA,MAAM,MAAM,GAAA,CAAI,OAAA,CAAQ,IAAI,YAAY,CAAA,IAAK,IAAI,WAAA,EAAY;AAC7D,EAAA,MAAM,WAAA,GACJ,EAAA,CAAG,QAAA,CAAS,SAAS,KAAK,EAAA,CAAG,QAAA,CAAS,QAAQ,CAAA,IAAK,GAAG,QAAA,CAAS,QAAQ,CAAA,IAAK,EAAA,CAAG,SAAS,SAAS,CAAA;AAEnG,EAAA,IAAI,CAAC,WAAA,EAAa,OAAO,EAAE,KAAA,EAAO,GAAG,OAAA,EAAS,EAAC,EAAG,MAAA,EAAQ,KAAA,EAAM;AAEhE,EAAA,IAAI,CAAC,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,EAAG;AACvC,IAAA,OAAA,CAAQ,KAAK,yBAAyB,CAAA;AAAA,EACxC;AACA,EAAA,IAAI,CAAC,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,gBAAgB,CAAA,EAAG;AACtC,IAAA,OAAA,CAAQ,KAAK,wBAAwB,CAAA;AAAA,EACvC;AACA,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,WAAW,CAAA;AAC3C,EAAA,IAAI,CAAC,OAAA,EAAS;AACZ,IAAA,OAAA,CAAQ,KAAK,mBAAmB,CAAA;AAAA,EAClC,WAAW,OAAA,CAAQ,WAAA,EAAY,CAAE,QAAA,CAAS,gBAAgB,CAAA,EAAG;AAC3D,IAAA,OAAA,CAAQ,KAAK,sBAAsB,CAAA;AAAA,EACrC;AACA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,CAAC,MAAA,IAAU,MAAA,KAAW,KAAA,EAAO;AAC/B,IAAA,OAAA,CAAQ,KAAK,wBAAwB,CAAA;AAAA,EACvC;AACA,EAAA,IAAA,CAAK,GAAA,CAAI,QAAQ,GAAA,CAAI,YAAY,KAAK,EAAA,EAAI,WAAA,OAAkB,OAAA,EAAS;AACnE,IAAA,OAAA,CAAQ,KAAK,kBAAkB,CAAA;AAAA,EACjC;AAEA,EAAA,MAAM,QAAQ,OAAA,CAAQ,MAAA;AACtB,EAAA,OAAO,EAAE,KAAA,EAAO,OAAA,EAAS,MAAA,EAAQ,SAAS,CAAA,EAAE;AAC9C;AA6CO,SAAS,cAAc,SAAA,EAA2D;AACvF,EAAA,MAAM,KAAA,GAAQ,aAAa,SAAS,CAAA;AACpC,EAAA,MAAM,KAAA,GAAQ,QAAQ,SAAS,CAAA;AAC/B,EAAA,MAAM,UAAA,GAAa,aAAa,SAAS,CAAA;AAEzC,EAAA,IAAI,IAAA;AACJ,EAAA,IAAI,OAAO,IAAA,GAAO,kBAAA;AAAA,OAAA,IACT,YAAY,IAAA,GAAO,mBAAA;AAAA,OAAA,IACnB,KAAA,KAAU,WAAW,IAAA,GAAO,SAAA;AAAA,OAChC,IAAA,GAAO,OAAA;AAEZ,EAAA,OAAO,EAAE,IAAA,EAAM,KAAA,EAAO,OAAA,EAAS,KAAA,EAAO,iBAAiB,UAAA,EAAW;AACpE;AAOO,SAAS,gBAAgB,GAAA,EAAmC;AACjE,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AACnD,EAAA,MAAM,IAAA,GAAO,cAAc,SAAS,CAAA;AACpC,EAAA,MAAM,QAAA,GAAW,eAAe,GAAG,CAAA;AAEnC,EAAA,IAAI,OAAO,IAAA,CAAK,IAAA;AAChB,EAAA,IAAI,IAAA,KAAS,SAAA,IAAa,QAAA,CAAS,MAAA,EAAQ;AACzC,IAAA,IAAA,GAAO,iBAAA;AAAA,EACT;AAEA,EAAA,OAAO,EAAE,GAAG,IAAA,EAAM,IAAA,EAAM,QAAA,EAAS;AACnC;;;AC3RO,SAAS,OAAO,KAAA,EAAuB;AAC5C,EAAA,IAAI,CAAA,GAAI,IAAA;AACR,EAAA,KAAA,IAAS,CAAA,GAAI,CAAA,EAAG,CAAA,GAAI,KAAA,CAAM,QAAQ,CAAA,EAAA,EAAK;AACrC,IAAA,CAAA,GAAA,CAAM,KAAK,CAAA,IAAK,CAAA,GAAI,KAAA,CAAM,UAAA,CAAW,CAAC,CAAA,GAAK,UAAA;AAAA,EAC7C;AACA,EAAA,OAAO,OAAA,GAAA,CAAW,CAAA,KAAM,CAAA,EAAG,QAAA,CAAS,EAAE,CAAA;AACxC;;;ACCA,eAAsB,UAAA,CACpB,KACA,IAAA,EACe;AACf,EAAA,MAAM,SAAA,GAAY,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAEnD,EAAA,MAAM,QAAA,GAAW,KAAK,QAAA,IAAY,KAAA;AAClC,EAAA,MAAM,YAAA,GAAe,KAAK,YAAA,IAAgB,KAAA;AAC1C,EAAA,IAAI,QAAA,IAAY,CAAC,OAAA,CAAQ,SAAS,CAAA,EAAG;AACrC,EAAA,IAAI,YAAA,IAAgB,CAAC,OAAA,CAAQ,SAAS,KAAK,CAAC,YAAA,CAAa,SAAS,CAAA,EAAG;AAInE,IAAA,IAAI,CAAC,cAAA,CAAe,GAAG,CAAA,CAAE,MAAA,EAAQ;AAAA,EACnC;AAEA,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI,aAAA,GAAgB,EAAA;AACpB,EAAA,IAAI;AACF,IAAA,MAAM,GAAA,GAAM,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA;AAC3B,IAAA,QAAA,GAAW,GAAA,CAAI,QAAA;AACf,IAAA,aAAA,GAAgB,GAAA,CAAI,MAAA;AAAA,EACtB,CAAA,CAAA,MAAQ;AAEN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AACA,EAAA,MAAM,MAAA,GAAS,KAAK,MAAA,IAAU,aAAA;AAE9B,EAAA,MAAM,YAAA,GAAe,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,iBAAiB,CAAA,IAAK,EAAA;AAC3D,EAAA,MAAM,EAAA,GAAK,aAAa,KAAA,CAAM,GAAG,EAAE,CAAC,CAAA,EAAG,MAAK,IAAK,EAAA;AACjD,EAAA,MAAM,OAAA,GAAU,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,SAAS,CAAA;AACzC,EAAA,MAAM,cAAA,GAAiB,gBAAgB,GAAG,CAAA;AAE1C,EAAA,MAAM,KAAA,GAAQ;AAAA,IACZ,KAAA,EAAO,KAAK,SAAA,IAAa,aAAA;AAAA,IACzB,YAAY,MAAA,CAAO,CAAA,EAAG,EAAE,CAAA,CAAA,EAAI,SAAS,CAAA,CAAE,CAAA;AAAA,IACvC,SAAA,EAAA,iBAAW,IAAI,IAAA,EAAK,EAAE,WAAA,EAAY;AAAA,IAClC,UAAA,EAAY;AAAA,MACV,uBAAA,EAAyB,KAAA;AAAA,MACzB,cAAc,MAAA,GAAS,CAAA,EAAG,MAAM,CAAA,EAAG,QAAQ,CAAA,CAAA,GAAK,QAAA;AAAA,MAChD,IAAA,EAAM,QAAA;AAAA,MACN,UAAA,EAAY,SAAA;AAAA,MACZ,WAAW,cAAA,CAAe,OAAA;AAAA,MAC1B,UAAU,cAAA,CAAe,KAAA;AAAA,MACzB,aAAa,cAAA,CAAe,IAAA;AAAA,MAC5B,mBAAmB,cAAA,CAAe,eAAA;AAAA,MAClC,cAAA,EAAgB,cAAA,CAAe,QAAA,EAAU,KAAA,IAAS,CAAA;AAAA,MAClD,eAAA,EAAiB,cAAA,CAAe,QAAA,EAAU,MAAA,IAAU,KAAA;AAAA,MACpD,OAAA;AAAA,MACA,MAAA,EAAQ,KAAK,MAAA,IAAU,IAAA;AAAA,MACvB,GAAG,IAAA,CAAK;AAAA;AACV,GACF;AAEA,EAAA,IAAI;AACF,IAAA,MAAM,IAAA,CAAK,SAAA,CAAU,OAAA,CAAQ,KAAK,CAAA;AAAA,EACpC,CAAA,CAAA,MAAQ;AAAA,EAER;AACF;;;AC3CO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,OAAA,GAAU,OAAO,IAAA,IAAQ,0BAAA;AAC/B,EAAA,MAAM,IAAA,GAAA,CAAQ,cAAA,CAAe,IAAA,CAAK,OAAO,CAAA,GAAI,OAAA,GAAU,CAAA,QAAA,EAAW,OAAO,CAAA,CAAA,EAAI,OAAA,CAAQ,KAAA,EAAO,EAAE,CAAA;AAC9F,EAAA,MAAM,QAAQ,MAAA,CAAO,IAAA,IAAQ,UAAA,EAAY,OAAA,CAAQ,WAAW,GAAG,CAAA;AAC/D,EAAA,MAAM,QAAA,GAAW,CAAA,EAAG,IAAI,CAAA,EAAG,IAAI,CAAA,CAAA;AAC/B,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AAEtC,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,OAAA,GAAU;AAAA,QACd,SAAS,MAAA,CAAO,MAAA;AAAA,QAChB,OAAO,KAAA,CAAM,KAAA;AAAA,QACb,aAAa,KAAA,CAAM,UAAA;AAAA,QACnB,WAAW,KAAA,CAAM,SAAA;AAAA,QACjB,YAAY,KAAA,CAAM;AAAA,OACpB;AACA,MAAA,MAAM,UAAU,QAAA,EAAU;AAAA,QACxB,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS,EAAE,cAAA,EAAgB,kBAAA,EAAmB;AAAA,QAC9C,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,OAAO,CAAA;AAAA,QAC5B,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC/BO,SAAS,iBAAiB,MAAA,EAAgD;AAC/E,EAAA,MAAM,SAAA,GAAY,OAAO,SAAA,IAAa,KAAA;AACtC,EAAA,MAAM,SAAA,GAAY,MAAA,CAAO,SAAA,KAAc,CAAC,CAAA,KAA6B,CAAA,CAAA;AAErE,EAAA,OAAO;AAAA,IACL,MAAM,QAAQ,KAAA,EAAoC;AAChD,MAAA,MAAM,SAAA,CAAU,OAAO,GAAA,EAAK;AAAA,QAC1B,MAAA,EAAQ,MAAA;AAAA,QACR,OAAA,EAAS;AAAA,UACP,cAAA,EAAgB,kBAAA;AAAA,UAChB,GAAI,MAAA,CAAO,OAAA,IAAW;AAAC,SACzB;AAAA,QACA,IAAA,EAAM,IAAA,CAAK,SAAA,CAAU,SAAA,CAAU,KAAK,CAAC,CAAA;AAAA,QACrC,SAAA,EAAW;AAAA,OACZ,CAAA;AAAA,IACH;AAAA,GACF;AACF;;;AC3BO,SAAS,gBACd,OAAA,EACkB;AAClB,EAAA,OAAO,EAAE,OAAA,EAAQ;AACnB","file":"index.js","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // SEO crawlers and monitoring bots.\n if (s.includes('ahrefsbot')) return 'Ahrefs'\n if (s.includes('semrushbot')) return 'Semrush'\n if (s.includes('mj12bot')) return 'Majestic'\n if (s.includes('dotbot')) return 'Moz'\n if (s.includes('rogerbot')) return 'Moz'\n if (s.includes('screaming frog')) return 'Screaming Frog'\n if (s.includes('sitebulb')) return 'Sitebulb'\n if (s.includes('linkfluence')) return 'Linkfluence'\n if (s.includes('dataforseo')) return 'DataForSEO'\n if (s.includes('serpstatbot')) return 'Serpstat'\n\n // Monitoring and feed bots.\n if (s.includes('uptimerobot')) return 'UptimeRobot'\n if (s.includes('pingdom')) return 'Pingdom'\n if (s.includes('statuscake')) return 'StatusCake'\n if (s.includes('newrelicpinger')) return 'New Relic'\n if (s.includes('datadogagent') || s.includes('datadog')) return 'Datadog'\n if (s.includes('slackbot')) return 'Slack'\n if (s.includes('twitterbot')) return 'Twitter'\n if (s.includes('linkedinbot')) return 'LinkedIn'\n if (s.includes('discordbot')) return 'Discord'\n if (s.includes('telegrambot')) return 'Telegram'\n if (s.includes('whatsapp')) return 'WhatsApp'\n\n // AI search and indexing bots.\n if (s.includes('linkupbot')) return 'Linkup'\n if (s.includes('sogou')) return 'Sogou'\n if (s.includes('yandexbot')) return 'Yandex'\n if (s.includes('baiduspider')) return 'Baidu'\n\n // Link preview fetchers.\n if (s.includes('facebookexternalhit')) return 'Facebook'\n if (s.includes('com.apple.webkit')) return 'Apple URL Preview'\n\n // Uptime and monitoring.\n if (s.includes('ohdear')) return 'Oh Dear'\n\n // Generic scrapers.\n if (s.includes('scrapy')) return 'Scrapy'\n if (s.includes('headlesschrome')) return 'Headless Chrome'\n if (s.includes('phantomjs')) return 'PhantomJS'\n if (s.includes('wget')) return 'wget'\n if (s.includes('httpie')) return 'HTTPie'\n if (s.includes('guzzlehttp')) return 'Guzzle'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\n/**\n * Detect likely headless/automated browsers by checking for missing headers\n * that real browsers always send. Playwright, Puppeteer, and similar tools\n * spoof the UA but often omit standard browser headers.\n *\n * Signals checked (each scores 1 point):\n * - Missing `Accept-Language` — every real browser sends this\n * - Missing `Sec-Fetch-Mode` — sent by all modern browsers\n * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+\n * - `Sec-CH-UA` contains \"HeadlessChrome\"\n * - Missing or bare Accept header — browsers send detailed accept lists\n * - `Connection: close` with browser UA — browsers use keep-alive\n *\n * Returns a score (0-6), the signals that fired, and a boolean `likely`\n * flag (score >= 2 with a browser-like UA).\n */\nexport function detectHeadless(req: Request): HeadlessDetection {\n const signals: string[] = []\n const ua = (req.headers.get('user-agent') || '').toLowerCase()\n const isBrowserUA =\n ua.includes('mozilla') || ua.includes('chrome') || ua.includes('safari') || ua.includes('firefox')\n\n if (!isBrowserUA) return { score: 0, signals: [], likely: false }\n\n if (!req.headers.get('accept-language')) {\n signals.push('missing-accept-language')\n }\n if (!req.headers.get('sec-fetch-mode')) {\n signals.push('missing-sec-fetch-mode')\n }\n const secChUa = req.headers.get('sec-ch-ua')\n if (!secChUa) {\n signals.push('missing-sec-ch-ua')\n } else if (secChUa.toLowerCase().includes('headlesschrome')) {\n signals.push('headless-chrome-hint')\n }\n const accept = req.headers.get('accept') || ''\n if (!accept || accept === '*/*') {\n signals.push('missing-or-bare-accept')\n }\n if ((req.headers.get('connection') || '').toLowerCase() === 'close') {\n signals.push('connection-close')\n }\n\n const score = signals.length\n return { score, signals, likely: score >= 2 }\n}\n\nexport interface HeadlessDetection {\n /** Number of suspicious signals found (0-6). */\n score: number\n /** Names of the specific signals that fired. */\n signals: string[]\n /** True when score >= 2 — strong headless indication. */\n likely: boolean\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'headless-likely'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the request:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'headless-likely'` — Browser-like UA but missing standard headers.\n * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).\n * - `'browser'` — Looks like a real browser with expected headers present.\n * - `'other'` — Unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n /** Headless browser detection result. Only populated when `req` is passed. */\n headless?: HeadlessDetection\n}\n\n/**\n * UA-only classification. Use {@link classifyRequest} for full detection\n * including headless browser heuristics.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n\n/**\n * Full request classification — combines UA parsing with header-based\n * headless detection. When a browser-like UA is missing standard headers,\n * the kind is promoted from `'browser'` to `'headless-likely'`.\n */\nexport function classifyRequest(req: Request): AgentClassification {\n const userAgent = req.headers.get('user-agent') || ''\n const base = classifyAgent(userAgent)\n const headless = detectHeadless(req)\n\n let kind = base.kind\n if (kind === 'browser' && headless.likely) {\n kind = 'headless-likely'\n }\n\n return { ...base, kind, headless }\n}\n","/**\n * djb2 hash returning an 8-char hex string prefixed with `anon_`. Used to\n * build stable anonymous distinct-ids from `ip:ua:...` tuples without\n * collecting identifying data. Not cryptographic — collisions are fine for\n * analytics segmentation.\n */\nexport function hashId(input: string): string {\n let h = 5381\n for (let i = 0; i < input.length; i++) {\n h = ((h << 5) + h + input.charCodeAt(i)) & 0xffffffff\n }\n return 'anon_' + (h >>> 0).toString(16)\n}\n","import { classifyRequest, detectHeadless, isAiBot, isHttpClient } from './bots.js'\nimport { hashId } from './hash.js'\nimport type { TrackVisitOptions } from './types.js'\n\n/**\n * Capture an event describing the incoming request. Fire-and-forget: awaits\n * the adapter but swallows errors so a downed analytics backend never breaks\n * the response path. Callers typically don't await the returned promise.\n *\n * By default, captures every request so coding-agent traffic (axios, curl,\n * Electron, …) shows up alongside branded crawlers. Set `onlyBots: true` to\n * restrict capture to UAs matching {@link AI_BOT_PATTERN}.\n */\nexport async function trackVisit(\n req: Request,\n opts: TrackVisitOptions\n): Promise<void> {\n const userAgent = req.headers.get('user-agent') || ''\n\n const onlyBots = opts.onlyBots ?? false\n const skipBrowsers = opts.skipBrowsers ?? false\n if (onlyBots && !isAiBot(userAgent)) return\n if (skipBrowsers && !isAiBot(userAgent) && !isHttpClient(userAgent)) {\n // Not a declared bot or HTTP client — check headless heuristics.\n // Playwright-based agents (Aider, OpenCode) will pass if they're missing\n // standard browser headers. Real browsers get skipped.\n if (!detectHeadless(req).likely) return\n }\n\n let pathname = '/'\n let originFromUrl = ''\n try {\n const url = new URL(req.url)\n pathname = url.pathname\n originFromUrl = url.origin\n } catch {\n // Some runtimes hand us a relative URL; fall back to the raw string.\n pathname = req.url || '/'\n }\n const origin = opts.origin ?? originFromUrl\n\n const forwardedFor = req.headers.get('x-forwarded-for') || ''\n const ip = forwardedFor.split(',')[0]?.trim() ?? ''\n const referer = req.headers.get('referer')\n const classification = classifyRequest(req)\n\n const event = {\n event: opts.eventName ?? 'agent_visit',\n distinctId: hashId(`${ip}:${userAgent}`),\n timestamp: new Date().toISOString(),\n properties: {\n $process_person_profile: false,\n $current_url: origin ? `${origin}${pathname}` : pathname,\n path: pathname,\n user_agent: userAgent,\n is_ai_bot: classification.isAiBot,\n bot_name: classification.label,\n ua_category: classification.kind,\n coding_agent_hint: classification.codingAgentHint,\n headless_score: classification.headless?.score ?? 0,\n headless_likely: classification.headless?.likely ?? false,\n referer,\n source: opts.source ?? null,\n ...opts.properties\n }\n }\n\n try {\n await opts.analytics.capture(event)\n } catch {\n // Intentional swallow — analytics failures must not affect the response.\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface PostHogAdapterConfig {\n /** PostHog project API key (the public one used by the JS SDK). */\n apiKey: string\n /**\n * PostHog host, with or without scheme. Defaults to `https://us.i.posthog.com`.\n * Use `https://eu.i.posthog.com` for EU cloud, or your own reverse-proxy\n * domain (e.g. `https://svc.example.com`).\n */\n host?: string\n /**\n * Path on the host that accepts single-event captures. Defaults to\n * `/i/v0/e/` which is PostHog's current endpoint for this.\n */\n path?: string\n /**\n * Override the `fetch` implementation (useful for tests or custom runtimes\n * that need a pinned fetch).\n */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that posts each event to the PostHog capture endpoint. Uses\n * `keepalive: true` so the request survives after a serverless response\n * returns — events aren't guaranteed (fire-and-forget), but that's the\n * trade we want to keep the hot path fast.\n */\nexport function posthogAnalytics(config: PostHogAdapterConfig): AnalyticsAdapter {\n const hostRaw = config.host ?? 'https://us.i.posthog.com'\n const base = (/^https?:\\/\\//.test(hostRaw) ? hostRaw : `https://${hostRaw}`).replace(/\\/$/, '')\n const path = (config.path ?? '/i/v0/e/').replace(/^(?!\\/)/, '/')\n const endpoint = `${base}${path}`\n const fetchImpl = config.fetchImpl ?? fetch\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n const payload = {\n api_key: config.apiKey,\n event: event.event,\n distinct_id: event.distinctId,\n timestamp: event.timestamp,\n properties: event.properties\n }\n await fetchImpl(endpoint, {\n method: 'POST',\n headers: { 'Content-Type': 'application/json' },\n body: JSON.stringify(payload),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\nexport interface WebhookAdapterConfig {\n /** Destination URL that receives a POST for each event. */\n url: string\n /** Extra headers merged onto the POST (useful for shared-secret auth). */\n headers?: Record<string, string>\n /**\n * Transform the event into the exact JSON body the destination expects.\n * Defaults to sending the {@link CaptureEvent} as-is.\n */\n transform?: (event: CaptureEvent) => unknown\n /** Override the `fetch` implementation. */\n fetchImpl?: typeof fetch\n}\n\n/**\n * Adapter that POSTs each event to an arbitrary webhook URL. Keeps the\n * library analytics-backend-agnostic — use this when PostHog isn't your\n * analytics of record, or when you want to multiplex events through your\n * own ingestion layer.\n */\nexport function webhookAnalytics(config: WebhookAdapterConfig): AnalyticsAdapter {\n const fetchImpl = config.fetchImpl ?? fetch\n const transform = config.transform ?? ((e: CaptureEvent): unknown => e)\n\n return {\n async capture(event: CaptureEvent): Promise<void> {\n await fetchImpl(config.url, {\n method: 'POST',\n headers: {\n 'Content-Type': 'application/json',\n ...(config.headers ?? {})\n },\n body: JSON.stringify(transform(event)),\n keepalive: true\n })\n }\n }\n}\n","import type { AnalyticsAdapter, CaptureEvent } from '../types.js'\n\n/**\n * Escape hatch for wiring a callback directly as an analytics adapter.\n * Useful when you want to log events, pipe them through your own SDK, or\n * compose multiple adapters.\n *\n * @example\n * ```ts\n * const devAnalytics = customAnalytics((e) => console.log('[doc_view]', e))\n * ```\n */\nexport function customAnalytics(\n capture: (event: CaptureEvent) => Promise<void> | void\n): AnalyticsAdapter {\n return { capture }\n}\n"]}
package/dist/markdown.cjs.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/bots.ts","../src/markdown.ts"],"names":[],"mappings":";;;AAgBO,IAAM,cAAA,GACX,2QAAA;AAsBK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;;;ACZO,SAAS,sBAAsB,GAAA,EAAuC;AAC3E,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI;AACF,IAAA,QAAA,GAAW,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA,CAAE,QAAA;AAAA,EAC9B,CAAA,CAAA,MAAQ;AACN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AAEA,EAAA,MAAM,EAAA,GAAK,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,OAAA,CAAQ,EAAE,CAAA,EAAG;AACf,IAAA,OAAO,EAAE,MAAA,EAAQ,YAAA,EAAc,YAAA,EAAc,QAAA,EAAS;AAAA,EACxD;AAEA,EAAA,IAAI,QAAA,CAAS,QAAA,CAAS,KAAK,CAAA,EAAG;AAC5B,IAAA,OAAO,EAAE,QAAQ,WAAA,EAAa,YAAA,EAAc,SAAS,OAAA,CAAQ,OAAA,EAAS,EAAE,CAAA,EAAE;AAAA,EAC5E;AAEA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,MAAA,CAAO,QAAA,CAAS,eAAe,CAAA,EAAG;AACpC,IAAA,OAAO,EAAE,MAAA,EAAQ,eAAA,EAAiB,YAAA,EAAc,QAAA,EAAS;AAAA,EAC3D;AAEA,EAAA,OAAO,IAAA;AACT;AAqBO,SAAS,eAAA,CAAgB,KAAA,GAA8B,EAAC,EAA2B;AACxF,EAAA,MAAM,OAAA,GAAkC;AAAA,IACtC,cAAA,EAAgB,8BAAA;AAAA,IAChB,gBAAA,EAAkB,MAAM,aAAA,IAAiB,uCAAA;AAAA,IACzC,IAAA,EAAM;AAAA,GACR;AACA,EAAA,IAAI,OAAO,KAAA,CAAM,MAAA,KAAW,QAAA,IAAY,KAAA,CAAM,SAAS,CAAA,EAAG;AACxD,IAAA,OAAA,CAAQ,mBAAmB,CAAA,GAAI,IAAA,CAAK,GAAA,CAAI,CAAA,EAAG,IAAA,CAAK,IAAA,CAAK,KAAA,CAAM,MAAM,CAAC,CAAA,CAAE,QAAA,EAAS;AAAA,EAC/E;AACA,EAAA,OAAO,OAAA;AACT;AAoBO,SAAS,0BAA0B,KAAA,EAAuC;AAC/E,EAAA,MAAM,IAAA,GACJ,KAAA,CAAM,QAAA,IAAA,CACL,MAAM;AACL,IAAA,IAAI;AACF,MAAA,OAAO,IAAI,GAAA,CAAI,KAAA,CAAM,MAAM,CAAA,CAAE,QAAA;AAAA,IAC/B,CAAA,CAAA,MAAQ;AACN,MAAA,OAAO,KAAA,CAAM,MAAA;AAAA,IACf;AAAA,EACF,CAAA,GAAG;AACL,EAAA,MAAM,MAAM,CAAA,EAAG,KAAA,CAAM,MAAM,CAAA,EAAG,MAAM,QAAQ,CAAA,CAAA;AAC5C,EAAA,MAAM,KAAA,GAAkB,CAAC,CAAA,EAAA,EAAK,IAAI,IAAI,EAAA,EAAI,CAAA,WAAA,EAAc,GAAG,CAAA,gDAAA,CAAA,EAAoD,EAAE,CAAA;AACjH,EAAA,MAAM,QAAkB,EAAC;AACzB,EAAA,IAAI,KAAA,CAAM,UAAA,EAAY,KAAA,CAAM,IAAA,CAAK,CAAA,GAAA,EAAM,MAAM,UAAU,CAAA,EAAA,EAAK,KAAA,CAAM,UAAU,CAAA,8BAAA,CAA2B,CAAA;AACvG,EAAA,IAAI,KAAA,CAAM,cAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,cAAc,CAAA,EAAA,EAAK,KAAA,CAAM,cAAc,CAAA,8BAAA,CAA2B,CAAA;AAC3F,EAAA,IAAI,KAAA,CAAM,gBAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,gBAAgB,CAAA,EAAA,EAAK,KAAA,CAAM,gBAAgB,CAAA,yCAAA,CAAsC,CAAA;AAC1G,EAAA,IAAI,MAAM,MAAA,EAAQ;AAChB,IAAA,KAAA,CAAM,IAAA,CAAK,0CAAA,EAA4C,EAAA,EAAI,GAAG,OAAO,EAAE,CAAA;AAAA,EACzE;AACA,EAAA,OAAO,KAAA,CAAM,KAAK,IAAI,CAAA;AACxB","file":"markdown.cjs","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the UA:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'browser'` — looks like a real browser. Could be a genuine user or\n * a Playwright-based agent (Aider, OpenCode) that can't be distinguished\n * at the UA layer.\n * - `'other'` — unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n}\n\n/**\n * One-stop classification of a user-agent. Combines {@link isAiBot},\n * {@link isHttpClient}, and {@link parseBotName} into a single structured\n * result. Used internally by `trackVisit` to populate event properties;\n * useful in consumer code when you need all signals at once.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n","import { isAiBot } from './bots.js'\n\nexport type MarkdownServeReason =\n | 'ua-rewrite'\n | 'md-suffix'\n | 'accept-header'\n\nexport interface MarkdownDecision {\n /** Why this request should be served Markdown. */\n reason: MarkdownServeReason\n /**\n * The request's original logical path, with any trailing `.md` stripped.\n * Use this when mapping to a mirror file.\n */\n strippedPath: string\n}\n\n/**\n * Decide whether the request should be served Markdown instead of HTML.\n * Returns `null` when the request should go through your normal handler.\n *\n * Covers three triggers:\n * - Known AI-bot UA on any URL (`ua-rewrite`)\n * - Explicit `.md` suffix on the URL (`md-suffix`)\n * - `Accept: text/markdown` header (`accept-header`)\n *\n * This helper intentionally does not perform the rewrite itself — routing is\n * framework-specific (NextResponse.rewrite for Next.js, ctx.rewrite for\n * Hono, etc.). Use the returned decision to build the appropriate response.\n */\nexport function markdownServeDecision(req: Request): MarkdownDecision | null {\n let pathname = '/'\n try {\n pathname = new URL(req.url).pathname\n } catch {\n pathname = req.url || '/'\n }\n\n const ua = req.headers.get('user-agent') || ''\n if (isAiBot(ua)) {\n return { reason: 'ua-rewrite', strippedPath: pathname }\n }\n\n if (pathname.endsWith('.md')) {\n return { reason: 'md-suffix', strippedPath: pathname.replace(/\\.md$/, '') }\n }\n\n const accept = req.headers.get('accept') || ''\n if (accept.includes('text/markdown')) {\n return { reason: 'accept-header', strippedPath: pathname }\n }\n\n return null\n}\n\nexport interface MarkdownHeadersInput {\n /**\n * If provided, rendered as `x-markdown-tokens` so agents can budget context\n * before parsing the body. Typically `Math.ceil(body.length / 4)`.\n */\n tokens?: number\n /**\n * Content-Signal directive (see contentsignals.org). Defaults to\n * `'search=yes, ai-input=yes, ai-train=no'` — change if you want to permit\n * training or restrict indexing.\n */\n contentSignal?: string\n}\n\n/**\n * Build the set of response headers to attach to a Markdown response. Safe\n * defaults: UTF-8 text/markdown, Vary: accept, and a Content-Signal directive\n * that permits search + agent input but denies training.\n */\nexport function markdownHeaders(input: MarkdownHeadersInput = {}): Record<string, string> {\n const headers: Record<string, string> = {\n 'Content-Type': 'text/markdown; charset=utf-8',\n 'Content-Signal': input.contentSignal ?? 'search=yes, ai-input=yes, ai-train=no',\n Vary: 'accept'\n }\n if (typeof input.tokens === 'number' && input.tokens > 0) {\n headers['x-markdown-tokens'] = Math.max(1, Math.ceil(input.tokens)).toString()\n }\n return headers\n}\n\nexport interface SynthesizePointerInput {\n origin: string\n pathname: string\n /** URL of the site's curated index, usually `/llms.txt`. */\n llmsTxtUrl?: string\n /** URL of the full enumerated index, usually `/llms-full.txt`. */\n llmsFullTxtUrl?: string\n /** URL of the machine-readable path manifest, usually `/md/index.json`. */\n markdownIndexUrl?: string\n /** Site name to title the pointer document. Defaults to the origin hostname. */\n siteName?: string\n}\n\n/**\n * Generate a minimal pointer Markdown document for URLs that don't have a\n * pre-built mirror. Keeps the `Accept: text/markdown` contract intact\n * site-wide — agents always get *something* parseable, not a 404.\n */\nexport function synthesizeMarkdownPointer(input: SynthesizePointerInput): string {\n const site =\n input.siteName ??\n (() => {\n try {\n return new URL(input.origin).hostname\n } catch {\n return input.origin\n }\n })()\n const url = `${input.origin}${input.pathname}`\n const lines: string[] = [`# ${site}`, '', `This page (${url}) does not have a dedicated Markdown mirror yet.`, '']\n const links: string[] = []\n if (input.llmsTxtUrl) links.push(`- [${input.llmsTxtUrl}](${input.llmsTxtUrl}) — curated index of docs`)\n if (input.llmsFullTxtUrl)\n links.push(`- [${input.llmsFullTxtUrl}](${input.llmsFullTxtUrl}) — full enumerated index`)\n if (input.markdownIndexUrl)\n links.push(`- [${input.markdownIndexUrl}](${input.markdownIndexUrl}) — JSON index of all Markdown paths`)\n if (links.length) {\n lines.push('For machine-readable documentation, see:', '', ...links, '')\n }\n return lines.join('\\n')\n}\n"]}
1
+ {"version":3,"sources":["../src/bots.ts","../src/markdown.ts"],"names":[],"mappings":";;;AAgBO,IAAM,cAAA,GACX,2QAAA;AAsBK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;;;ACZO,SAAS,sBAAsB,GAAA,EAAuC;AAC3E,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI;AACF,IAAA,QAAA,GAAW,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA,CAAE,QAAA;AAAA,EAC9B,CAAA,CAAA,MAAQ;AACN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AAEA,EAAA,MAAM,EAAA,GAAK,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,OAAA,CAAQ,EAAE,CAAA,EAAG;AACf,IAAA,OAAO,EAAE,MAAA,EAAQ,YAAA,EAAc,YAAA,EAAc,QAAA,EAAS;AAAA,EACxD;AAEA,EAAA,IAAI,QAAA,CAAS,QAAA,CAAS,KAAK,CAAA,EAAG;AAC5B,IAAA,OAAO,EAAE,QAAQ,WAAA,EAAa,YAAA,EAAc,SAAS,OAAA,CAAQ,OAAA,EAAS,EAAE,CAAA,EAAE;AAAA,EAC5E;AAEA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,MAAA,CAAO,QAAA,CAAS,eAAe,CAAA,EAAG;AACpC,IAAA,OAAO,EAAE,MAAA,EAAQ,eAAA,EAAiB,YAAA,EAAc,QAAA,EAAS;AAAA,EAC3D;AAEA,EAAA,OAAO,IAAA;AACT;AAqBO,SAAS,eAAA,CAAgB,KAAA,GAA8B,EAAC,EAA2B;AACxF,EAAA,MAAM,OAAA,GAAkC;AAAA,IACtC,cAAA,EAAgB,8BAAA;AAAA,IAChB,gBAAA,EAAkB,MAAM,aAAA,IAAiB,uCAAA;AAAA,IACzC,IAAA,EAAM;AAAA,GACR;AACA,EAAA,IAAI,OAAO,KAAA,CAAM,MAAA,KAAW,QAAA,IAAY,KAAA,CAAM,SAAS,CAAA,EAAG;AACxD,IAAA,OAAA,CAAQ,mBAAmB,CAAA,GAAI,IAAA,CAAK,GAAA,CAAI,CAAA,EAAG,IAAA,CAAK,IAAA,CAAK,KAAA,CAAM,MAAM,CAAC,CAAA,CAAE,QAAA,EAAS;AAAA,EAC/E;AACA,EAAA,OAAO,OAAA;AACT;AAoBO,SAAS,0BAA0B,KAAA,EAAuC;AAC/E,EAAA,MAAM,IAAA,GACJ,KAAA,CAAM,QAAA,IAAA,CACL,MAAM;AACL,IAAA,IAAI;AACF,MAAA,OAAO,IAAI,GAAA,CAAI,KAAA,CAAM,MAAM,CAAA,CAAE,QAAA;AAAA,IAC/B,CAAA,CAAA,MAAQ;AACN,MAAA,OAAO,KAAA,CAAM,MAAA;AAAA,IACf;AAAA,EACF,CAAA,GAAG;AACL,EAAA,MAAM,MAAM,CAAA,EAAG,KAAA,CAAM,MAAM,CAAA,EAAG,MAAM,QAAQ,CAAA,CAAA;AAC5C,EAAA,MAAM,KAAA,GAAkB,CAAC,CAAA,EAAA,EAAK,IAAI,IAAI,EAAA,EAAI,CAAA,WAAA,EAAc,GAAG,CAAA,gDAAA,CAAA,EAAoD,EAAE,CAAA;AACjH,EAAA,MAAM,QAAkB,EAAC;AACzB,EAAA,IAAI,KAAA,CAAM,UAAA,EAAY,KAAA,CAAM,IAAA,CAAK,CAAA,GAAA,EAAM,MAAM,UAAU,CAAA,EAAA,EAAK,KAAA,CAAM,UAAU,CAAA,8BAAA,CAA2B,CAAA;AACvG,EAAA,IAAI,KAAA,CAAM,cAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,cAAc,CAAA,EAAA,EAAK,KAAA,CAAM,cAAc,CAAA,8BAAA,CAA2B,CAAA;AAC3F,EAAA,IAAI,KAAA,CAAM,gBAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,gBAAgB,CAAA,EAAA,EAAK,KAAA,CAAM,gBAAgB,CAAA,yCAAA,CAAsC,CAAA;AAC1G,EAAA,IAAI,MAAM,MAAA,EAAQ;AAChB,IAAA,KAAA,CAAM,IAAA,CAAK,0CAAA,EAA4C,EAAA,EAAI,GAAG,OAAO,EAAE,CAAA;AAAA,EACzE;AACA,EAAA,OAAO,KAAA,CAAM,KAAK,IAAI,CAAA;AACxB","file":"markdown.cjs","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // SEO crawlers and monitoring bots.\n if (s.includes('ahrefsbot')) return 'Ahrefs'\n if (s.includes('semrushbot')) return 'Semrush'\n if (s.includes('mj12bot')) return 'Majestic'\n if (s.includes('dotbot')) return 'Moz'\n if (s.includes('rogerbot')) return 'Moz'\n if (s.includes('screaming frog')) return 'Screaming Frog'\n if (s.includes('sitebulb')) return 'Sitebulb'\n if (s.includes('linkfluence')) return 'Linkfluence'\n if (s.includes('dataforseo')) return 'DataForSEO'\n if (s.includes('serpstatbot')) return 'Serpstat'\n\n // Monitoring and feed bots.\n if (s.includes('uptimerobot')) return 'UptimeRobot'\n if (s.includes('pingdom')) return 'Pingdom'\n if (s.includes('statuscake')) return 'StatusCake'\n if (s.includes('newrelicpinger')) return 'New Relic'\n if (s.includes('datadogagent') || s.includes('datadog')) return 'Datadog'\n if (s.includes('slackbot')) return 'Slack'\n if (s.includes('twitterbot')) return 'Twitter'\n if (s.includes('linkedinbot')) return 'LinkedIn'\n if (s.includes('discordbot')) return 'Discord'\n if (s.includes('telegrambot')) return 'Telegram'\n if (s.includes('whatsapp')) return 'WhatsApp'\n\n // AI search and indexing bots.\n if (s.includes('linkupbot')) return 'Linkup'\n if (s.includes('sogou')) return 'Sogou'\n if (s.includes('yandexbot')) return 'Yandex'\n if (s.includes('baiduspider')) return 'Baidu'\n\n // Link preview fetchers.\n if (s.includes('facebookexternalhit')) return 'Facebook'\n if (s.includes('com.apple.webkit')) return 'Apple URL Preview'\n\n // Uptime and monitoring.\n if (s.includes('ohdear')) return 'Oh Dear'\n\n // Generic scrapers.\n if (s.includes('scrapy')) return 'Scrapy'\n if (s.includes('headlesschrome')) return 'Headless Chrome'\n if (s.includes('phantomjs')) return 'PhantomJS'\n if (s.includes('wget')) return 'wget'\n if (s.includes('httpie')) return 'HTTPie'\n if (s.includes('guzzlehttp')) return 'Guzzle'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\n/**\n * Detect likely headless/automated browsers by checking for missing headers\n * that real browsers always send. Playwright, Puppeteer, and similar tools\n * spoof the UA but often omit standard browser headers.\n *\n * Signals checked (each scores 1 point):\n * - Missing `Accept-Language` — every real browser sends this\n * - Missing `Sec-Fetch-Mode` — sent by all modern browsers\n * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+\n * - `Sec-CH-UA` contains \"HeadlessChrome\"\n * - Missing or bare Accept header — browsers send detailed accept lists\n * - `Connection: close` with browser UA — browsers use keep-alive\n *\n * Returns a score (0-6), the signals that fired, and a boolean `likely`\n * flag (score >= 2 with a browser-like UA).\n */\nexport function detectHeadless(req: Request): HeadlessDetection {\n const signals: string[] = []\n const ua = (req.headers.get('user-agent') || '').toLowerCase()\n const isBrowserUA =\n ua.includes('mozilla') || ua.includes('chrome') || ua.includes('safari') || ua.includes('firefox')\n\n if (!isBrowserUA) return { score: 0, signals: [], likely: false }\n\n if (!req.headers.get('accept-language')) {\n signals.push('missing-accept-language')\n }\n if (!req.headers.get('sec-fetch-mode')) {\n signals.push('missing-sec-fetch-mode')\n }\n const secChUa = req.headers.get('sec-ch-ua')\n if (!secChUa) {\n signals.push('missing-sec-ch-ua')\n } else if (secChUa.toLowerCase().includes('headlesschrome')) {\n signals.push('headless-chrome-hint')\n }\n const accept = req.headers.get('accept') || ''\n if (!accept || accept === '*/*') {\n signals.push('missing-or-bare-accept')\n }\n if ((req.headers.get('connection') || '').toLowerCase() === 'close') {\n signals.push('connection-close')\n }\n\n const score = signals.length\n return { score, signals, likely: score >= 2 }\n}\n\nexport interface HeadlessDetection {\n /** Number of suspicious signals found (0-6). */\n score: number\n /** Names of the specific signals that fired. */\n signals: string[]\n /** True when score >= 2 — strong headless indication. */\n likely: boolean\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'headless-likely'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the request:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'headless-likely'` — Browser-like UA but missing standard headers.\n * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).\n * - `'browser'` — Looks like a real browser with expected headers present.\n * - `'other'` — Unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n /** Headless browser detection result. Only populated when `req` is passed. */\n headless?: HeadlessDetection\n}\n\n/**\n * UA-only classification. Use {@link classifyRequest} for full detection\n * including headless browser heuristics.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n\n/**\n * Full request classification — combines UA parsing with header-based\n * headless detection. When a browser-like UA is missing standard headers,\n * the kind is promoted from `'browser'` to `'headless-likely'`.\n */\nexport function classifyRequest(req: Request): AgentClassification {\n const userAgent = req.headers.get('user-agent') || ''\n const base = classifyAgent(userAgent)\n const headless = detectHeadless(req)\n\n let kind = base.kind\n if (kind === 'browser' && headless.likely) {\n kind = 'headless-likely'\n }\n\n return { ...base, kind, headless }\n}\n","import { isAiBot } from './bots.js'\n\nexport type MarkdownServeReason =\n | 'ua-rewrite'\n | 'md-suffix'\n | 'accept-header'\n\nexport interface MarkdownDecision {\n /** Why this request should be served Markdown. */\n reason: MarkdownServeReason\n /**\n * The request's original logical path, with any trailing `.md` stripped.\n * Use this when mapping to a mirror file.\n */\n strippedPath: string\n}\n\n/**\n * Decide whether the request should be served Markdown instead of HTML.\n * Returns `null` when the request should go through your normal handler.\n *\n * Covers three triggers:\n * - Known AI-bot UA on any URL (`ua-rewrite`)\n * - Explicit `.md` suffix on the URL (`md-suffix`)\n * - `Accept: text/markdown` header (`accept-header`)\n *\n * This helper intentionally does not perform the rewrite itself — routing is\n * framework-specific (NextResponse.rewrite for Next.js, ctx.rewrite for\n * Hono, etc.). Use the returned decision to build the appropriate response.\n */\nexport function markdownServeDecision(req: Request): MarkdownDecision | null {\n let pathname = '/'\n try {\n pathname = new URL(req.url).pathname\n } catch {\n pathname = req.url || '/'\n }\n\n const ua = req.headers.get('user-agent') || ''\n if (isAiBot(ua)) {\n return { reason: 'ua-rewrite', strippedPath: pathname }\n }\n\n if (pathname.endsWith('.md')) {\n return { reason: 'md-suffix', strippedPath: pathname.replace(/\\.md$/, '') }\n }\n\n const accept = req.headers.get('accept') || ''\n if (accept.includes('text/markdown')) {\n return { reason: 'accept-header', strippedPath: pathname }\n }\n\n return null\n}\n\nexport interface MarkdownHeadersInput {\n /**\n * If provided, rendered as `x-markdown-tokens` so agents can budget context\n * before parsing the body. Typically `Math.ceil(body.length / 4)`.\n */\n tokens?: number\n /**\n * Content-Signal directive (see contentsignals.org). Defaults to\n * `'search=yes, ai-input=yes, ai-train=no'` — change if you want to permit\n * training or restrict indexing.\n */\n contentSignal?: string\n}\n\n/**\n * Build the set of response headers to attach to a Markdown response. Safe\n * defaults: UTF-8 text/markdown, Vary: accept, and a Content-Signal directive\n * that permits search + agent input but denies training.\n */\nexport function markdownHeaders(input: MarkdownHeadersInput = {}): Record<string, string> {\n const headers: Record<string, string> = {\n 'Content-Type': 'text/markdown; charset=utf-8',\n 'Content-Signal': input.contentSignal ?? 'search=yes, ai-input=yes, ai-train=no',\n Vary: 'accept'\n }\n if (typeof input.tokens === 'number' && input.tokens > 0) {\n headers['x-markdown-tokens'] = Math.max(1, Math.ceil(input.tokens)).toString()\n }\n return headers\n}\n\nexport interface SynthesizePointerInput {\n origin: string\n pathname: string\n /** URL of the site's curated index, usually `/llms.txt`. */\n llmsTxtUrl?: string\n /** URL of the full enumerated index, usually `/llms-full.txt`. */\n llmsFullTxtUrl?: string\n /** URL of the machine-readable path manifest, usually `/md/index.json`. */\n markdownIndexUrl?: string\n /** Site name to title the pointer document. Defaults to the origin hostname. */\n siteName?: string\n}\n\n/**\n * Generate a minimal pointer Markdown document for URLs that don't have a\n * pre-built mirror. Keeps the `Accept: text/markdown` contract intact\n * site-wide — agents always get *something* parseable, not a 404.\n */\nexport function synthesizeMarkdownPointer(input: SynthesizePointerInput): string {\n const site =\n input.siteName ??\n (() => {\n try {\n return new URL(input.origin).hostname\n } catch {\n return input.origin\n }\n })()\n const url = `${input.origin}${input.pathname}`\n const lines: string[] = [`# ${site}`, '', `This page (${url}) does not have a dedicated Markdown mirror yet.`, '']\n const links: string[] = []\n if (input.llmsTxtUrl) links.push(`- [${input.llmsTxtUrl}](${input.llmsTxtUrl}) — curated index of docs`)\n if (input.llmsFullTxtUrl)\n links.push(`- [${input.llmsFullTxtUrl}](${input.llmsFullTxtUrl}) — full enumerated index`)\n if (input.markdownIndexUrl)\n links.push(`- [${input.markdownIndexUrl}](${input.markdownIndexUrl}) — JSON index of all Markdown paths`)\n if (links.length) {\n lines.push('For machine-readable documentation, see:', '', ...links, '')\n }\n return lines.join('\\n')\n}\n"]}
package/dist/markdown.js.map CHANGED
@@ -1 +1 @@
1
- {"version":3,"sources":["../src/bots.ts","../src/markdown.ts"],"names":[],"mappings":";AAgBO,IAAM,cAAA,GACX,2QAAA;AAsBK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;;;ACZO,SAAS,sBAAsB,GAAA,EAAuC;AAC3E,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI;AACF,IAAA,QAAA,GAAW,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA,CAAE,QAAA;AAAA,EAC9B,CAAA,CAAA,MAAQ;AACN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AAEA,EAAA,MAAM,EAAA,GAAK,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,OAAA,CAAQ,EAAE,CAAA,EAAG;AACf,IAAA,OAAO,EAAE,MAAA,EAAQ,YAAA,EAAc,YAAA,EAAc,QAAA,EAAS;AAAA,EACxD;AAEA,EAAA,IAAI,QAAA,CAAS,QAAA,CAAS,KAAK,CAAA,EAAG;AAC5B,IAAA,OAAO,EAAE,QAAQ,WAAA,EAAa,YAAA,EAAc,SAAS,OAAA,CAAQ,OAAA,EAAS,EAAE,CAAA,EAAE;AAAA,EAC5E;AAEA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,MAAA,CAAO,QAAA,CAAS,eAAe,CAAA,EAAG;AACpC,IAAA,OAAO,EAAE,MAAA,EAAQ,eAAA,EAAiB,YAAA,EAAc,QAAA,EAAS;AAAA,EAC3D;AAEA,EAAA,OAAO,IAAA;AACT;AAqBO,SAAS,eAAA,CAAgB,KAAA,GAA8B,EAAC,EAA2B;AACxF,EAAA,MAAM,OAAA,GAAkC;AAAA,IACtC,cAAA,EAAgB,8BAAA;AAAA,IAChB,gBAAA,EAAkB,MAAM,aAAA,IAAiB,uCAAA;AAAA,IACzC,IAAA,EAAM;AAAA,GACR;AACA,EAAA,IAAI,OAAO,KAAA,CAAM,MAAA,KAAW,QAAA,IAAY,KAAA,CAAM,SAAS,CAAA,EAAG;AACxD,IAAA,OAAA,CAAQ,mBAAmB,CAAA,GAAI,IAAA,CAAK,GAAA,CAAI,CAAA,EAAG,IAAA,CAAK,IAAA,CAAK,KAAA,CAAM,MAAM,CAAC,CAAA,CAAE,QAAA,EAAS;AAAA,EAC/E;AACA,EAAA,OAAO,OAAA;AACT;AAoBO,SAAS,0BAA0B,KAAA,EAAuC;AAC/E,EAAA,MAAM,IAAA,GACJ,KAAA,CAAM,QAAA,IAAA,CACL,MAAM;AACL,IAAA,IAAI;AACF,MAAA,OAAO,IAAI,GAAA,CAAI,KAAA,CAAM,MAAM,CAAA,CAAE,QAAA;AAAA,IAC/B,CAAA,CAAA,MAAQ;AACN,MAAA,OAAO,KAAA,CAAM,MAAA;AAAA,IACf;AAAA,EACF,CAAA,GAAG;AACL,EAAA,MAAM,MAAM,CAAA,EAAG,KAAA,CAAM,MAAM,CAAA,EAAG,MAAM,QAAQ,CAAA,CAAA;AAC5C,EAAA,MAAM,KAAA,GAAkB,CAAC,CAAA,EAAA,EAAK,IAAI,IAAI,EAAA,EAAI,CAAA,WAAA,EAAc,GAAG,CAAA,gDAAA,CAAA,EAAoD,EAAE,CAAA;AACjH,EAAA,MAAM,QAAkB,EAAC;AACzB,EAAA,IAAI,KAAA,CAAM,UAAA,EAAY,KAAA,CAAM,IAAA,CAAK,CAAA,GAAA,EAAM,MAAM,UAAU,CAAA,EAAA,EAAK,KAAA,CAAM,UAAU,CAAA,8BAAA,CAA2B,CAAA;AACvG,EAAA,IAAI,KAAA,CAAM,cAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,cAAc,CAAA,EAAA,EAAK,KAAA,CAAM,cAAc,CAAA,8BAAA,CAA2B,CAAA;AAC3F,EAAA,IAAI,KAAA,CAAM,gBAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,gBAAgB,CAAA,EAAA,EAAK,KAAA,CAAM,gBAAgB,CAAA,yCAAA,CAAsC,CAAA;AAC1G,EAAA,IAAI,MAAM,MAAA,EAAQ;AAChB,IAAA,KAAA,CAAM,IAAA,CAAK,0CAAA,EAA4C,EAAA,EAAI,GAAG,OAAO,EAAE,CAAA;AAAA,EACzE;AACA,EAAA,OAAO,KAAA,CAAM,KAAK,IAAI,CAAA;AACxB","file":"markdown.js","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the UA:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'browser'` — looks like a real browser. Could be a genuine user or\n * a Playwright-based agent (Aider, OpenCode) that can't be distinguished\n * at the UA layer.\n * - `'other'` — unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n}\n\n/**\n * One-stop classification of a user-agent. Combines {@link isAiBot},\n * {@link isHttpClient}, and {@link parseBotName} into a single structured\n * result. Used internally by `trackVisit` to populate event properties;\n * useful in consumer code when you need all signals at once.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n","import { isAiBot } from './bots.js'\n\nexport type MarkdownServeReason =\n | 'ua-rewrite'\n | 'md-suffix'\n | 'accept-header'\n\nexport interface MarkdownDecision {\n /** Why this request should be served Markdown. */\n reason: MarkdownServeReason\n /**\n * The request's original logical path, with any trailing `.md` stripped.\n * Use this when mapping to a mirror file.\n */\n strippedPath: string\n}\n\n/**\n * Decide whether the request should be served Markdown instead of HTML.\n * Returns `null` when the request should go through your normal handler.\n *\n * Covers three triggers:\n * - Known AI-bot UA on any URL (`ua-rewrite`)\n * - Explicit `.md` suffix on the URL (`md-suffix`)\n * - `Accept: text/markdown` header (`accept-header`)\n *\n * This helper intentionally does not perform the rewrite itself — routing is\n * framework-specific (NextResponse.rewrite for Next.js, ctx.rewrite for\n * Hono, etc.). Use the returned decision to build the appropriate response.\n */\nexport function markdownServeDecision(req: Request): MarkdownDecision | null {\n let pathname = '/'\n try {\n pathname = new URL(req.url).pathname\n } catch {\n pathname = req.url || '/'\n }\n\n const ua = req.headers.get('user-agent') || ''\n if (isAiBot(ua)) {\n return { reason: 'ua-rewrite', strippedPath: pathname }\n }\n\n if (pathname.endsWith('.md')) {\n return { reason: 'md-suffix', strippedPath: pathname.replace(/\\.md$/, '') }\n }\n\n const accept = req.headers.get('accept') || ''\n if (accept.includes('text/markdown')) {\n return { reason: 'accept-header', strippedPath: pathname }\n }\n\n return null\n}\n\nexport interface MarkdownHeadersInput {\n /**\n * If provided, rendered as `x-markdown-tokens` so agents can budget context\n * before parsing the body. Typically `Math.ceil(body.length / 4)`.\n */\n tokens?: number\n /**\n * Content-Signal directive (see contentsignals.org). Defaults to\n * `'search=yes, ai-input=yes, ai-train=no'` — change if you want to permit\n * training or restrict indexing.\n */\n contentSignal?: string\n}\n\n/**\n * Build the set of response headers to attach to a Markdown response. Safe\n * defaults: UTF-8 text/markdown, Vary: accept, and a Content-Signal directive\n * that permits search + agent input but denies training.\n */\nexport function markdownHeaders(input: MarkdownHeadersInput = {}): Record<string, string> {\n const headers: Record<string, string> = {\n 'Content-Type': 'text/markdown; charset=utf-8',\n 'Content-Signal': input.contentSignal ?? 'search=yes, ai-input=yes, ai-train=no',\n Vary: 'accept'\n }\n if (typeof input.tokens === 'number' && input.tokens > 0) {\n headers['x-markdown-tokens'] = Math.max(1, Math.ceil(input.tokens)).toString()\n }\n return headers\n}\n\nexport interface SynthesizePointerInput {\n origin: string\n pathname: string\n /** URL of the site's curated index, usually `/llms.txt`. */\n llmsTxtUrl?: string\n /** URL of the full enumerated index, usually `/llms-full.txt`. */\n llmsFullTxtUrl?: string\n /** URL of the machine-readable path manifest, usually `/md/index.json`. */\n markdownIndexUrl?: string\n /** Site name to title the pointer document. Defaults to the origin hostname. */\n siteName?: string\n}\n\n/**\n * Generate a minimal pointer Markdown document for URLs that don't have a\n * pre-built mirror. Keeps the `Accept: text/markdown` contract intact\n * site-wide — agents always get *something* parseable, not a 404.\n */\nexport function synthesizeMarkdownPointer(input: SynthesizePointerInput): string {\n const site =\n input.siteName ??\n (() => {\n try {\n return new URL(input.origin).hostname\n } catch {\n return input.origin\n }\n })()\n const url = `${input.origin}${input.pathname}`\n const lines: string[] = [`# ${site}`, '', `This page (${url}) does not have a dedicated Markdown mirror yet.`, '']\n const links: string[] = []\n if (input.llmsTxtUrl) links.push(`- [${input.llmsTxtUrl}](${input.llmsTxtUrl}) — curated index of docs`)\n if (input.llmsFullTxtUrl)\n links.push(`- [${input.llmsFullTxtUrl}](${input.llmsFullTxtUrl}) — full enumerated index`)\n if (input.markdownIndexUrl)\n links.push(`- [${input.markdownIndexUrl}](${input.markdownIndexUrl}) — JSON index of all Markdown paths`)\n if (links.length) {\n lines.push('For machine-readable documentation, see:', '', ...links, '')\n }\n return lines.join('\\n')\n}\n"]}
1
+ {"version":3,"sources":["../src/bots.ts","../src/markdown.ts"],"names":[],"mappings":";AAgBO,IAAM,cAAA,GACX,2QAAA;AAsBK,SAAS,QAAQ,SAAA,EAA+C;AACrE,EAAA,IAAI,CAAC,WAAW,OAAO,KAAA;AACvB,EAAA,OAAO,cAAA,CAAe,KAAK,SAAS,CAAA;AACtC;;;ACZO,SAAS,sBAAsB,GAAA,EAAuC;AAC3E,EAAA,IAAI,QAAA,GAAW,GAAA;AACf,EAAA,IAAI;AACF,IAAA,QAAA,GAAW,IAAI,GAAA,CAAI,GAAA,CAAI,GAAG,CAAA,CAAE,QAAA;AAAA,EAC9B,CAAA,CAAA,MAAQ;AACN,IAAA,QAAA,GAAW,IAAI,GAAA,IAAO,GAAA;AAAA,EACxB;AAEA,EAAA,MAAM,EAAA,GAAK,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,YAAY,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,OAAA,CAAQ,EAAE,CAAA,EAAG;AACf,IAAA,OAAO,EAAE,MAAA,EAAQ,YAAA,EAAc,YAAA,EAAc,QAAA,EAAS;AAAA,EACxD;AAEA,EAAA,IAAI,QAAA,CAAS,QAAA,CAAS,KAAK,CAAA,EAAG;AAC5B,IAAA,OAAO,EAAE,QAAQ,WAAA,EAAa,YAAA,EAAc,SAAS,OAAA,CAAQ,OAAA,EAAS,EAAE,CAAA,EAAE;AAAA,EAC5E;AAEA,EAAA,MAAM,MAAA,GAAS,GAAA,CAAI,OAAA,CAAQ,GAAA,CAAI,QAAQ,CAAA,IAAK,EAAA;AAC5C,EAAA,IAAI,MAAA,CAAO,QAAA,CAAS,eAAe,CAAA,EAAG;AACpC,IAAA,OAAO,EAAE,MAAA,EAAQ,eAAA,EAAiB,YAAA,EAAc,QAAA,EAAS;AAAA,EAC3D;AAEA,EAAA,OAAO,IAAA;AACT;AAqBO,SAAS,eAAA,CAAgB,KAAA,GAA8B,EAAC,EAA2B;AACxF,EAAA,MAAM,OAAA,GAAkC;AAAA,IACtC,cAAA,EAAgB,8BAAA;AAAA,IAChB,gBAAA,EAAkB,MAAM,aAAA,IAAiB,uCAAA;AAAA,IACzC,IAAA,EAAM;AAAA,GACR;AACA,EAAA,IAAI,OAAO,KAAA,CAAM,MAAA,KAAW,QAAA,IAAY,KAAA,CAAM,SAAS,CAAA,EAAG;AACxD,IAAA,OAAA,CAAQ,mBAAmB,CAAA,GAAI,IAAA,CAAK,GAAA,CAAI,CAAA,EAAG,IAAA,CAAK,IAAA,CAAK,KAAA,CAAM,MAAM,CAAC,CAAA,CAAE,QAAA,EAAS;AAAA,EAC/E;AACA,EAAA,OAAO,OAAA;AACT;AAoBO,SAAS,0BAA0B,KAAA,EAAuC;AAC/E,EAAA,MAAM,IAAA,GACJ,KAAA,CAAM,QAAA,IAAA,CACL,MAAM;AACL,IAAA,IAAI;AACF,MAAA,OAAO,IAAI,GAAA,CAAI,KAAA,CAAM,MAAM,CAAA,CAAE,QAAA;AAAA,IAC/B,CAAA,CAAA,MAAQ;AACN,MAAA,OAAO,KAAA,CAAM,MAAA;AAAA,IACf;AAAA,EACF,CAAA,GAAG;AACL,EAAA,MAAM,MAAM,CAAA,EAAG,KAAA,CAAM,MAAM,CAAA,EAAG,MAAM,QAAQ,CAAA,CAAA;AAC5C,EAAA,MAAM,KAAA,GAAkB,CAAC,CAAA,EAAA,EAAK,IAAI,IAAI,EAAA,EAAI,CAAA,WAAA,EAAc,GAAG,CAAA,gDAAA,CAAA,EAAoD,EAAE,CAAA;AACjH,EAAA,MAAM,QAAkB,EAAC;AACzB,EAAA,IAAI,KAAA,CAAM,UAAA,EAAY,KAAA,CAAM,IAAA,CAAK,CAAA,GAAA,EAAM,MAAM,UAAU,CAAA,EAAA,EAAK,KAAA,CAAM,UAAU,CAAA,8BAAA,CAA2B,CAAA;AACvG,EAAA,IAAI,KAAA,CAAM,cAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,cAAc,CAAA,EAAA,EAAK,KAAA,CAAM,cAAc,CAAA,8BAAA,CAA2B,CAAA;AAC3F,EAAA,IAAI,KAAA,CAAM,gBAAA;AACR,IAAA,KAAA,CAAM,KAAK,CAAA,GAAA,EAAM,KAAA,CAAM,gBAAgB,CAAA,EAAA,EAAK,KAAA,CAAM,gBAAgB,CAAA,yCAAA,CAAsC,CAAA;AAC1G,EAAA,IAAI,MAAM,MAAA,EAAQ;AAChB,IAAA,KAAA,CAAM,IAAA,CAAK,0CAAA,EAA4C,EAAA,EAAI,GAAG,OAAO,EAAE,CAAA;AAAA,EACzE;AACA,EAAA,OAAO,KAAA,CAAM,KAAK,IAAI,CAAA;AACxB","file":"markdown.js","sourcesContent":["/**\n * User-agent substrings that identify **publicly declared** AI crawlers — the\n * branded bots that identify themselves by name (OpenAI's GPTBot, Anthropic's\n * ClaudeBot, Perplexity-User, Google-Extended, etc.). High-confidence: when\n * this matches, the request almost certainly comes from that vendor's crawler\n * fleet.\n *\n * Does NOT include **coding-agent traffic** (Claude Code, Cline, Cursor,\n * Windsurf, Aider, OpenCode, VS Code). Those tools use generic HTTP library\n * UAs (axios, curl, got, colly, Electron) or spoof full browser UAs — they\n * can't be distinguished from non-AI traffic by UA alone. See\n * {@link HTTP_CLIENT_PATTERN} for the loose heuristic layer.\n *\n * Sources consulted when updating: darkvisitors.com, vendor docs from OpenAI,\n * Anthropic, Google, Perplexity, Cohere, Apple, Bytedance.\n */\nexport const AI_BOT_PATTERN =\n /ClaudeBot|Claude-User|Anthropic|ChatGPT-User|GPTBot|OAI-SearchBot|PerplexityBot|Perplexity-User|Google-Extended|Applebot-Extended|cohere-ai|Bytespider|CCBot|Amazonbot|Meta-ExternalAgent|FacebookBot|DuckAssistBot|MistralAI-User|YouBot|AI2Bot|Diffbot|Cursor|Windsurf/i\n\n/**\n * HTTP library / runtime signatures frequently used by coding agents. Matching\n * any of these is a **loose** signal — legitimate curl scripts, CI jobs, and\n * server-to-server traffic use the same libraries. Use this for the wider\n * net (`coding_agent_hint: true`) and pair with other signals (request\n * shape, JA4 fingerprint, path patterns) for higher confidence.\n *\n * Based on behavioural signatures observed by Addy Osmani:\n * Claude Code → axios/1.8.4\n * Cline, Junie → curl/8.4.0\n * Cursor → got (sindresorhus/got)\n * Windsurf → colly\n * VS Code → Electron / Chromium\n *\n * Aider and OpenCode use Playwright-driven full Mozilla/Safari UAs and are\n * indistinguishable from real browsers at the UA layer.\n */\nexport const HTTP_CLIENT_PATTERN =\n /axios\\/|curl\\/|(?:^|[\\s(])got(?:\\/|[\\s(])|\\bcolly\\b|Electron\\/|node-fetch\\/|python-requests\\/|Go-http-client\\/|okhttp\\/|aiohttp\\/|Deno\\//i\n\nexport function isAiBot(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return AI_BOT_PATTERN.test(userAgent)\n}\n\nexport function isHttpClient(userAgent: string | null | undefined): boolean {\n if (!userAgent) return false\n return HTTP_CLIENT_PATTERN.test(userAgent)\n}\n\n/**\n * Map a user-agent string to a coarse, human-readable label. Returns one of:\n *\n * - A branded-crawler name (`'Claude'`, `'ChatGPT'`, …) — pair with\n * {@link isAiBot} for `is_ai_bot: true` segmentation.\n * - An HTTP-library name (`'curl'`, `'axios'`, `'got'`, `'colly'`,\n * `'Electron'`, …) — hint of a coding agent or automation; not\n * conclusive. Pair with {@link isHttpClient}.\n * - `'Browser'` for typical desktop browsers (possibly spoofed by\n * Playwright-based agents like Aider/OpenCode — this label alone can't\n * tell you).\n * - `'Other'` for anything unrecognised or empty input.\n */\nexport function parseBotName(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const s = userAgent.toLowerCase()\n\n // Publicly declared AI crawlers (high confidence).\n if (s.includes('chatgpt-user') || s.includes('gptbot') || s.includes('oai-searchbot') || s.includes('openai'))\n return 'ChatGPT'\n if (s.includes('claudebot') || s.includes('claude-user') || s.includes('anthropic')) return 'Claude'\n if (s.includes('perplexitybot') || s.includes('perplexity-user')) return 'Perplexity'\n if (s.includes('ccbot')) return 'Common Crawl'\n if (s.includes('google-extended') || s.includes('googlebot')) return 'Google'\n if (s.includes('applebot-extended') || s.includes('applebot')) return 'Apple'\n if (s.includes('bingbot')) return 'Bing'\n if (s.includes('bytespider')) return 'Bytespider'\n if (s.includes('amazonbot')) return 'Amazon'\n if (s.includes('meta-externalagent') || s.includes('facebookbot')) return 'Meta'\n if (s.includes('mistralai-user')) return 'Mistral'\n if (s.includes('duckassistbot')) return 'DuckDuckGo'\n if (s.includes('youbot')) return 'You.com'\n if (s.includes('diffbot')) return 'Diffbot'\n if (s.includes('ai2bot')) return 'AI2'\n if (s.includes('cohere')) return 'Cohere'\n if (s.includes('cursor')) return 'Cursor'\n if (s.includes('windsurf')) return 'Windsurf'\n if (s.includes('petalbot')) return 'PetalBot'\n\n // SEO crawlers and monitoring bots.\n if (s.includes('ahrefsbot')) return 'Ahrefs'\n if (s.includes('semrushbot')) return 'Semrush'\n if (s.includes('mj12bot')) return 'Majestic'\n if (s.includes('dotbot')) return 'Moz'\n if (s.includes('rogerbot')) return 'Moz'\n if (s.includes('screaming frog')) return 'Screaming Frog'\n if (s.includes('sitebulb')) return 'Sitebulb'\n if (s.includes('linkfluence')) return 'Linkfluence'\n if (s.includes('dataforseo')) return 'DataForSEO'\n if (s.includes('serpstatbot')) return 'Serpstat'\n\n // Monitoring and feed bots.\n if (s.includes('uptimerobot')) return 'UptimeRobot'\n if (s.includes('pingdom')) return 'Pingdom'\n if (s.includes('statuscake')) return 'StatusCake'\n if (s.includes('newrelicpinger')) return 'New Relic'\n if (s.includes('datadogagent') || s.includes('datadog')) return 'Datadog'\n if (s.includes('slackbot')) return 'Slack'\n if (s.includes('twitterbot')) return 'Twitter'\n if (s.includes('linkedinbot')) return 'LinkedIn'\n if (s.includes('discordbot')) return 'Discord'\n if (s.includes('telegrambot')) return 'Telegram'\n if (s.includes('whatsapp')) return 'WhatsApp'\n\n // AI search and indexing bots.\n if (s.includes('linkupbot')) return 'Linkup'\n if (s.includes('sogou')) return 'Sogou'\n if (s.includes('yandexbot')) return 'Yandex'\n if (s.includes('baiduspider')) return 'Baidu'\n\n // Link preview fetchers.\n if (s.includes('facebookexternalhit')) return 'Facebook'\n if (s.includes('com.apple.webkit')) return 'Apple URL Preview'\n\n // Uptime and monitoring.\n if (s.includes('ohdear')) return 'Oh Dear'\n\n // Generic scrapers.\n if (s.includes('scrapy')) return 'Scrapy'\n if (s.includes('headlesschrome')) return 'Headless Chrome'\n if (s.includes('phantomjs')) return 'PhantomJS'\n if (s.includes('wget')) return 'wget'\n if (s.includes('httpie')) return 'HTTPie'\n if (s.includes('guzzlehttp')) return 'Guzzle'\n\n // HTTP library / runtime signatures (loose — coding agent or automation).\n // Check Electron before Browser since Electron UAs contain Chrome/Safari.\n if (s.includes('electron/')) return 'Electron'\n if (/curl\\//.test(s)) return 'curl'\n if (/axios\\//.test(s)) return 'axios'\n if (/(?:^|[\\s(])got(?:\\/|[\\s(])/.test(s)) return 'got'\n if (/\\bcolly\\b/.test(s)) return 'colly'\n if (/node-fetch\\//.test(s)) return 'node-fetch'\n if (/python-requests\\//.test(s)) return 'python-requests'\n if (/go-http-client\\//.test(s)) return 'Go http client'\n if (/okhttp\\//.test(s)) return 'OkHttp'\n if (/aiohttp\\//.test(s)) return 'aiohttp'\n if (/deno\\//.test(s)) return 'Deno'\n\n // Real browsers (or UAs spoofed to look like them — see Aider/OpenCode note).\n if (s.includes('mozilla') || s.includes('chrome') || s.includes('safari') || s.includes('firefox'))\n return 'Browser'\n\n return 'Other'\n}\n\n/**\n * Return the first product token from a UA header, useful for segmenting by\n * client without hard-coding every bot name. Falls back to `'Other'` for empty\n * input.\n */\nexport function firstUserAgentProduct(userAgent: string | null | undefined): string {\n if (!userAgent || typeof userAgent !== 'string') return 'Other'\n const compatibleMatch = userAgent.match(/compatible;\\s*([^/;\\s]+)(?:\\/[^\\s;]*)?/i)\n if (compatibleMatch && compatibleMatch[1]) return compatibleMatch[1].trim()\n const first = userAgent.trim().split('/')[0]?.trim().split(/\\s+/)[0]?.trim()\n return first || 'Other'\n}\n\n/**\n * Detect likely headless/automated browsers by checking for missing headers\n * that real browsers always send. Playwright, Puppeteer, and similar tools\n * spoof the UA but often omit standard browser headers.\n *\n * Signals checked (each scores 1 point):\n * - Missing `Accept-Language` — every real browser sends this\n * - Missing `Sec-Fetch-Mode` — sent by all modern browsers\n * - Missing `Sec-CH-UA` — Client Hints, Chromium 89+\n * - `Sec-CH-UA` contains \"HeadlessChrome\"\n * - Missing or bare Accept header — browsers send detailed accept lists\n * - `Connection: close` with browser UA — browsers use keep-alive\n *\n * Returns a score (0-6), the signals that fired, and a boolean `likely`\n * flag (score >= 2 with a browser-like UA).\n */\nexport function detectHeadless(req: Request): HeadlessDetection {\n const signals: string[] = []\n const ua = (req.headers.get('user-agent') || '').toLowerCase()\n const isBrowserUA =\n ua.includes('mozilla') || ua.includes('chrome') || ua.includes('safari') || ua.includes('firefox')\n\n if (!isBrowserUA) return { score: 0, signals: [], likely: false }\n\n if (!req.headers.get('accept-language')) {\n signals.push('missing-accept-language')\n }\n if (!req.headers.get('sec-fetch-mode')) {\n signals.push('missing-sec-fetch-mode')\n }\n const secChUa = req.headers.get('sec-ch-ua')\n if (!secChUa) {\n signals.push('missing-sec-ch-ua')\n } else if (secChUa.toLowerCase().includes('headlesschrome')) {\n signals.push('headless-chrome-hint')\n }\n const accept = req.headers.get('accept') || ''\n if (!accept || accept === '*/*') {\n signals.push('missing-or-bare-accept')\n }\n if ((req.headers.get('connection') || '').toLowerCase() === 'close') {\n signals.push('connection-close')\n }\n\n const score = signals.length\n return { score, signals, likely: score >= 2 }\n}\n\nexport interface HeadlessDetection {\n /** Number of suspicious signals found (0-6). */\n score: number\n /** Names of the specific signals that fired. */\n signals: string[]\n /** True when score >= 2 — strong headless indication. */\n likely: boolean\n}\n\nexport type AgentKind =\n | 'declared-crawler'\n | 'coding-agent-hint'\n | 'headless-likely'\n | 'browser'\n | 'other'\n\nexport interface AgentClassification {\n /**\n * Categorical tag for the request:\n *\n * - `'declared-crawler'` — {@link AI_BOT_PATTERN} matched. High confidence.\n * - `'coding-agent-hint'` — {@link HTTP_CLIENT_PATTERN} matched. Loose\n * signal; could be a coding agent, a curl script, or any automation.\n * - `'headless-likely'` — Browser-like UA but missing standard headers.\n * Strong signal of Playwright/Puppeteer automation (Aider, OpenCode, etc.).\n * - `'browser'` — Looks like a real browser with expected headers present.\n * - `'other'` — Unrecognised or empty.\n */\n kind: AgentKind\n /** Human-readable label, same string {@link parseBotName} returns. */\n label: string\n /** Strict: `true` only when the UA matches a branded AI crawler. */\n isAiBot: boolean\n /** Loose: `true` for known HTTP-library / automation UAs. */\n codingAgentHint: boolean\n /** Headless browser detection result. Only populated when `req` is passed. */\n headless?: HeadlessDetection\n}\n\n/**\n * UA-only classification. Use {@link classifyRequest} for full detection\n * including headless browser heuristics.\n */\nexport function classifyAgent(userAgent: string | null | undefined): AgentClassification {\n const label = parseBotName(userAgent)\n const aiBot = isAiBot(userAgent)\n const httpClient = isHttpClient(userAgent)\n\n let kind: AgentKind\n if (aiBot) kind = 'declared-crawler'\n else if (httpClient) kind = 'coding-agent-hint'\n else if (label === 'Browser') kind = 'browser'\n else kind = 'other'\n\n return { kind, label, isAiBot: aiBot, codingAgentHint: httpClient }\n}\n\n/**\n * Full request classification — combines UA parsing with header-based\n * headless detection. When a browser-like UA is missing standard headers,\n * the kind is promoted from `'browser'` to `'headless-likely'`.\n */\nexport function classifyRequest(req: Request): AgentClassification {\n const userAgent = req.headers.get('user-agent') || ''\n const base = classifyAgent(userAgent)\n const headless = detectHeadless(req)\n\n let kind = base.kind\n if (kind === 'browser' && headless.likely) {\n kind = 'headless-likely'\n }\n\n return { ...base, kind, headless }\n}\n","import { isAiBot } from './bots.js'\n\nexport type MarkdownServeReason =\n | 'ua-rewrite'\n | 'md-suffix'\n | 'accept-header'\n\nexport interface MarkdownDecision {\n /** Why this request should be served Markdown. */\n reason: MarkdownServeReason\n /**\n * The request's original logical path, with any trailing `.md` stripped.\n * Use this when mapping to a mirror file.\n */\n strippedPath: string\n}\n\n/**\n * Decide whether the request should be served Markdown instead of HTML.\n * Returns `null` when the request should go through your normal handler.\n *\n * Covers three triggers:\n * - Known AI-bot UA on any URL (`ua-rewrite`)\n * - Explicit `.md` suffix on the URL (`md-suffix`)\n * - `Accept: text/markdown` header (`accept-header`)\n *\n * This helper intentionally does not perform the rewrite itself — routing is\n * framework-specific (NextResponse.rewrite for Next.js, ctx.rewrite for\n * Hono, etc.). Use the returned decision to build the appropriate response.\n */\nexport function markdownServeDecision(req: Request): MarkdownDecision | null {\n let pathname = '/'\n try {\n pathname = new URL(req.url).pathname\n } catch {\n pathname = req.url || '/'\n }\n\n const ua = req.headers.get('user-agent') || ''\n if (isAiBot(ua)) {\n return { reason: 'ua-rewrite', strippedPath: pathname }\n }\n\n if (pathname.endsWith('.md')) {\n return { reason: 'md-suffix', strippedPath: pathname.replace(/\\.md$/, '') }\n }\n\n const accept = req.headers.get('accept') || ''\n if (accept.includes('text/markdown')) {\n return { reason: 'accept-header', strippedPath: pathname }\n }\n\n return null\n}\n\nexport interface MarkdownHeadersInput {\n /**\n * If provided, rendered as `x-markdown-tokens` so agents can budget context\n * before parsing the body. Typically `Math.ceil(body.length / 4)`.\n */\n tokens?: number\n /**\n * Content-Signal directive (see contentsignals.org). Defaults to\n * `'search=yes, ai-input=yes, ai-train=no'` — change if you want to permit\n * training or restrict indexing.\n */\n contentSignal?: string\n}\n\n/**\n * Build the set of response headers to attach to a Markdown response. Safe\n * defaults: UTF-8 text/markdown, Vary: accept, and a Content-Signal directive\n * that permits search + agent input but denies training.\n */\nexport function markdownHeaders(input: MarkdownHeadersInput = {}): Record<string, string> {\n const headers: Record<string, string> = {\n 'Content-Type': 'text/markdown; charset=utf-8',\n 'Content-Signal': input.contentSignal ?? 'search=yes, ai-input=yes, ai-train=no',\n Vary: 'accept'\n }\n if (typeof input.tokens === 'number' && input.tokens > 0) {\n headers['x-markdown-tokens'] = Math.max(1, Math.ceil(input.tokens)).toString()\n }\n return headers\n}\n\nexport interface SynthesizePointerInput {\n origin: string\n pathname: string\n /** URL of the site's curated index, usually `/llms.txt`. */\n llmsTxtUrl?: string\n /** URL of the full enumerated index, usually `/llms-full.txt`. */\n llmsFullTxtUrl?: string\n /** URL of the machine-readable path manifest, usually `/md/index.json`. */\n markdownIndexUrl?: string\n /** Site name to title the pointer document. Defaults to the origin hostname. */\n siteName?: string\n}\n\n/**\n * Generate a minimal pointer Markdown document for URLs that don't have a\n * pre-built mirror. Keeps the `Accept: text/markdown` contract intact\n * site-wide — agents always get *something* parseable, not a 404.\n */\nexport function synthesizeMarkdownPointer(input: SynthesizePointerInput): string {\n const site =\n input.siteName ??\n (() => {\n try {\n return new URL(input.origin).hostname\n } catch {\n return input.origin\n }\n })()\n const url = `${input.origin}${input.pathname}`\n const lines: string[] = [`# ${site}`, '', `This page (${url}) does not have a dedicated Markdown mirror yet.`, '']\n const links: string[] = []\n if (input.llmsTxtUrl) links.push(`- [${input.llmsTxtUrl}](${input.llmsTxtUrl}) — curated index of docs`)\n if (input.llmsFullTxtUrl)\n links.push(`- [${input.llmsFullTxtUrl}](${input.llmsFullTxtUrl}) — full enumerated index`)\n if (input.markdownIndexUrl)\n links.push(`- [${input.markdownIndexUrl}](${input.markdownIndexUrl}) — JSON index of all Markdown paths`)\n if (links.length) {\n lines.push('For machine-readable documentation, see:', '', ...links, '')\n }\n return lines.join('\\n')\n}\n"]}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@apideck/agent-analytics",
3
- "version": "0.5.0",
3
+ "version": "0.7.0",
4
4
  "description": "Track AI agent and bot traffic to your Next.js / Vercel app — PostHog, webhooks, or any custom analytics backend. Detects Claude, ChatGPT, Perplexity, Google-Extended, and more.",
5
5
  "keywords": [
6
6
  "ai",