@absolutejs/rag 0.22.12 → 0.22.14
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.
- package/CHANGELOG.md +16 -0
- package/changelog.json +36 -0
- package/dist/index.js +4 -1
- package/dist/index.js.map +3 -3
- package/dist/research/index.js +98 -14
- package/dist/research/index.js.map +6 -6
- package/dist/src/research/types.d.ts +16 -0
- package/dist/src/web/index.d.ts +1 -1
- package/dist/src/web/transport.d.ts +4 -0
- package/dist/web/index.js +8 -4
- package/dist/web/index.js.map +5 -5
- package/dist/web/playwright.js +4 -1
- package/dist/web/playwright.js.map +3 -3
- package/dist/web-index/index.js +98 -14
- package/dist/web-index/index.js.map +6 -6
- package/package.json +1 -1
package/dist/web/playwright.js
CHANGED
|
@@ -10,15 +10,18 @@ import {
|
|
|
10
10
|
import { lookup } from "dns/promises";
|
|
11
11
|
import { isIP } from "net";
|
|
12
12
|
import { brotliDecompressSync, gunzipSync, inflateSync } from "zlib";
|
|
13
|
+
var WEB_READ_ERROR = Symbol.for("@absolutejs/rag/WebReadError");
|
|
13
14
|
|
|
14
15
|
class WebReadError extends Error {
|
|
15
16
|
code;
|
|
17
|
+
[WEB_READ_ERROR] = true;
|
|
16
18
|
constructor(code, message) {
|
|
17
19
|
super(message);
|
|
18
20
|
this.code = code;
|
|
19
21
|
this.name = "WebReadError";
|
|
20
22
|
}
|
|
21
23
|
}
|
|
24
|
+
var isWebReadError = (error) => error instanceof Error && Reflect.get(error, WEB_READ_ERROR) === true && typeof Reflect.get(error, "code") === "string";
|
|
22
25
|
var isPublicWebAddress = (address) => isPublicNetworkAddress(address);
|
|
23
26
|
var validatePublicWebUrl = (raw) => {
|
|
24
27
|
let url;
|
|
@@ -248,5 +251,5 @@ export {
|
|
|
248
251
|
createPlaywrightWebRenderer
|
|
249
252
|
};
|
|
250
253
|
|
|
251
|
-
//# debugId=
|
|
254
|
+
//# debugId=D0DCB66F403B29EA64756E2164756E21
|
|
252
255
|
//# sourceMappingURL=playwright.js.map
|
|
@@ -3,9 +3,9 @@
|
|
|
3
3
|
"sources": ["../src/web/playwright.ts", "../src/web/transport.ts"],
|
|
4
4
|
"sourcesContent": [
|
|
5
5
|
"import { chromium, type Browser } from \"playwright-core\";\nimport {\n fetchPublicWebResource,\n validatePublicWebUrl,\n WebReadError,\n} from \"./transport\";\nimport type { WebRedirect } from \"./transport\";\nimport type { WebRenderer } from \"./index\";\n/** Fresh contexts share a browser, never cookies; all HTTP traffic uses pinned public DNS. */\nexport const createPlaywrightWebRenderer = (\n options: { executablePath?: string } = {},\n) => {\n let browser: Browser | undefined;\n let launching: Promise<Browser> | undefined;\n const launch = async () => {\n if (browser?.isConnected()) return browser;\n launching ??= chromium.launch({\n headless: true,\n executablePath: options.executablePath,\n args: [\"--disable-dev-shm-usage\"],\n });\n try {\n browser = await launching;\n return browser;\n } finally {\n launching = undefined;\n }\n };\n const renderOnce: WebRenderer = async (url, { signal }) => {\n validatePublicWebUrl(url);\n signal.throwIfAborted();\n const instance = await launch();\n const context = await instance.newContext({\n serviceWorkers: \"block\",\n acceptDownloads: false,\n });\n const abort = () => {\n void context.close().catch(() => {});\n };\n signal.addEventListener(\"abort\", abort, { once: true });\n const redirects: WebRedirect[] = [];\n let count = 0,\n bytes = 0;\n try {\n await context.routeWebSocket(\"**/*\", (socket) => socket.close());\n await context.route(\"**/*\", async (route) => {\n const request = route.request();\n try {\n if (\n ++count > 80 ||\n bytes > 20_000_000 ||\n [\"image\", \"media\", \"font\"].includes(request.resourceType())\n )\n return await route.abort();\n const body = request.postData() ?? undefined;\n if (body && body.length > 100_000) return await route.abort();\n const original = request.headers();\n const headers: Record<string, string> = {};\n for (const key of [\"accept\", \"content-type\", \"origin\", \"user-agent\"])\n if (original[key]) headers[key] = original[key];\n const response = await fetchPublicWebResource(request.url(), {\n signal,\n maxBytes: 5_000_000,\n method: request.method(),\n body,\n headers,\n redirect: \"manual\",\n });\n if (\n request.isNavigationRequest() &&\n request.frame() === context.pages()[0]?.mainFrame() &&\n [301, 302, 303, 307, 308].includes(response.status) &&\n response.headers.location\n ) {\n redirects.push({\n from: request.url(),\n to: new URL(response.headers.location, request.url()).href,\n kind: \"http\",\n status: response.status,\n });\n }\n bytes += response.body.byteLength;\n await route.fulfill({\n status: response.status,\n headers: response.headers,\n body: Buffer.from(response.body),\n });\n } catch {\n await route.abort().catch(() => {});\n }\n });\n const page = await context.newPage();\n let previousUrl = url;\n page.on(\"framenavigated\", (frame) => {\n if (frame !== page.mainFrame() || !/^https?:/u.test(frame.url()))\n return;\n const next = frame.url();\n if (next !== previousUrl && !redirects.some((hop) => hop.to === next))\n redirects.push({ from: previousUrl, to: next, kind: \"client\" });\n previousUrl = next;\n });\n const response = await page.goto(url, {\n waitUntil: \"domcontentloaded\",\n timeout: 25000,\n });\n const settled = await page\n .waitForFunction(\n () => {\n const root = document.querySelector(\n \"#root,#app,#__next,app-root,[ng-version]\",\n );\n const rootText = root?.textContent?.trim() ?? \"\";\n return (\n document.body.innerText.trim().length > 100 &&\n (!root ||\n (rootText.length > 20 &&\n !/^(loading|please wait)/i.test(rootText)))\n );\n },\n undefined,\n { timeout: 8000 },\n )\n .then(\n () => true,\n () => false,\n );\n // Allow hydration and async content to settle after the first readable frame.\n await page.waitForTimeout(1200);\n // Bounded scrolling exposes common lazy-loaded sections without clicking actions.\n for (let step = 0; step < 3; step++) {\n await page.evaluate(() => window.scrollBy(0, window.innerHeight));\n await page.waitForTimeout(350);\n }\n await page.evaluate(() => {\n document.querySelectorAll(\"details\").forEach((element) => {\n element.open = true;\n });\n window.scrollTo(0, 0);\n });\n signal.throwIfAborted();\n const html = await page.content();\n if (html.length > 5_000_000)\n throw new WebReadError(\n \"too_large\",\n \"Rendered page exceeded the content limit.\",\n );\n return {\n html,\n redirects,\n settled,\n text: await page.locator(\"body\").innerText(),\n url: page.url(),\n status: response?.status() ?? 200,\n };\n } finally {\n signal.removeEventListener(\"abort\", abort);\n await context.close().catch(() => {});\n }\n };\n const render: WebRenderer = async (url, options) => {\n try {\n return await renderOnce(url, options);\n } catch (error) {\n if (options.signal.aborted || browser?.isConnected()) throw error;\n // A crashed worker may be replaced once; the same total deadline still applies.\n return renderOnce(url, options);\n }\n };\n return {\n render,\n ready: async () => {\n await launch();\n },\n close: async () => {\n await browser?.close();\n browser = undefined;\n },\n };\n};\n",
|
|
6
|
-
"import {\n isPublicNetworkAddress,\n pinnedPublicRequest,\n} from \"@absolutejs/egress/transport\";\nimport { lookup } from \"node:dns/promises\";\nimport { isIP } from \"node:net\";\nimport { brotliDecompressSync, gunzipSync, inflateSync } from \"node:zlib\";\n\nexport class WebReadError extends Error {\n constructor(\n public code: string,\n message: string,\n ) {\n super(message);\n this.name = \"WebReadError\";\n }\n}\nexport const isPublicWebAddress = (address: string) =>\n isPublicNetworkAddress(address);\nexport type WebRedirect = {\n from: string;\n to: string;\n kind: \"http\" | \"client\";\n status?: number;\n};\nexport type WebFetchResult = {\n redirects?: WebRedirect[];\n url: string;\n status: number;\n headers: Record<string, string>;\n body: Uint8Array;\n};\nexport type WebFetchOptions = {\n signal?: AbortSignal;\n redirect?: \"follow\" | \"manual\";\n maxBytes?: number;\n method?: string;\n body?: string;\n headers?: Record<string, string>;\n};\nexport const validatePublicWebUrl = (raw: string) => {\n let url: URL;\n try {\n url = new URL(raw);\n } catch {\n throw new WebReadError(\n \"invalid_url\",\n \"Provide a valid public HTTP or HTTPS URL.\",\n );\n }\n if (\n ![\"http:\", \"https:\"].includes(url.protocol) ||\n url.username ||\n url.password ||\n (url.port && ![\"80\", \"443\"].includes(url.port))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"Only public HTTP/HTTPS URLs on standard ports are supported.\",\n );\n const host = url.hostname.replace(/^\\[|\\]$/gu, \"\").toLowerCase();\n if (\n host === \"localhost\" ||\n /\\.(?:localhost|local|internal)$/u.test(host) ||\n (isIP(host) && !isPublicWebAddress(host))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"Private and local network destinations are not allowed.\",\n );\n return url;\n};\n/** Pin the validated DNS answer into the connection, including every redirect. */\nexport const fetchPublicWebResource = async (\n raw: string,\n options: WebFetchOptions = {},\n): Promise<WebFetchResult> => {\n const signal = options.signal ?? AbortSignal.timeout(15000);\n const maxBytes = options.maxBytes ?? 5_000_000;\n let url = validatePublicWebUrl(raw);\n const redirects: WebRedirect[] = [];\n let method = options.method ?? \"GET\";\n let body = options.body;\n for (let hop = 0; hop <= 5; hop++) {\n signal.throwIfAborted();\n const host = url.hostname.replace(/^\\[|\\]$/gu, \"\");\n const addresses = await lookup(host, { all: true });\n if (\n !addresses.length ||\n addresses.some(({ address }) => !isPublicWebAddress(address))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"The destination resolved to a private or reserved network.\",\n );\n signal.throwIfAborted();\n const selected =\n addresses.find(({ family }) => family === 4) ?? addresses[0]!;\n const incoming = await pinnedPublicRequest(\n new Request(url, {\n method,\n body,\n signal,\n headers: {\n \"user-agent\": \"AbsoluteJSReader/1.0\",\n accept: \"*/*\",\n \"accept-encoding\": \"identity\",\n ...options.headers,\n host: url.host,\n },\n }),\n { hostname: host, address: selected.address, maxResponseBytes: maxBytes },\n );\n let data: Uint8Array = new Uint8Array(await incoming.arrayBuffer());\n const encoding = incoming.headers.get(\"content-encoding\");\n const limits = { maxOutputLength: maxBytes };\n if (encoding === \"gzip\") data = gunzipSync(data, limits);\n else if (encoding === \"br\") data = brotliDecompressSync(data, limits);\n else if (encoding === \"deflate\") data = inflateSync(data, limits);\n const response: WebFetchResult = {\n url: url.href,\n status: incoming.status,\n headers: Object.fromEntries(\n [...incoming.headers].filter(\n ([key]) =>\n ![\n \"content-encoding\",\n \"content-length\",\n \"transfer-encoding\",\n \"set-cookie\",\n ].includes(key),\n ),\n ),\n body: data,\n };\n if (options.redirect === \"manual\") return response;\n if (\n [301, 302, 303, 307, 308].includes(response.status) &&\n response.headers.location\n ) {\n const next = validatePublicWebUrl(\n new URL(response.headers.location, url).href,\n );\n redirects.push({\n from: url.href,\n to: next.href,\n kind: \"http\",\n status: response.status,\n });\n url = next;\n if (\n response.status === 303 ||\n ([301, 302].includes(response.status) && method === \"POST\")\n ) {\n method = \"GET\";\n body = undefined;\n }\n continue;\n }\n return { ...response, redirects };\n }\n throw new WebReadError(\n \"redirect_limit\",\n \"The site exceeded the redirect limit.\",\n );\n};\n"
|
|
6
|
+
"import {\n isPublicNetworkAddress,\n pinnedPublicRequest,\n} from \"@absolutejs/egress/transport\";\nimport { lookup } from \"node:dns/promises\";\nimport { isIP } from \"node:net\";\nimport { brotliDecompressSync, gunzipSync, inflateSync } from \"node:zlib\";\n\n// Package entrypoints are bundled independently; constructor identity is not shared.\nconst WEB_READ_ERROR = Symbol.for(\"@absolutejs/rag/WebReadError\");\nexport class WebReadError extends Error {\n readonly [WEB_READ_ERROR] = true;\n constructor(\n public code: string,\n message: string,\n ) {\n super(message);\n this.name = \"WebReadError\";\n }\n}\nexport const isWebReadError = (error: unknown): error is WebReadError =>\n error instanceof Error &&\n Reflect.get(error, WEB_READ_ERROR) === true &&\n typeof Reflect.get(error, \"code\") === \"string\";\nexport const isPublicWebAddress = (address: string) =>\n isPublicNetworkAddress(address);\nexport type WebRedirect = {\n from: string;\n to: string;\n kind: \"http\" | \"client\";\n status?: number;\n};\nexport type WebFetchResult = {\n redirects?: WebRedirect[];\n url: string;\n status: number;\n headers: Record<string, string>;\n body: Uint8Array;\n};\nexport type WebFetchOptions = {\n signal?: AbortSignal;\n redirect?: \"follow\" | \"manual\";\n maxBytes?: number;\n method?: string;\n body?: string;\n headers?: Record<string, string>;\n};\nexport const validatePublicWebUrl = (raw: string) => {\n let url: URL;\n try {\n url = new URL(raw);\n } catch {\n throw new WebReadError(\n \"invalid_url\",\n \"Provide a valid public HTTP or HTTPS URL.\",\n );\n }\n if (\n ![\"http:\", \"https:\"].includes(url.protocol) ||\n url.username ||\n url.password ||\n (url.port && ![\"80\", \"443\"].includes(url.port))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"Only public HTTP/HTTPS URLs on standard ports are supported.\",\n );\n const host = url.hostname.replace(/^\\[|\\]$/gu, \"\").toLowerCase();\n if (\n host === \"localhost\" ||\n /\\.(?:localhost|local|internal)$/u.test(host) ||\n (isIP(host) && !isPublicWebAddress(host))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"Private and local network destinations are not allowed.\",\n );\n return url;\n};\n/** Pin the validated DNS answer into the connection, including every redirect. */\nexport const fetchPublicWebResource = async (\n raw: string,\n options: WebFetchOptions = {},\n): Promise<WebFetchResult> => {\n const signal = options.signal ?? AbortSignal.timeout(15000);\n const maxBytes = options.maxBytes ?? 5_000_000;\n let url = validatePublicWebUrl(raw);\n const redirects: WebRedirect[] = [];\n let method = options.method ?? \"GET\";\n let body = options.body;\n for (let hop = 0; hop <= 5; hop++) {\n signal.throwIfAborted();\n const host = url.hostname.replace(/^\\[|\\]$/gu, \"\");\n const addresses = await lookup(host, { all: true });\n if (\n !addresses.length ||\n addresses.some(({ address }) => !isPublicWebAddress(address))\n )\n throw new WebReadError(\n \"blocked_url\",\n \"The destination resolved to a private or reserved network.\",\n );\n signal.throwIfAborted();\n const selected =\n addresses.find(({ family }) => family === 4) ?? addresses[0]!;\n const incoming = await pinnedPublicRequest(\n new Request(url, {\n method,\n body,\n signal,\n headers: {\n \"user-agent\": \"AbsoluteJSReader/1.0\",\n accept: \"*/*\",\n \"accept-encoding\": \"identity\",\n ...options.headers,\n host: url.host,\n },\n }),\n { hostname: host, address: selected.address, maxResponseBytes: maxBytes },\n );\n let data: Uint8Array = new Uint8Array(await incoming.arrayBuffer());\n const encoding = incoming.headers.get(\"content-encoding\");\n const limits = { maxOutputLength: maxBytes };\n if (encoding === \"gzip\") data = gunzipSync(data, limits);\n else if (encoding === \"br\") data = brotliDecompressSync(data, limits);\n else if (encoding === \"deflate\") data = inflateSync(data, limits);\n const response: WebFetchResult = {\n url: url.href,\n status: incoming.status,\n headers: Object.fromEntries(\n [...incoming.headers].filter(\n ([key]) =>\n ![\n \"content-encoding\",\n \"content-length\",\n \"transfer-encoding\",\n \"set-cookie\",\n ].includes(key),\n ),\n ),\n body: data,\n };\n if (options.redirect === \"manual\") return response;\n if (\n [301, 302, 303, 307, 308].includes(response.status) &&\n response.headers.location\n ) {\n const next = validatePublicWebUrl(\n new URL(response.headers.location, url).href,\n );\n redirects.push({\n from: url.href,\n to: next.href,\n kind: \"http\",\n status: response.status,\n });\n url = next;\n if (\n response.status === 303 ||\n ([301, 302].includes(response.status) && method === \"POST\")\n ) {\n method = \"GET\";\n body = undefined;\n }\n continue;\n }\n return { ...response, redirects };\n }\n throw new WebReadError(\n \"redirect_limit\",\n \"The site exceeded the redirect limit.\",\n );\n};\n"
|
|
7
7
|
],
|
|
8
|
-
"mappings": ";;AAAA;;;ACAA;AAAA;AAAA;AAAA;AAIA;AACA;AACA;AAAA;
|
|
9
|
-
"debugId": "
|
|
8
|
+
"mappings": ";;AAAA;;;ACAA;AAAA;AAAA;AAAA;AAIA;AACA;AACA;AAGA,IAAM,iBAAiB,OAAO,IAAI,8BAA8B;AAAA;AACzD,MAAM,qBAAqB,MAAM;AAAA,EAG7B;AAAA,GAFC,kBAAkB;AAAA,EAC5B,WAAW,CACF,MACP,SACA;AAAA,IACA,MAAM,OAAO;AAAA,IAHN;AAAA,IAIP,KAAK,OAAO;AAAA;AAEhB;AACO,IAAM,iBAAiB,CAAC,UAC7B,iBAAiB,SACjB,QAAQ,IAAI,OAAO,cAAc,MAAM,QACvC,OAAO,QAAQ,IAAI,OAAO,MAAM,MAAM;AACjC,IAAM,qBAAqB,CAAC,YACjC,uBAAuB,OAAO;AAsBzB,IAAM,uBAAuB,CAAC,QAAgB;AAAA,EACnD,IAAI;AAAA,EACJ,IAAI;AAAA,IACF,MAAM,IAAI,IAAI,GAAG;AAAA,IACjB,MAAM;AAAA,IACN,MAAM,IAAI,aACR,eACA,2CACF;AAAA;AAAA,EAEF,IACE,CAAC,CAAC,SAAS,QAAQ,EAAE,SAAS,IAAI,QAAQ,KAC1C,IAAI,YACJ,IAAI,YACH,IAAI,QAAQ,CAAC,CAAC,MAAM,KAAK,EAAE,SAAS,IAAI,IAAI;AAAA,IAE7C,MAAM,IAAI,aACR,eACA,8DACF;AAAA,EACF,MAAM,OAAO,IAAI,SAAS,QAAQ,aAAa,EAAE,EAAE,YAAY;AAAA,EAC/D,IACE,SAAS,eACT,mCAAmC,KAAK,IAAI,KAC3C,KAAK,IAAI,KAAK,CAAC,mBAAmB,IAAI;AAAA,IAEvC,MAAM,IAAI,aACR,eACA,yDACF;AAAA,EACF,OAAO;AAAA;AAGF,IAAM,yBAAyB,OACpC,KACA,UAA2B,CAAC,MACA;AAAA,EAC5B,MAAM,SAAS,QAAQ,UAAU,YAAY,QAAQ,KAAK;AAAA,EAC1D,MAAM,WAAW,QAAQ,YAAY;AAAA,EACrC,IAAI,MAAM,qBAAqB,GAAG;AAAA,EAClC,MAAM,YAA2B,CAAC;AAAA,EAClC,IAAI,SAAS,QAAQ,UAAU;AAAA,EAC/B,IAAI,OAAO,QAAQ;AAAA,EACnB,SAAS,MAAM,EAAG,OAAO,GAAG,OAAO;AAAA,IACjC,OAAO,eAAe;AAAA,IACtB,MAAM,OAAO,IAAI,SAAS,QAAQ,aAAa,EAAE;AAAA,IACjD,MAAM,YAAY,MAAM,OAAO,MAAM,EAAE,KAAK,KAAK,CAAC;AAAA,IAClD,IACE,CAAC,UAAU,UACX,UAAU,KAAK,GAAG,cAAc,CAAC,mBAAmB,OAAO,CAAC;AAAA,MAE5D,MAAM,IAAI,aACR,eACA,4DACF;AAAA,IACF,OAAO,eAAe;AAAA,IACtB,MAAM,WACJ,UAAU,KAAK,GAAG,aAAa,WAAW,CAAC,KAAK,UAAU;AAAA,IAC5D,MAAM,WAAW,MAAM,oBACrB,IAAI,QAAQ,KAAK;AAAA,MACf;AAAA,MACA;AAAA,MACA;AAAA,MACA,SAAS;AAAA,QACP,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,mBAAmB;AAAA,WAChB,QAAQ;AAAA,QACX,MAAM,IAAI;AAAA,MACZ;AAAA,IACF,CAAC,GACD,EAAE,UAAU,MAAM,SAAS,SAAS,SAAS,kBAAkB,SAAS,CAC1E;AAAA,IACA,IAAI,OAAmB,IAAI,WAAW,MAAM,SAAS,YAAY,CAAC;AAAA,IAClE,MAAM,WAAW,SAAS,QAAQ,IAAI,kBAAkB;AAAA,IACxD,MAAM,SAAS,EAAE,iBAAiB,SAAS;AAAA,IAC3C,IAAI,aAAa;AAAA,MAAQ,OAAO,WAAW,MAAM,MAAM;AAAA,IAClD,SAAI,aAAa;AAAA,MAAM,OAAO,qBAAqB,MAAM,MAAM;AAAA,IAC/D,SAAI,aAAa;AAAA,MAAW,OAAO,YAAY,MAAM,MAAM;AAAA,IAChE,MAAM,WAA2B;AAAA,MAC/B,KAAK,IAAI;AAAA,MACT,QAAQ,SAAS;AAAA,MACjB,SAAS,OAAO,YACd,CAAC,GAAG,SAAS,OAAO,EAAE,OACpB,EAAE,SACA,CAAC;AAAA,QACC;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,EAAE,SAAS,GAAG,CAClB,CACF;AAAA,MACA,MAAM;AAAA,IACR;AAAA,IACA,IAAI,QAAQ,aAAa;AAAA,MAAU,OAAO;AAAA,IAC1C,IACE,CAAC,KAAK,KAAK,KAAK,KAAK,GAAG,EAAE,SAAS,SAAS,MAAM,KAClD,SAAS,QAAQ,UACjB;AAAA,MACA,MAAM,OAAO,qBACX,IAAI,IAAI,SAAS,QAAQ,UAAU,GAAG,EAAE,IAC1C;AAAA,MACA,UAAU,KAAK;AAAA,QACb,MAAM,IAAI;AAAA,QACV,IAAI,KAAK;AAAA,QACT,MAAM;AAAA,QACN,QAAQ,SAAS;AAAA,MACnB,CAAC;AAAA,MACD,MAAM;AAAA,MACN,IACE,SAAS,WAAW,OACnB,CAAC,KAAK,GAAG,EAAE,SAAS,SAAS,MAAM,KAAK,WAAW,QACpD;AAAA,QACA,SAAS;AAAA,QACT,OAAO;AAAA,MACT;AAAA,MACA;AAAA,IACF;AAAA,IACA,OAAO,KAAK,UAAU,UAAU;AAAA,EAClC;AAAA,EACA,MAAM,IAAI,aACR,kBACA,uCACF;AAAA;;;ADlKK,IAAM,8BAA8B,CACzC,UAAuC,CAAC,MACrC;AAAA,EACH,IAAI;AAAA,EACJ,IAAI;AAAA,EACJ,MAAM,SAAS,YAAY;AAAA,IACzB,IAAI,SAAS,YAAY;AAAA,MAAG,OAAO;AAAA,IACnC,cAAc,SAAS,OAAO;AAAA,MAC5B,UAAU;AAAA,MACV,gBAAgB,QAAQ;AAAA,MACxB,MAAM,CAAC,yBAAyB;AAAA,IAClC,CAAC;AAAA,IACD,IAAI;AAAA,MACF,UAAU,MAAM;AAAA,MAChB,OAAO;AAAA,cACP;AAAA,MACA,YAAY;AAAA;AAAA;AAAA,EAGhB,MAAM,aAA0B,OAAO,OAAO,aAAa;AAAA,IACzD,qBAAqB,GAAG;AAAA,IACxB,OAAO,eAAe;AAAA,IACtB,MAAM,WAAW,MAAM,OAAO;AAAA,IAC9B,MAAM,UAAU,MAAM,SAAS,WAAW;AAAA,MACxC,gBAAgB;AAAA,MAChB,iBAAiB;AAAA,IACnB,CAAC;AAAA,IACD,MAAM,QAAQ,MAAM;AAAA,MACb,QAAQ,MAAM,EAAE,MAAM,MAAM,EAAE;AAAA;AAAA,IAErC,OAAO,iBAAiB,SAAS,OAAO,EAAE,MAAM,KAAK,CAAC;AAAA,IACtD,MAAM,YAA2B,CAAC;AAAA,IAClC,IAAI,QAAQ,GACV,QAAQ;AAAA,IACV,IAAI;AAAA,MACF,MAAM,QAAQ,eAAe,QAAQ,CAAC,WAAW,OAAO,MAAM,CAAC;AAAA,MAC/D,MAAM,QAAQ,MAAM,QAAQ,OAAO,UAAU;AAAA,QAC3C,MAAM,UAAU,MAAM,QAAQ;AAAA,QAC9B,IAAI;AAAA,UACF,IACE,EAAE,QAAQ,MACV,QAAQ,YACR,CAAC,SAAS,SAAS,MAAM,EAAE,SAAS,QAAQ,aAAa,CAAC;AAAA,YAE1D,OAAO,MAAM,MAAM,MAAM;AAAA,UAC3B,MAAM,OAAO,QAAQ,SAAS,KAAK;AAAA,UACnC,IAAI,QAAQ,KAAK,SAAS;AAAA,YAAS,OAAO,MAAM,MAAM,MAAM;AAAA,UAC5D,MAAM,WAAW,QAAQ,QAAQ;AAAA,UACjC,MAAM,UAAkC,CAAC;AAAA,UACzC,WAAW,OAAO,CAAC,UAAU,gBAAgB,UAAU,YAAY;AAAA,YACjE,IAAI,SAAS;AAAA,cAAM,QAAQ,OAAO,SAAS;AAAA,UAC7C,MAAM,WAAW,MAAM,uBAAuB,QAAQ,IAAI,GAAG;AAAA,YAC3D;AAAA,YACA,UAAU;AAAA,YACV,QAAQ,QAAQ,OAAO;AAAA,YACvB;AAAA,YACA;AAAA,YACA,UAAU;AAAA,UACZ,CAAC;AAAA,UACD,IACE,QAAQ,oBAAoB,KAC5B,QAAQ,MAAM,MAAM,QAAQ,MAAM,EAAE,IAAI,UAAU,KAClD,CAAC,KAAK,KAAK,KAAK,KAAK,GAAG,EAAE,SAAS,SAAS,MAAM,KAClD,SAAS,QAAQ,UACjB;AAAA,YACA,UAAU,KAAK;AAAA,cACb,MAAM,QAAQ,IAAI;AAAA,cAClB,IAAI,IAAI,IAAI,SAAS,QAAQ,UAAU,QAAQ,IAAI,CAAC,EAAE;AAAA,cACtD,MAAM;AAAA,cACN,QAAQ,SAAS;AAAA,YACnB,CAAC;AAAA,UACH;AAAA,UACA,SAAS,SAAS,KAAK;AAAA,UACvB,MAAM,MAAM,QAAQ;AAAA,YAClB,QAAQ,SAAS;AAAA,YACjB,SAAS,SAAS;AAAA,YAClB,MAAM,OAAO,KAAK,SAAS,IAAI;AAAA,UACjC,CAAC;AAAA,UACD,MAAM;AAAA,UACN,MAAM,MAAM,MAAM,EAAE,MAAM,MAAM,EAAE;AAAA;AAAA,OAErC;AAAA,MACD,MAAM,OAAO,MAAM,QAAQ,QAAQ;AAAA,MACnC,IAAI,cAAc;AAAA,MAClB,KAAK,GAAG,kBAAkB,CAAC,UAAU;AAAA,QACnC,IAAI,UAAU,KAAK,UAAU,KAAK,CAAC,YAAY,KAAK,MAAM,IAAI,CAAC;AAAA,UAC7D;AAAA,QACF,MAAM,OAAO,MAAM,IAAI;AAAA,QACvB,IAAI,SAAS,eAAe,CAAC,UAAU,KAAK,CAAC,QAAQ,IAAI,OAAO,IAAI;AAAA,UAClE,UAAU,KAAK,EAAE,MAAM,aAAa,IAAI,MAAM,MAAM,SAAS,CAAC;AAAA,QAChE,cAAc;AAAA,OACf;AAAA,MACD,MAAM,WAAW,MAAM,KAAK,KAAK,KAAK;AAAA,QACpC,WAAW;AAAA,QACX,SAAS;AAAA,MACX,CAAC;AAAA,MACD,MAAM,UAAU,MAAM,KACnB,gBACC,MAAM;AAAA,QACJ,MAAM,OAAO,SAAS,cACpB,0CACF;AAAA,QACA,MAAM,WAAW,MAAM,aAAa,KAAK,KAAK;AAAA,QAC9C,OACE,SAAS,KAAK,UAAU,KAAK,EAAE,SAAS,QACvC,CAAC,QACC,SAAS,SAAS,MACjB,CAAC,0BAA0B,KAAK,QAAQ;AAAA,SAGhD,WACA,EAAE,SAAS,KAAK,CAClB,EACC,KACC,MAAM,MACN,MAAM,KACR;AAAA,MAEF,MAAM,KAAK,eAAe,IAAI;AAAA,MAE9B,SAAS,OAAO,EAAG,OAAO,GAAG,QAAQ;AAAA,QACnC,MAAM,KAAK,SAAS,MAAM,OAAO,SAAS,GAAG,OAAO,WAAW,CAAC;AAAA,QAChE,MAAM,KAAK,eAAe,GAAG;AAAA,MAC/B;AAAA,MACA,MAAM,KAAK,SAAS,MAAM;AAAA,QACxB,SAAS,iBAAiB,SAAS,EAAE,QAAQ,CAAC,YAAY;AAAA,UACxD,QAAQ,OAAO;AAAA,SAChB;AAAA,QACD,OAAO,SAAS,GAAG,CAAC;AAAA,OACrB;AAAA,MACD,OAAO,eAAe;AAAA,MACtB,MAAM,OAAO,MAAM,KAAK,QAAQ;AAAA,MAChC,IAAI,KAAK,SAAS;AAAA,QAChB,MAAM,IAAI,aACR,aACA,2CACF;AAAA,MACF,OAAO;AAAA,QACL;AAAA,QACA;AAAA,QACA;AAAA,QACA,MAAM,MAAM,KAAK,QAAQ,MAAM,EAAE,UAAU;AAAA,QAC3C,KAAK,KAAK,IAAI;AAAA,QACd,QAAQ,UAAU,OAAO,KAAK;AAAA,MAChC;AAAA,cACA;AAAA,MACA,OAAO,oBAAoB,SAAS,KAAK;AAAA,MACzC,MAAM,QAAQ,MAAM,EAAE,MAAM,MAAM,EAAE;AAAA;AAAA;AAAA,EAGxC,MAAM,SAAsB,OAAO,KAAK,YAAY;AAAA,IAClD,IAAI;AAAA,MACF,OAAO,MAAM,WAAW,KAAK,OAAO;AAAA,MACpC,OAAO,OAAO;AAAA,MACd,IAAI,QAAQ,OAAO,WAAW,SAAS,YAAY;AAAA,QAAG,MAAM;AAAA,MAE5D,OAAO,WAAW,KAAK,OAAO;AAAA;AAAA;AAAA,EAGlC,OAAO;AAAA,IACL;AAAA,IACA,OAAO,YAAY;AAAA,MACjB,MAAM,OAAO;AAAA;AAAA,IAEf,OAAO,YAAY;AAAA,MACjB,MAAM,SAAS,MAAM;AAAA,MACrB,UAAU;AAAA;AAAA,EAEd;AAAA;",
|
|
9
|
+
"debugId": "D0DCB66F403B29EA64756E2164756E21",
|
|
10
10
|
"names": []
|
|
11
11
|
}
|
package/dist/web-index/index.js
CHANGED
|
@@ -101639,15 +101639,18 @@ import {
|
|
|
101639
101639
|
import { lookup } from "dns/promises";
|
|
101640
101640
|
import { isIP } from "net";
|
|
101641
101641
|
import { brotliDecompressSync, gunzipSync, inflateSync } from "zlib";
|
|
101642
|
+
var WEB_READ_ERROR = Symbol.for("@absolutejs/rag/WebReadError");
|
|
101642
101643
|
|
|
101643
101644
|
class WebReadError extends Error {
|
|
101644
101645
|
code;
|
|
101646
|
+
[WEB_READ_ERROR] = true;
|
|
101645
101647
|
constructor(code, message) {
|
|
101646
101648
|
super(message);
|
|
101647
101649
|
this.code = code;
|
|
101648
101650
|
this.name = "WebReadError";
|
|
101649
101651
|
}
|
|
101650
101652
|
}
|
|
101653
|
+
var isWebReadError = (error) => error instanceof Error && Reflect.get(error, WEB_READ_ERROR) === true && typeof Reflect.get(error, "code") === "string";
|
|
101651
101654
|
var isPublicWebAddress = (address) => isPublicNetworkAddress(address);
|
|
101652
101655
|
var validatePublicWebUrl = (raw) => {
|
|
101653
101656
|
let url;
|
|
@@ -123063,7 +123066,7 @@ var createWebpageReaderClient = (options) => {
|
|
|
123063
123066
|
});
|
|
123064
123067
|
if (result.status === "error") {
|
|
123065
123068
|
const error = result.error;
|
|
123066
|
-
throw new
|
|
123069
|
+
throw new WebReadError(error?.code ?? "reader_unavailable", error?.message ?? "Isolated page read failed");
|
|
123067
123070
|
}
|
|
123068
123071
|
const documents = result.documents;
|
|
123069
123072
|
const sources = result.sources;
|
|
@@ -123097,7 +123100,7 @@ var createWebpageReaderClient = (options) => {
|
|
|
123097
123100
|
|
|
123098
123101
|
// src/web/index.ts
|
|
123099
123102
|
var messageFor = (error) => error instanceof Error ? error.message : "Website retrieval failed.";
|
|
123100
|
-
var codeFor = (error) => error
|
|
123103
|
+
var codeFor = (error) => isWebReadError(error) ? error.code : error instanceof Error && /^(?:AbortError|TimeoutError)$/u.test(error.name) ? "timeout" : "fetch_failed";
|
|
123101
123104
|
var cleanHtml = websiteHtmlText;
|
|
123102
123105
|
var titleOf = (html) => {
|
|
123103
123106
|
const title = /<title\b[^>]*>([\s\S]*?)<\/title>/iu.exec(html)?.[1];
|
|
@@ -123144,7 +123147,7 @@ var readRAGWebpage = async (options) => {
|
|
|
123144
123147
|
response = await (options.fetchResource ?? fetchPublicWebResource)(options.url, { signal, maxBytes: 5000000 });
|
|
123145
123148
|
} catch (error) {
|
|
123146
123149
|
attempts.push({ method: "http", status: codeFor(error) });
|
|
123147
|
-
if (error
|
|
123150
|
+
if (isWebReadError(error) && error.code === "blocked_url")
|
|
123148
123151
|
throw error;
|
|
123149
123152
|
if (!options.render)
|
|
123150
123153
|
throw error;
|
|
@@ -123402,6 +123405,8 @@ var createResearch = (config) => {
|
|
|
123402
123405
|
const extract = async (task, input) => {
|
|
123403
123406
|
if (!input.query.trim() || input.query.length > 8000)
|
|
123404
123407
|
throw new Error("Research query must contain 1\u20138000 characters");
|
|
123408
|
+
if (task.collection && (!/^\/(?:[^~]|~[01])+$/.test(task.collection.path) || !Number.isSafeInteger(task.collection.minimum) || task.collection.minimum < 1 || !task.collection.requiredFields.length || task.collection.requiredFields.some((key) => !key || key.includes("/"))))
|
|
123409
|
+
throw new Error("Invalid research collection requirement");
|
|
123405
123410
|
const signal = AbortSignal.any([
|
|
123406
123411
|
AbortSignal.timeout(limits.timeoutMs),
|
|
123407
123412
|
...input.signal ? [input.signal] : []
|
|
@@ -123415,7 +123420,14 @@ var createResearch = (config) => {
|
|
|
123415
123420
|
searches: [],
|
|
123416
123421
|
limitations: [],
|
|
123417
123422
|
generatedAt: new Date().toISOString(),
|
|
123418
|
-
operations: []
|
|
123423
|
+
operations: [],
|
|
123424
|
+
...task.collection ? {
|
|
123425
|
+
coverage: {
|
|
123426
|
+
requested: task.collection.minimum,
|
|
123427
|
+
supported: 0,
|
|
123428
|
+
complete: false
|
|
123429
|
+
}
|
|
123430
|
+
} : {}
|
|
123419
123431
|
};
|
|
123420
123432
|
const operationLimit = limits.searches + limits.reads + limits.rounds + 2 + 2 * validationRepairs;
|
|
123421
123433
|
const operate = async (kind, call, usage = () => ({})) => {
|
|
@@ -123588,16 +123600,68 @@ Correct the previous schema-invalid response using the validation diagnostics. T
|
|
|
123588
123600
|
if (!discoveredLinks.has(url) && !result.searches.some((search) => search.sources.some((source) => source.url === url)))
|
|
123589
123601
|
return;
|
|
123590
123602
|
urls.add(url);
|
|
123591
|
-
const page = await operate("read", () =>
|
|
123592
|
-
|
|
123593
|
-
|
|
123594
|
-
|
|
123595
|
-
|
|
123596
|
-
|
|
123603
|
+
const page = await operate("read", async () => {
|
|
123604
|
+
try {
|
|
123605
|
+
return await (config.reader ?? readRAGWebpage)({
|
|
123606
|
+
url,
|
|
123607
|
+
signal,
|
|
123608
|
+
render: config.render,
|
|
123609
|
+
maxChars: Math.min(limits.evidenceChars, 24000)
|
|
123610
|
+
});
|
|
123611
|
+
} catch (error) {
|
|
123612
|
+
signal.throwIfAborted();
|
|
123613
|
+
const messages = {
|
|
123614
|
+
access_blocked: "Website blocked automated access",
|
|
123615
|
+
authentication_required: "Page requires sign-in",
|
|
123616
|
+
http_error: "Page returned an HTTP error",
|
|
123617
|
+
reader_unavailable: "Website reader unavailable",
|
|
123618
|
+
fetch_failed: "Page could not be retrieved",
|
|
123619
|
+
timeout: "Page retrieval timed out"
|
|
123620
|
+
};
|
|
123621
|
+
if (!isWebReadError(error) || !messages[error.code])
|
|
123622
|
+
throw error;
|
|
123623
|
+
return {
|
|
123624
|
+
status: "error",
|
|
123625
|
+
url,
|
|
123626
|
+
finalUrl: url,
|
|
123627
|
+
title: null,
|
|
123628
|
+
text: "",
|
|
123629
|
+
method: "http",
|
|
123630
|
+
truncated: false,
|
|
123631
|
+
attempts: [],
|
|
123632
|
+
fetchedAt: new Date().toISOString(),
|
|
123633
|
+
redirects: [],
|
|
123634
|
+
evidence: { links: [], images: [], media: [], canonicalUrl: null },
|
|
123635
|
+
limitations: [],
|
|
123636
|
+
error: { code: error.code, message: messages[error.code] }
|
|
123637
|
+
};
|
|
123638
|
+
}
|
|
123639
|
+
});
|
|
123640
|
+
if (page.status === "error") {
|
|
123641
|
+
const code = page.error?.code ?? "read_failed";
|
|
123642
|
+
if ([
|
|
123643
|
+
"unsafe_url",
|
|
123644
|
+
"blocked_address",
|
|
123645
|
+
"blocked_url",
|
|
123646
|
+
"invalid_url"
|
|
123647
|
+
].includes(code))
|
|
123648
|
+
throw new WebReadError(code, "Unsafe research source URL");
|
|
123649
|
+
const message = "Source unavailable (" + code + ")";
|
|
123650
|
+
(result.sourceFailures ??= []).push({ url, code, message });
|
|
123651
|
+
result.limitations.push(`${message}: ${url}`);
|
|
123652
|
+
try {
|
|
123653
|
+
config.onError?.({
|
|
123654
|
+
runId: result.id,
|
|
123655
|
+
phase: "read",
|
|
123656
|
+
error: new WebReadError(code, message)
|
|
123657
|
+
});
|
|
123658
|
+
} catch {}
|
|
123659
|
+
return;
|
|
123660
|
+
}
|
|
123597
123661
|
result.limitations.push(...page.limitations);
|
|
123598
123662
|
if (page.status !== "ok")
|
|
123599
123663
|
result.limitations.push(`Page read ${page.status}: ${url}`);
|
|
123600
|
-
if (page.text
|
|
123664
|
+
if (page.text) {
|
|
123601
123665
|
for (const link of page.evidence.links) {
|
|
123602
123666
|
if (discoveredLinks.size >= 64)
|
|
123603
123667
|
break;
|
|
@@ -123634,7 +123698,7 @@ Correct the previous schema-invalid response using the validation diagnostics. T
|
|
|
123634
123698
|
try {
|
|
123635
123699
|
const needsInitialPlan = !input.queries && (input.queryMode === "task" || boundedQueries(input.query).length > 1) && limits.rounds > 0;
|
|
123636
123700
|
if (needsInitialPlan) {
|
|
123637
|
-
const initial = await generate("plan", PlanSchema, "Turn the research task into concise search-engine queries. Preserve the requested company identity and target criteria, but never copy instructions, JSON, internal markers, or the whole brief into a query. Prioritize the most useful query first. Return at most remainingSearches queries. Return no URLs: source URLs must be discovered rather than invented.", {
|
|
123701
|
+
const initial = await generate("plan", PlanSchema, "Turn the research task into concise search-engine queries. Search for the requested answer entities, not merely the client named in the brief. For new partners, pursue external candidate companies across the requested categories; client background and existing partnerships do not fulfill that task. Preserve the requested company identity and target criteria, but never copy instructions, JSON, internal markers, or the whole brief into a query. Prioritize the most useful query first. Return at most remainingSearches queries. Return no URLs: source URLs must be discovered rather than invented.", {
|
|
123638
123702
|
query: input.query,
|
|
123639
123703
|
instructions: task.instructions,
|
|
123640
123704
|
remainingSearches: limits.searches
|
|
@@ -123700,12 +123764,32 @@ Correct the previous schema-invalid response using the validation diagnostics. T
|
|
|
123700
123764
|
instructions: task.instructions,
|
|
123701
123765
|
schema: task.schema,
|
|
123702
123766
|
fields: leaves,
|
|
123703
|
-
sources: reviewSources
|
|
123767
|
+
sources: reviewSources.map(({ passages, ...source }) => ({
|
|
123768
|
+
...source,
|
|
123769
|
+
passages: passages.map(({ id, text }) => ({ id, text }))
|
|
123770
|
+
}))
|
|
123704
123771
|
});
|
|
123705
123772
|
result.fields = bindResearchReview(leaves, reviewed.fields.map((field) => ({
|
|
123706
123773
|
...field,
|
|
123707
123774
|
citations: resolveResearchPassages(field.citations, reviewSources)
|
|
123708
123775
|
})), result.sources);
|
|
123776
|
+
if (task.collection) {
|
|
123777
|
+
const { path, minimum, requiredFields } = task.collection;
|
|
123778
|
+
const indices = new Set(result.fields.flatMap((field) => {
|
|
123779
|
+
if (!field.path.startsWith(path + "/"))
|
|
123780
|
+
return [];
|
|
123781
|
+
const index = field.path.slice(path.length + 1).split("/")[0];
|
|
123782
|
+
return /^\d+$/.test(index) ? [index] : [];
|
|
123783
|
+
}));
|
|
123784
|
+
const supported = [...indices].filter((index) => requiredFields.every((key) => result.fields.some((field) => field.path === `${path}/${index}/${key}` && field.verdict === "supported" && field.value !== null && field.value !== ""))).length;
|
|
123785
|
+
result.coverage = {
|
|
123786
|
+
requested: minimum,
|
|
123787
|
+
supported,
|
|
123788
|
+
complete: supported >= minimum
|
|
123789
|
+
};
|
|
123790
|
+
if (!result.coverage.complete)
|
|
123791
|
+
result.limitations.push(`Requested ${minimum} evidence-backed items; established ${supported}.`);
|
|
123792
|
+
}
|
|
123709
123793
|
const complete = result.fields.length > 0 && result.fields.every((field) => field.verdict === "supported");
|
|
123710
123794
|
result.data = complete ? data : null;
|
|
123711
123795
|
result.status = complete && !result.limitations.length ? "reviewed" : "partial";
|
|
@@ -123837,5 +123921,5 @@ export {
|
|
|
123837
123921
|
webIndexPostgresSchemaSql
|
|
123838
123922
|
};
|
|
123839
123923
|
|
|
123840
|
-
//# debugId=
|
|
123924
|
+
//# debugId=FEE9813D4409B5C064756E2164756E21
|
|
123841
123925
|
//# sourceMappingURL=index.js.map
|