Spaces:
Build error
Build error
fix: add cache tolerance
Browse files
backend/functions/src/cloud-functions/crawler.ts
CHANGED
|
@@ -426,6 +426,7 @@ ${this.content}
|
|
| 426 |
const customMode = ctx.req.get('x-respond-with') || 'default';
|
| 427 |
const withGeneratedAlt = Boolean(ctx.req.get('x-with-generated-alt'));
|
| 428 |
const noCache = Boolean(ctx.req.get('x-no-cache'));
|
|
|
|
| 429 |
const cookies: CookieParam[] = [];
|
| 430 |
const setCookieHeaders = ctx.req.headers['x-set-cookie'];
|
| 431 |
if (Array.isArray(setCookieHeaders)) {
|
|
@@ -454,7 +455,7 @@ ${this.content}
|
|
| 454 |
rpcReflect.return(sseStream);
|
| 455 |
|
| 456 |
try {
|
| 457 |
-
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts,
|
| 458 |
if (!scrapped) {
|
| 459 |
continue;
|
| 460 |
}
|
|
@@ -481,7 +482,7 @@ ${this.content}
|
|
| 481 |
|
| 482 |
let lastScrapped;
|
| 483 |
if (!ctx.req.accepts('text/plain') && (ctx.req.accepts('text/json') || ctx.req.accepts('application/json'))) {
|
| 484 |
-
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts,
|
| 485 |
lastScrapped = scrapped;
|
| 486 |
if (!scrapped?.parsed?.content || !(scrapped.title?.trim())) {
|
| 487 |
continue;
|
|
@@ -503,7 +504,7 @@ ${this.content}
|
|
| 503 |
return formatted;
|
| 504 |
}
|
| 505 |
|
| 506 |
-
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts,
|
| 507 |
lastScrapped = scrapped;
|
| 508 |
if (!scrapped?.parsed?.content || !(scrapped.title?.trim())) {
|
| 509 |
continue;
|
|
@@ -546,7 +547,7 @@ ${this.content}
|
|
| 546 |
return digest;
|
| 547 |
}
|
| 548 |
|
| 549 |
-
async queryCache(urlToCrawl: URL) {
|
| 550 |
const digest = this.getUrlDigest(urlToCrawl);
|
| 551 |
|
| 552 |
const cache = (await Crawled.fromFirestoreQuery(Crawled.COLLECTION.where('urlPathDigest', '==', digest).orderBy('createdAt', 'desc').limit(1)))?.[0];
|
|
@@ -556,9 +557,9 @@ ${this.content}
|
|
| 556 |
}
|
| 557 |
|
| 558 |
const age = Date.now() - cache.createdAt.valueOf();
|
| 559 |
-
const stale = cache.createdAt.valueOf() < (Date.now() -
|
| 560 |
-
this.logger.info(`${stale ? 'Stale cache exists' : 'Cache hit'} for ${urlToCrawl}, normalized digest: ${digest}, ${age}ms old`, {
|
| 561 |
-
url: urlToCrawl, digest, age, stale
|
| 562 |
});
|
| 563 |
|
| 564 |
let snapshot: PageSnapshot | undefined;
|
|
@@ -641,10 +642,10 @@ ${this.content}
|
|
| 641 |
return r;
|
| 642 |
}
|
| 643 |
|
| 644 |
-
async *cachedScrap(urlToCrawl: URL, crawlOpts?: ScrappingOptions,
|
| 645 |
let cache;
|
| 646 |
-
if (
|
| 647 |
-
cache = await this.queryCache(urlToCrawl);
|
| 648 |
}
|
| 649 |
|
| 650 |
if (cache?.isFresh && (!crawlOpts?.favorScreenshot || (crawlOpts?.favorScreenshot && cache?.screenshotAvailable))) {
|
|
@@ -687,10 +688,10 @@ ${this.content}
|
|
| 687 |
}
|
| 688 |
|
| 689 |
|
| 690 |
-
async *scrapMany(urls: URL[], options?: ScrappingOptions,
|
| 691 |
-
const iterators = urls.map((url) => this.cachedScrap(url, options,
|
| 692 |
|
| 693 |
-
const results: (PageSnapshot | undefined)[] = iterators.map((_x)=> undefined);
|
| 694 |
|
| 695 |
let nextDeferred = Defer();
|
| 696 |
let concluded = false;
|
|
|
|
| 426 |
const customMode = ctx.req.get('x-respond-with') || 'default';
|
| 427 |
const withGeneratedAlt = Boolean(ctx.req.get('x-with-generated-alt'));
|
| 428 |
const noCache = Boolean(ctx.req.get('x-no-cache'));
|
| 429 |
+
const cacheTolerance = noCache ? 0 : this.cacheValidMs;
|
| 430 |
const cookies: CookieParam[] = [];
|
| 431 |
const setCookieHeaders = ctx.req.headers['x-set-cookie'];
|
| 432 |
if (Array.isArray(setCookieHeaders)) {
|
|
|
|
| 455 |
rpcReflect.return(sseStream);
|
| 456 |
|
| 457 |
try {
|
| 458 |
+
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts, cacheTolerance)) {
|
| 459 |
if (!scrapped) {
|
| 460 |
continue;
|
| 461 |
}
|
|
|
|
| 482 |
|
| 483 |
let lastScrapped;
|
| 484 |
if (!ctx.req.accepts('text/plain') && (ctx.req.accepts('text/json') || ctx.req.accepts('application/json'))) {
|
| 485 |
+
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts, cacheTolerance)) {
|
| 486 |
lastScrapped = scrapped;
|
| 487 |
if (!scrapped?.parsed?.content || !(scrapped.title?.trim())) {
|
| 488 |
continue;
|
|
|
|
| 504 |
return formatted;
|
| 505 |
}
|
| 506 |
|
| 507 |
+
for await (const scrapped of this.cachedScrap(urlToCrawl, crawlOpts, cacheTolerance)) {
|
| 508 |
lastScrapped = scrapped;
|
| 509 |
if (!scrapped?.parsed?.content || !(scrapped.title?.trim())) {
|
| 510 |
continue;
|
|
|
|
| 547 |
return digest;
|
| 548 |
}
|
| 549 |
|
| 550 |
+
async queryCache(urlToCrawl: URL, cacheTolerance: number) {
|
| 551 |
const digest = this.getUrlDigest(urlToCrawl);
|
| 552 |
|
| 553 |
const cache = (await Crawled.fromFirestoreQuery(Crawled.COLLECTION.where('urlPathDigest', '==', digest).orderBy('createdAt', 'desc').limit(1)))?.[0];
|
|
|
|
| 557 |
}
|
| 558 |
|
| 559 |
const age = Date.now() - cache.createdAt.valueOf();
|
| 560 |
+
const stale = cache.createdAt.valueOf() < (Date.now() - cacheTolerance);
|
| 561 |
+
this.logger.info(`${stale ? 'Stale cache exists' : 'Cache hit'} for ${urlToCrawl}, normalized digest: ${digest}, ${age}ms old, tolerance ${cacheTolerance}ms`, {
|
| 562 |
+
url: urlToCrawl, digest, age, stale, cacheTolerance
|
| 563 |
});
|
| 564 |
|
| 565 |
let snapshot: PageSnapshot | undefined;
|
|
|
|
| 642 |
return r;
|
| 643 |
}
|
| 644 |
|
| 645 |
+
async *cachedScrap(urlToCrawl: URL, crawlOpts?: ScrappingOptions, cacheTolerance: number = this.cacheValidMs) {
|
| 646 |
let cache;
|
| 647 |
+
if (cacheTolerance && !crawlOpts?.cookies?.length) {
|
| 648 |
+
cache = await this.queryCache(urlToCrawl, cacheTolerance);
|
| 649 |
}
|
| 650 |
|
| 651 |
if (cache?.isFresh && (!crawlOpts?.favorScreenshot || (crawlOpts?.favorScreenshot && cache?.screenshotAvailable))) {
|
|
|
|
| 688 |
}
|
| 689 |
|
| 690 |
|
| 691 |
+
async *scrapMany(urls: URL[], options?: ScrappingOptions, cacheTolerance?: number) {
|
| 692 |
+
const iterators = urls.map((url) => this.cachedScrap(url, options, cacheTolerance));
|
| 693 |
|
| 694 |
+
const results: (PageSnapshot | undefined)[] = iterators.map((_x) => undefined);
|
| 695 |
|
| 696 |
let nextDeferred = Defer();
|
| 697 |
let concluded = false;
|
backend/functions/src/cloud-functions/searcher.ts
CHANGED
|
@@ -18,7 +18,7 @@ import { CookieParam } from 'puppeteer';
|
|
| 18 |
import { parseString as parseSetCookieString } from 'set-cookie-parser';
|
| 19 |
import { WebSearchQueryParams } from '../shared/3rd-party/brave-search';
|
| 20 |
import { SearchResult } from '../db/searched';
|
| 21 |
-
import { WebSearchApiResponse } from '../shared/3rd-party/brave-types';
|
| 22 |
|
| 23 |
|
| 24 |
@singleton()
|
|
@@ -27,6 +27,9 @@ export class SearcherHost extends RPCHost {
|
|
| 27 |
|
| 28 |
cacheRetentionMs = 1000 * 3600 * 24 * 7;
|
| 29 |
cacheValidMs = 1000 * 3600;
|
|
|
|
|
|
|
|
|
|
| 30 |
|
| 31 |
constructor(
|
| 32 |
protected globalLogger: Logger,
|
|
@@ -178,6 +181,7 @@ export class SearcherHost extends RPCHost {
|
|
| 178 |
const customMode = ctx.req.get('x-respond-with') || 'default';
|
| 179 |
const withGeneratedAlt = Boolean(ctx.req.get('x-with-generated-alt'));
|
| 180 |
const noCache = Boolean(ctx.req.get('x-no-cache'));
|
|
|
|
| 181 |
const cookies: CookieParam[] = [];
|
| 182 |
const setCookieHeaders = ctx.req.headers['x-set-cookie'];
|
| 183 |
if (Array.isArray(setCookieHeaders)) {
|
|
@@ -204,8 +208,7 @@ export class SearcherHost extends RPCHost {
|
|
| 204 |
count: 5
|
| 205 |
});
|
| 206 |
|
| 207 |
-
const
|
| 208 |
-
const it = this.fetchSearchResults(customMode, urls, crawlOpts, noCache);
|
| 209 |
|
| 210 |
if (!ctx.req.accepts('text/plain') && ctx.req.accepts('text/event-stream')) {
|
| 211 |
const sseStream = new OutputServerEventStream();
|
|
@@ -238,12 +241,14 @@ export class SearcherHost extends RPCHost {
|
|
| 238 |
return sseStream;
|
| 239 |
}
|
| 240 |
|
|
|
|
|
|
|
| 241 |
let lastScrapped;
|
| 242 |
if (!ctx.req.accepts('text/plain') && (ctx.req.accepts('text/json') || ctx.req.accepts('application/json'))) {
|
| 243 |
for await (const scrapped of it) {
|
| 244 |
lastScrapped = scrapped;
|
| 245 |
|
| 246 |
-
if (!this.qualified(scrapped)) {
|
| 247 |
continue;
|
| 248 |
}
|
| 249 |
|
|
@@ -264,7 +269,7 @@ export class SearcherHost extends RPCHost {
|
|
| 264 |
for await (const scrapped of it) {
|
| 265 |
lastScrapped = scrapped;
|
| 266 |
|
| 267 |
-
if (!this.qualified(scrapped)) {
|
| 268 |
continue;
|
| 269 |
}
|
| 270 |
chargeAmount = this.getChargeAmount(scrapped);
|
|
@@ -282,18 +287,27 @@ export class SearcherHost extends RPCHost {
|
|
| 282 |
}
|
| 283 |
|
| 284 |
async *fetchSearchResults(mode: string | 'markdown' | 'html' | 'text' | 'screenshot',
|
| 285 |
-
|
| 286 |
-
|
| 287 |
-
for await (const scrapped of this.crawler.scrapMany(urls, options,
|
| 288 |
const mapped = scrapped.map((x, i) => {
|
| 289 |
-
|
|
|
|
| 290 |
const p = {
|
| 291 |
-
toString() {
|
| 292 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 293 |
}
|
| 294 |
};
|
| 295 |
const r = Object.create(p);
|
| 296 |
-
r.url =
|
|
|
|
|
|
|
| 297 |
|
| 298 |
return r;
|
| 299 |
}
|
|
@@ -317,7 +331,7 @@ export class SearcherHost extends RPCHost {
|
|
| 317 |
[${i + 1}] URL Source: ${this.url}${mixins.length ? `\n${mixins.join('\n')}` : ''}
|
| 318 |
[${i + 1}] Markdown Content:
|
| 319 |
${this.content}
|
| 320 |
-
|
| 321 |
};
|
| 322 |
}
|
| 323 |
}
|
|
|
|
| 18 |
import { parseString as parseSetCookieString } from 'set-cookie-parser';
|
| 19 |
import { WebSearchQueryParams } from '../shared/3rd-party/brave-search';
|
| 20 |
import { SearchResult } from '../db/searched';
|
| 21 |
+
import { WebSearchApiResponse, SearchResult as WebSearchResult } from '../shared/3rd-party/brave-types';
|
| 22 |
|
| 23 |
|
| 24 |
@singleton()
|
|
|
|
| 27 |
|
| 28 |
cacheRetentionMs = 1000 * 3600 * 24 * 7;
|
| 29 |
cacheValidMs = 1000 * 3600;
|
| 30 |
+
pageCacheToleranceMs = 1000 * 3600 * 24;
|
| 31 |
+
|
| 32 |
+
reasonableDelayMs = 10_000;
|
| 33 |
|
| 34 |
constructor(
|
| 35 |
protected globalLogger: Logger,
|
|
|
|
| 181 |
const customMode = ctx.req.get('x-respond-with') || 'default';
|
| 182 |
const withGeneratedAlt = Boolean(ctx.req.get('x-with-generated-alt'));
|
| 183 |
const noCache = Boolean(ctx.req.get('x-no-cache'));
|
| 184 |
+
const pageCacheTolerance = noCache ? 0 : this.pageCacheToleranceMs;
|
| 185 |
const cookies: CookieParam[] = [];
|
| 186 |
const setCookieHeaders = ctx.req.headers['x-set-cookie'];
|
| 187 |
if (Array.isArray(setCookieHeaders)) {
|
|
|
|
| 208 |
count: 5
|
| 209 |
});
|
| 210 |
|
| 211 |
+
const it = this.fetchSearchResults(customMode, r.web.results, crawlOpts, pageCacheTolerance);
|
|
|
|
| 212 |
|
| 213 |
if (!ctx.req.accepts('text/plain') && ctx.req.accepts('text/event-stream')) {
|
| 214 |
const sseStream = new OutputServerEventStream();
|
|
|
|
| 241 |
return sseStream;
|
| 242 |
}
|
| 243 |
|
| 244 |
+
const t0 = Date.now();
|
| 245 |
+
|
| 246 |
let lastScrapped;
|
| 247 |
if (!ctx.req.accepts('text/plain') && (ctx.req.accepts('text/json') || ctx.req.accepts('application/json'))) {
|
| 248 |
for await (const scrapped of it) {
|
| 249 |
lastScrapped = scrapped;
|
| 250 |
|
| 251 |
+
if (!this.qualified(scrapped) && ((Date.now() - t0) < this.reasonableDelayMs)) {
|
| 252 |
continue;
|
| 253 |
}
|
| 254 |
|
|
|
|
| 269 |
for await (const scrapped of it) {
|
| 270 |
lastScrapped = scrapped;
|
| 271 |
|
| 272 |
+
if (!this.qualified(scrapped) && ((Date.now() - t0) < this.reasonableDelayMs)) {
|
| 273 |
continue;
|
| 274 |
}
|
| 275 |
chargeAmount = this.getChargeAmount(scrapped);
|
|
|
|
| 287 |
}
|
| 288 |
|
| 289 |
async *fetchSearchResults(mode: string | 'markdown' | 'html' | 'text' | 'screenshot',
|
| 290 |
+
searchResults: WebSearchResult[], options?: ScrappingOptions, pageCacheTolerance?: number) {
|
| 291 |
+
const urls = searchResults.map((x) => new URL(x.url));
|
| 292 |
+
for await (const scrapped of this.crawler.scrapMany(urls, options, pageCacheTolerance)) {
|
| 293 |
const mapped = scrapped.map((x, i) => {
|
| 294 |
+
const upstreamSearchResult = searchResults[i];
|
| 295 |
+
if (!x || (!x.parsed && mode !== 'markdown')) {
|
| 296 |
const p = {
|
| 297 |
+
toString(this: any) {
|
| 298 |
+
if (this.title && this.description) {
|
| 299 |
+
return `[${i + 1}] Title: ${this.title}
|
| 300 |
+
[${i + 1}] URL Source: ${this.url}
|
| 301 |
+
[${i + 1}] Description: ${this.description}
|
| 302 |
+
`;
|
| 303 |
+
}
|
| 304 |
+
return `[${i + 1}] No content available for ${this.url}`;
|
| 305 |
}
|
| 306 |
};
|
| 307 |
const r = Object.create(p);
|
| 308 |
+
r.url = upstreamSearchResult.url;
|
| 309 |
+
r.title = upstreamSearchResult.title;
|
| 310 |
+
r.description = upstreamSearchResult.description;
|
| 311 |
|
| 312 |
return r;
|
| 313 |
}
|
|
|
|
| 331 |
[${i + 1}] URL Source: ${this.url}${mixins.length ? `\n${mixins.join('\n')}` : ''}
|
| 332 |
[${i + 1}] Markdown Content:
|
| 333 |
${this.content}
|
| 334 |
+
`;
|
| 335 |
};
|
| 336 |
}
|
| 337 |
}
|