Merge pull request #1 from mendableai/main

Fix FIRECRAWL_API_URL bug, also various PyLint fixes

.github/workflows/ci.yml

@@ -25,6 +25,9 @@ env:
   SUPABASE_SERVICE_TOKEN: ${{ secrets.SUPABASE_SERVICE_TOKEN }}
   SUPABASE_URL: ${{ secrets.SUPABASE_URL }}
   TEST_API_KEY: ${{ secrets.TEST_API_KEY }}
+  HYPERDX_API_KEY: ${{ secrets.HYPERDX_API_KEY }}
+  HDX_NODE_BETA_MODE: 1
+
 
 jobs:
   pre-deploy:

apps/api/.env.example

@@ -31,3 +31,8 @@ POSTHOG_HOST= # set if you'd like to send posthog events like job logs
 
 STRIPE_PRICE_ID_STANDARD=
 STRIPE_PRICE_ID_SCALE=
+
+HYPERDX_API_KEY=
+HDX_NODE_BETA_MODE=1
+
+FIRE_ENGINE_BETA_URL= # set if you'd like to use the fire engine closed beta

apps/api/openapi.json

@@ -479,6 +479,16 @@
                     "format": "uri"
                   }
                 }
+              },
+              "llm_extraction": {
+                "type": "object",
+                "description": "Displayed when using LLM Extraction. Extracted data from the page following the schema defined.",
+                "nullable": true
+              },
+              "warning": {
+                "type": "string",
+                "nullable": true,
+                "description": "Can be displayed when using LLM Extraction. Warning message will let you know any issues with the extraction."
               }
             }
           }

apps/api/package.json

@@ -48,6 +48,7 @@
     "@bull-board/express": "^5.8.0",
     "@devil7softwares/pos": "^1.0.2",
     "@dqbd/tiktoken": "^1.0.13",
+    "@hyperdx/node-opentelemetry": "^0.7.0",
     "@logtail/node": "^0.4.12",
     "@nangohq/node": "^0.36.33",
     "@sentry/node": "^7.48.0",

apps/api/src/__tests__/e2e_withAuth/index.test.ts

@@ -81,7 +81,7 @@ describe("E2E Tests for API Routes", () => {
       expect(response.body.data).toHaveProperty("markdown");
       expect(response.body.data).toHaveProperty("metadata");
       expect(response.body.data).not.toHaveProperty("html");
-      expect(response.body.data.content).toContain("🔥 FireCrawl");
+      expect(response.body.data.content).toContain("🔥 Firecrawl");
     }, 30000); // 30 seconds timeout
 
     it("should return a successful response with a valid API key and includeHtml set to true", async () => {
@@ -99,8 +99,8 @@ describe("E2E Tests for API Routes", () => {
       expect(response.body.data).toHaveProperty("markdown");
       expect(response.body.data).toHaveProperty("html");
       expect(response.body.data).toHaveProperty("metadata");
-      expect(response.body.data.content).toContain("🔥 FireCrawl");
-      expect(response.body.data.markdown).toContain("🔥 FireCrawl");
+      expect(response.body.data.content).toContain("🔥 Firecrawl");
+      expect(response.body.data.markdown).toContain("🔥 Firecrawl");
       expect(response.body.data.html).toContain("<h1");
     }, 30000); // 30 seconds timeout
 
@@ -266,7 +266,7 @@ describe("E2E Tests for API Routes", () => {
       urls.forEach((url: string) => {
         expect(url.startsWith("https://wwww.mendable.ai/blog/")).toBeFalsy();
       });
-    }, 60000); // 60 seconds
+    }, 90000); // 90 seconds
 
     it("should return a successful response with a valid API key and limit to 3", async () => {
       const crawlResponse = await request(TEST_URL)
@@ -440,8 +440,8 @@ describe("E2E Tests for API Routes", () => {
       // 120 seconds
       expect(completedResponse.body.data[0]).toHaveProperty("html");
       expect(completedResponse.body.data[0]).toHaveProperty("metadata");
-      expect(completedResponse.body.data[0].content).toContain("🔥 FireCrawl");
-      expect(completedResponse.body.data[0].markdown).toContain("FireCrawl");
+      expect(completedResponse.body.data[0].content).toContain("🔥 Firecrawl");
+      expect(completedResponse.body.data[0].markdown).toContain("Firecrawl");
       expect(completedResponse.body.data[0].html).toContain("<h1");
     }, 60000);
   });
@@ -576,7 +576,7 @@ describe("E2E Tests for API Routes", () => {
       expect(completedResponse.body.data[0]).toHaveProperty("content");
       expect(completedResponse.body.data[0]).toHaveProperty("markdown");
       expect(completedResponse.body.data[0]).toHaveProperty("metadata");
-      expect(completedResponse.body.data[0].content).toContain("🔥 FireCrawl");
+      expect(completedResponse.body.data[0].content).toContain("🔥 Firecrawl");
     }, 60000); // 60 seconds
 
     it('should return a successful response for a valid crawl job with PDF files without explicit .pdf extension', async () => {
@@ -697,8 +697,8 @@ describe("E2E Tests for API Routes", () => {
       // 120 seconds
       expect(completedResponse.body.data[0]).toHaveProperty("html");
       expect(completedResponse.body.data[0]).toHaveProperty("metadata");
-      expect(completedResponse.body.data[0].content).toContain("🔥 FireCrawl");
-      expect(completedResponse.body.data[0].markdown).toContain("FireCrawl");
+      expect(completedResponse.body.data[0].content).toContain("🔥 Firecrawl");
+      expect(completedResponse.body.data[0].markdown).toContain("Firecrawl");
       expect(completedResponse.body.data[0].html).toContain("<h1");
     }, 60000);
   }); // 60 seconds

apps/api/src/controllers/auth.ts

@@ -4,11 +4,22 @@ import { AuthResponse, RateLimiterMode } from "../../src/types";
 import { supabase_service } from "../../src/services/supabase";
 import { withAuth } from "../../src/lib/withAuth";
 import { RateLimiterRedis } from "rate-limiter-flexible";
+import { setTraceAttributes } from '@hyperdx/node-opentelemetry';
 
 export async function authenticateUser(req, res, mode?: RateLimiterMode) : Promise<AuthResponse> {
   return withAuth(supaAuthenticateUser)(req, res, mode);
 }
-
+function setTrace(team_id: string, api_key: string) {
+  try {
+    setTraceAttributes({
+      team_id,
+      api_key
+    });
+  } catch (error) {
+    console.error('Error setting trace attributes:', error);
+  }
+
+}
 export async function supaAuthenticateUser(
   req,
   res,
@@ -78,11 +89,13 @@ export async function supaAuthenticateUser(
         status: 401,
       };
     }
-
-
+    const team_id = data[0].team_id;
+    const plan = getPlanByPriceId(data[0].price_id);
+    // HyperDX Logging
+    setTrace(team_id, normalizedApi);
     subscriptionData = {
-      team_id: data[0].team_id,
-      plan: getPlanByPriceId(data[0].price_id)
+      team_id: team_id,
+      plan: plan
     }
     switch (mode) { 
       case RateLimiterMode.Crawl:

apps/api/src/controllers/scrape.ts

@@ -106,6 +106,9 @@ export async function scrapeController(req: Request, res: Response) {
     const extractorOptions = req.body.extractorOptions ?? {
       mode: "markdown"
     }
+    if (extractorOptions.mode === "llm-extraction") {
+      pageOptions.onlyMainContent = true;
+    }
     const origin = req.body.origin ?? "api";
     const timeout = req.body.timeout ?? 30000; // Default timeout of 30 seconds
 

apps/api/src/index.ts

@@ -5,6 +5,8 @@ import "dotenv/config";
 import { getWebScraperQueue } from "./services/queue-service";
 import { redisClient } from "./services/rate-limiter";
 import { v0Router } from "./routes/v0";
+import { initSDK } from '@hyperdx/node-opentelemetry';
+
 const { createBullBoard } = require("@bull-board/api");
 const { BullAdapter } = require("@bull-board/api/bullAdapter");
 const { ExpressAdapter } = require("@bull-board/express");
@@ -47,6 +49,11 @@ const DEFAULT_PORT = process.env.PORT ?? 3002;
 const HOST = process.env.HOST ?? "localhost";
 redisClient.connect();
 
+// HyperDX OpenTelemetry
+if(process.env.ENV === 'production') {
+  initSDK({ consoleCapture: true, additionalInstrumentations: []});
+}
+
 
 export function startServer(port = DEFAULT_PORT) {
   const server = app.listen(Number(port), HOST, () => {

apps/api/src/lib/LLM-extraction/models.ts

@@ -1,25 +1,38 @@
 import OpenAI from "openai";
 import { Document } from "../../lib/entities";
+import { numTokensFromString } from "./helpers";
 
 export type ScraperCompletionResult = {
   data: any | null;
   url: string;
 };
 
+const maxTokens = 32000;
+const modifier = 4;
 const defaultPrompt =
   "You are a professional web scraper. Extract the contents of the webpage";
 
 function prepareOpenAIDoc(
   document: Document
-): OpenAI.Chat.Completions.ChatCompletionContentPart[] {
-  // Check if the markdown content exists in the document
-  if (!document.markdown) {
+): [OpenAI.Chat.Completions.ChatCompletionContentPart[], number] {
+  let markdown = document.markdown;
+
+// Check if the markdown content exists in the document
+  if (!markdown) {
     throw new Error(
       "Markdown content is missing in the document. This is likely due to an error in the scraping process. Please try again or reach out to [email protected]"
     );
   }
 
-  return [{ type: "text", text: document.markdown }];
+  // count number of tokens
+  const numTokens = numTokensFromString(document.markdown, "gpt-4");
+
+  if (numTokens > maxTokens) {
+    // trim the document to the maximum number of tokens, tokens != characters
+    markdown = markdown.slice(0, (maxTokens * modifier));
+  }
+
+  return [[{ type: "text", text: markdown }], numTokens];
 }
 
 export async function generateOpenAICompletions({
@@ -38,7 +51,7 @@ export async function generateOpenAICompletions({
   temperature?: number;
 }): Promise<Document> {
   const openai = client as OpenAI;
-  const content = prepareOpenAIDoc(document);
+  const [content, numTokens] = prepareOpenAIDoc(document);
 
   const completion = await openai.chat.completions.create({
     model,
@@ -72,6 +85,7 @@ export async function generateOpenAICompletions({
   return {
     ...document,
     llm_extraction: llmExtraction,
+    warning: numTokens > maxTokens ? `Page was trimmed to fit the maximum token limit defined by the LLM model (Max: ${maxTokens} tokens, Attemped: ${numTokens} tokens). If results are not good, email us at [email protected] so we can help you.` : undefined,
   };
 }
 

apps/api/src/lib/entities.ts

@@ -72,6 +72,7 @@ export class Document {
   };
   childrenLinks?: string[];
   provider?: string;
+  warning?: string;
 
   constructor(data: Partial<Document>) {
     if (!data.content) {

apps/api/src/lib/load-testing-example.ts

@@ -0,0 +1,42 @@
+import { scrapWithFireEngine } from "../../src/scraper/WebScraper/single_url";
+
+const delay = (ms: number) => new Promise((resolve) => setTimeout(resolve, ms));
+
+const scrapInBatches = async (
+  urls: string[],
+  batchSize: number,
+  delayMs: number
+) => {
+  let successCount = 0;
+  let errorCount = 0;
+
+  for (let i = 0; i < urls.length; i += batchSize) {
+    const batch = urls
+      .slice(i, i + batchSize)
+      .map((url) => scrapWithFireEngine(url));
+    try {
+      const results = await Promise.all(batch);
+      results.forEach((data, index) => {
+        if (data.trim() === "") {
+          errorCount++;
+        } else {
+          successCount++;
+          console.log(
+            `Scraping result ${i + index + 1}:`,
+            data.trim().substring(0, 20) + "..."
+          );
+        }
+      });
+    } catch (error) {
+      console.error("Error during scraping:", error);
+    }
+    await delay(delayMs);
+  }
+
+  console.log(`Total successful scrapes: ${successCount}`);
+  console.log(`Total errored scrapes: ${errorCount}`);
+};
+function run() {
+  const urls = Array.from({ length: 200 }, () => "https://scrapethissite.com");
+  scrapInBatches(urls, 10, 1000);
+}

apps/api/src/scraper/WebScraper/crawler.ts

@@ -25,7 +25,7 @@ export class WebCrawler {
     initialUrl,
     includes,
     excludes,
-    maxCrawledLinks,
+    maxCrawledLinks = 10000,
     limit = 10000,
     generateImgAltText = false,
     maxCrawledDepth = 10,
@@ -117,7 +117,7 @@ export class WebCrawler {
       const response = await axios.get(this.robotsTxtUrl);
       this.robots = robotsParser(this.robotsTxtUrl, response.data);
     } catch (error) {
-      console.error(`Failed to fetch robots.txt from ${this.robotsTxtUrl}`);
+      console.log(`Failed to fetch robots.txt from ${this.robotsTxtUrl}`);
 
     }
 
@@ -152,7 +152,7 @@ export class WebCrawler {
     inProgress?: (progress: Progress) => void,
   ): Promise<{ url: string, html: string }[]> {
     const queue = async.queue(async (task: string, callback) => {
-      if (this.crawledUrls.size >= this.maxCrawledLinks) {
+      if (this.crawledUrls.size >= Math.min(this.maxCrawledLinks, this.limit)) {
         if (callback && typeof callback === "function") {
           callback();
         }
@@ -176,14 +176,14 @@ export class WebCrawler {
       if (inProgress && newUrls.length > 0) {
         inProgress({
           current: this.crawledUrls.size,
-          total: this.maxCrawledLinks,
+          total: Math.min(this.maxCrawledLinks, this.limit),
           status: "SCRAPING",
           currentDocumentUrl: newUrls[newUrls.length - 1].url,
         });
       } else if (inProgress) {
         inProgress({
           current: this.crawledUrls.size,
-          total: this.maxCrawledLinks,
+          total: Math.min(this.maxCrawledLinks, this.limit),
           status: "SCRAPING",
           currentDocumentUrl: task,
         });
@@ -324,6 +324,12 @@ export class WebCrawler {
       // ".docx",
       ".xlsx",
       ".xml",
+      ".avi",
+      ".flv",
+      ".woff",
+      ".ttf",
+      ".woff2",
+      ".webp"
     ];
     return fileExtensions.some((ext) => url.endsWith(ext));
   }