A developer building a LlamaIndex pipeline that reads Intercom help centre articles needs clean page text before passing it into a vector index or query engine. Intercom pages mix article content with related article suggestions, chat widget, collection navigation, reaction buttons, breadcrumbs, and “Was this article helpful?” prompts, and scraping them by hand requires CSS selectors that break when Intercom updates its layout. EnConvert Perceive renders up to 10 Intercom pages in a single batch call and returns clean Markdown for each, with navigation and sidebars stripped by default. The extracted Markdown is ready to load into a LlamaIndex VectorStoreIndex with no HTML parsing.
How to Scrape Intercom with LlamaIndex
Scrape Intercom pages into a LlamaIndex pipeline with EnConvert Perceive, returning clean Markdown with navigation and sidebars stripped in a single batch call
Get API keyExample
python
import os
import requests
from llama_index.core import VectorStoreIndex
from llama_index.core.schema import TextNode
# LlamaIndex uses OpenAI for embeddings and answers, and reads OPENAI_API_KEY from the environment
BASE_URL = "https://api.enconvert.com/v2"
headers = {
"X-API-Key": os.environ["ENCONVERT_API_KEY"],
"Content-Type": "application/json",
}
page_urls = [url.strip() for url in os.environ["INTERCOM_ARTICLE_URLS"].split(",")] # public Intercom help article URLs
if len(page_urls) > 10:
raise ValueError("This example covers inline batches of 10 URLs or fewer")
response = requests.post(f"{BASE_URL}/perceive/batch", headers=headers,
json={"urls": page_urls, "options": {"outputs": ["markdown"]}}, timeout=300)
if response.status_code == 402:
raise RuntimeError("Monthly ops allowance exhausted")
response.raise_for_status()
nodes = []
for result in response.json()["items"]:
page_url = result["url"]
if result.get("status") != "completed":
print(f"Skipped {page_url}: {result.get('status')}")
continue
if result.get("is_blocked"):
print(f"Skipped {page_url}: blocked page")
continue
render_quality = result.get("render_quality", 1.0)
if render_quality < 0.40:
print(f"Skipped {page_url}: low render quality ({render_quality:.2f})")
continue
markdown_url = result["outputs"]["markdown"]["url"]
markdown = requests.get(markdown_url, timeout=60)
markdown.raise_for_status()
nodes.append(TextNode(
text=markdown.text,
id_=page_url,
metadata={"source_url": page_url, "render_quality": render_quality},
))
if not nodes:
raise RuntimeError("No Intercom articles passed the render checks")
index = VectorStoreIndex(nodes)
query_engine = index.as_query_engine(similarity_top_k=3)
print(query_engine.query("Summarise what these Intercom help articles cover."))What you get back
Perceive returns clean Markdown for each URL in the batch, with navigation and sidebars stripped by default, delivered as pre-signed URLs that you fetch once per page. Scraping Intercom by hand needs CSS selectors that break when the layout changes, and passing raw Intercom HTML into a LlamaIndex node fills the index with related article links, chat widget markup, and reaction prompts that add noise to every similarity search result.