Skip to main content

Collect web pages

Set a starting URL and collection scope on a knowledge detail screen to collect web pages. After processing finishes, collected content is available to search and AI Chat.

The crawler collects both static HTML and JavaScript-rendered pages. Collected content is divided with the configured method and registered in the search index.

Create a web-crawling job

On the Documents tab of a knowledge detail screen, select Add document → Web crawling.

Screenshot TODO

Capture: Web-crawling creation screen showing basic information, crawling strategy, exploration scope, and chunking options. Use docs.example.com as the URL.

Basic information

FieldRequiredDescription
NameRequiredEnter the document identifier used in metadata and the API.
AliasOptionalEnter the name shown in the list. If blank, it is set to the name.
URLRequiredEnter the web-page address where collection begins, for example https://docs.example.com.
DescriptionOptionalEnter notes about the crawling job.
TagsOptionalEnter tags used for classification and search.

After entering a name, select an AI generation button for the alias, description, or tags to display a suggestion. Review and revise it as needed.

Crawling options

After entering basic information, expand Crawling and indexing options to configure collection scope.

Crawling strategy

StrategyDescriptionUse when
Breadth first (BFS)Collects pages at the same depth before moving to the next depth.Collect several sections to the same depth. This is the default.
Depth first (DFS)Explores one path to its end before moving to the next path.Collect a particular child path first.
Best first (BEST_FIRST)Collects pages in descending link-importance score.Collect pages in score order.

Exploration scope

OptionDefaultRangeDescription
Maximum depth31–10Sets the link depth explored from the starting URL.
Maximum pages1001–10,000Sets the maximum number of pages collected.

Additional options

OptionDefaultDescription
Use sitemapEnabledUses the URL list from sitemap.xml for collection.
Respect robots.txtEnabledSkips paths disallowed by robots.txt.
Exclude external linksEnabledSkips links on domains other than the starting URL.

Indexing options

Select the search methods used to index collected content.

Search modeDescription
Semantic only (VECTOR)Finds similar content with semantic vectors.
Keyword only (TEXT)Finds content containing the query term.
Hybrid (HYBRID)Combines semantic and keyword search results.

See Chunking and search options for selection guidance. Compare results for the same query after processing.

Chunking options

These settings divide crawled content into searchable units. Expand Chunking options to review current values. See Chunking and search options for strategy and value selection guidance.

Monitor a crawling job

After the job starts, review progress on its detail screen. While active, status refreshes every five seconds.

Job states

StateDescription
ProcessingCrawling is in progress.
PausedA user paused the job.
CompleteCrawling and indexing have finished.
FailedCrawling stopped because of an error.
CanceledA user canceled the job.

Controls by state

Different controls appear in the header depending on job state.

Current stateAvailable buttonsBehavior
ProcessingPause / CancelPause or stop active crawling.
PausedResume / CancelResume or cancel paused crawling.
CompleteCrawl againStart crawling again with the same settings.
FailedCrawl againRetry failed crawling.
CanceledCrawl againRestart canceled crawling.

Progress display

During crawling, an indexing status card appears at the top of the detail page. It contains the current state and progress message and refreshes automatically every five seconds.

When crawling finishes, close the status card and review the generated chunks.

Chunk list

After crawling finishes, review generated chunks in the paginated table.

ColumnDescription
#Chunk sequence number
TypeChunk type (TEXT, IMAGE)
ContentChunk-content preview

Select a table row to open the Chunk details panel. Review the complete contents and move to the previous or next chunk.

Manage the document

On the job detail page, edit the description of the crawled document or delete the document.

  • Edit description: Add or change the document description.
  • Delete document: Permanently deletes the document and all associated chunks and embeddings.
Deletion warning

Deleting the document permanently removes all its chunks and embeddings. This action cannot be undone.

Next steps