General
Crawling Strategy
A comprehensive guide to Glean’s crawling system, detailing how it optimizes data retrieval while respecting API limits and source application performance.
Overview
The Glean crawling system implements a sophisticated approach to data retrieval and indexing. The system is built around two primary optimization goals:- Minimizing latency when incorporating updates from source applications
- Maintaining API call volumes within acceptable limits to prevent source application overload
Configuration Flexibility
All crawling frequencies described in this documentation represent default settings based on customer feedback. These values can be customized to meet specific organizational needs.
API Call Rate Management
API Call Rate Management
Administrators can configure:
- The rate of API calls per second
- The number of concurrent API calls
- Dynamic exponential backoff parameters for handling overload scenarios
Time-Based Controls
Time-Based Controls
The system supports granular scheduling with different rates for:
- Peak operational hours
- Off-peak periods
- Specific days of the week
- Custom time windows
Core Concepts
A comprehensive process that indexes the entire corpus of a datasource. These crawls are scheduled at regular intervals to ensure complete dataset accuracy in the search index.
An efficient update strategy that focuses on modified or newly added content since the previous crawl, optimizing resource usage by avoiding full repository scans.
Organizational information about individuals, encompassing names, titles, email addresses, departmental affiliations, and other relevant attributes.
A specialized process for retrieving and updating identity-related information across various datasources.
A continuous monitoring process that tracks and indexes specific changes within a datasource, including content additions, updates, deletions, and permission modifications.
The frequency at which the system performs incremental fetches to update or refresh data from different sources to ensure that the latest information is available.
Data Refresh Rates By Category
These tables present the complete refresh rate information for all supported datasources. Time periods are denoted as:
- d: days
- h: hours
- m: minutes
Communication & Collaboration Platforms
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| GChat | 28d | 5m | 30m | N/A | N/A | N/A | |
| Microsoft Teams | 30d | 1h | 1h | N/A | N/A | N/A | |
| Slack | 28d | 3h | 1h | N/A | <5m | webhook | Incremental crawl addresses cases where webhook wasn’t delivered |
| Slack Enterprise | 28d | N/A | 1h | 10m | 5m | N/A | SEG uses activity crawl to find channels with recent activity, then queues incremental crawls |
| Yammer | 1h | N/A | 1h | N/A | N/A | N/A |
Document Management & Storage
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Box | 28d | 10m | 1h | 1m | 10m | API | Uses Events API to identify new/modified/deleted docs |
| GDrive | 28d | 3h | 1h | 1m | 10m | API | Uses Reports API for 10-minute updates. Activity reports every 10m, with 12h recrawl for missed events |
| OneDrive/SharePoint | 28d | 1h | 10m | 10m | 10m | API | Uses User Insights API for 10-minute updates with hourly incremental fetch |
| Notion | 6h | N/A | 1h | N/A | N/A | N/A | |
| Quip | 28d | 10m | 10m | 10m | N/A | API |
Development & Project Management
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Asana | 7d | 10m | 1h | 10m | 10m | webhook | |
| Bitbucket | 28d | 10m | 1h | N/A | N/A | N/A | |
| GitHub | 28d | 10m | 10m | N/A | N/A | webhook | |
| GitLab | 28d | 10m | 1h | N/A | <5m | webhook | PR/issue/comment updates trigger immediate crawls; 10m incremental crawls |
| Jira | 7d | 3h | 10m | 30m | <5m | webhook | Webhooks for issue/comment modifications |
| Trello | 1d | N/A | 30m | N/A | N/A | webhook | Requires Jira or Confluence setup |
HR & People Management
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Azure | N/A | N/A | 1h | N/A | N/A | N/A | Hourly people data crawl with additional hour for indexing |
| BambooHR | N/A | N/A | 1h | N/A | N/A | N/A | Hourly people data crawl with additional hour for indexing |
| Greenhouse | 28d | 1h | 10m | N/A | N/A | N/A | |
| Lattice | 1d | N/A | 1h | N/A | N/A | N/A | Hourly people data crawl with additional hour for indexing |
| Lever | 1h | 1h | 10m | N/A | N/A | N/A | |
| People Data API | N/A | N/A | 1h | N/A | 1h | API | Real-time updates via push API |
| Pingboard | N/A | N/A | 1h | N/A | N/A | N/A |
Productivity & Knowledge Management
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Aha | 1d | 1h | 1h | N/A | N/A | N/A | |
| Airtable | 1d | 1h | 1h | N/A | N/A | N/A | |
| Coda | 7d | N/A | 1h | N/A | N/A | N/A | |
| Confluence | 7d | 1h | 10m | N/A | 5m | webhook | New DC versions use webhooks for 5m updates; hourly fetch for older versions |
| Guru | 1d | 10m | 1h | 10m | 1h | N/A | |
| Miro | 1d | 1h | 30m | 10m | 1m | N/A | |
| SmartSheet | 1h | 10m | 10m | N/A | N/A | N/A |
Business Applications
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Salesforce | 28d | 10m | 1h | N/A | 10m | N/A | |
| ServiceNow | 3d/30m* | 1h | 1h | 30m | 3d | N/A | *3d for Knowledge Articles, 30m for Catalog Items |
| Zendesk | 28d | 1h | 1h | N/A | 1h | N/A | |
| Freshservice | 1h | N/A | 10m | N/A | N/A | N/A | |
| Monday.com | 1h | N/A | 10m | N/A | N/A | N/A | |
| Pager Duty | 1h | 10m | 10m | 10m | N/A | API |
Web & Content Management
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Brightspot | 1h | N/A | 10m | N/A | N/A | N/A | |
| Google Sites | 1d | 4h | N/A | N/A | N/A | N/A | |
| Web pages | 1d | N/A | N/A | N/A | 1d | N/A | Default daily crawl; fully configurable |
| Wordpress | 12h | 10m | N/A | N/A | N/A | N/A |
Additional Services
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Docebo | 1h | N/A | 1h | N/A | N/A | N/A | |
| Egnyte | 1d | 10m | 1h | N/A | N/A | N/A | |
| Fifteen Five | 1h | N/A | 1h | N/A | N/A | N/A | |
| Figma | 30m | N/A | 30m | N/A | N/A | N/A | |
| Gmail | N/A | N/A | N/A | N/A | 0m | N/A | Real-time updates via federated API |
| Gong | 1h | N/A | 1h | N/A | N/A | N/A | |
| Google Groups | 7d | 1d | N/A | N/A | N/A | N/A | |
| Highspot | 4h | N/A | 1h | N/A | N/A | N/A | |
| Looker | 1h | N/A | 1h | N/A | N/A | N/A | External connector |
| LumApps | 1h | N/A | 10m | 10m | 1m | N/A | Hourly document fetch with corpus size-dependent processing |
| Okta | 3h | N/A | 3h | 1h | N/A | N/A | Additional hour for people data indexing |
| Seismic | 28d | 1d | 10m | 1d | N/A | API | |
| Simpplr | 28d | 1h | 1h | N/A | N/A | N/A | |
| Stack Overflow | 2h | N/A | 1h | N/A | 3h | N/A | |
| Tableau | 1d | 45m | 45m | N/A | N/A | N/A |
API Services
| Platform | Full Crawl | Incremental | People Data | Activity | Update Rate | Webhook | Notes |
|---|---|---|---|---|---|---|---|
| Push API for Content | N/A | N/A | N/A | N/A | N/A | N/A | Customer controls update frequency via bulk upload APIs and real-time CRUD operations |
Content Deletion Handling
Glean handles content deletion through two primary mechanisms, depending on the capabilities of the source application.
API/Webhook Deletion
For applications that provide deletion notifications through APIs or webhooks, content is removed from the index immediately upon notification.
Full Crawl Cleanup
For applications without deletion notifications, stale content is identified and removed during scheduled full crawls.
Deletion of derived information used in models and other auxiliary systems is governed by Glean’s privacy policy.
Support for Real-time Updates
Webhook Integration
Many modern platforms support webhook-based updates, enabling near real-time content synchronization. For example:
- GitHub/GitLab: Updates within 5 minutes of changes
- Slack: Near immediate updates via webhook
- Confluence: 5-minute update window for newer versions
For datasources that don’t support webhooks, Glean implements regular polling with optimized intervals to maintain data freshness while respecting API limits.