Agent skill · product strategy · jeremylongshore
glean-reference-architecture
Enterprise architecture: Source Systems to Connectors (Cloud Run/Lambda,
Why this skill is useful
Adds a comprehensive architecture design and executable code snippets for integrating various source systems with Glean's indexing API, which the AI wouldn't generate on its own.
What it needs
Requires curl, npm installed locally. About 4k tokens when loaded. Last updated 2026-08-07. 2,604 stars on the source repository.
What this skill does
Glean Reference Architecture Overview Enterprise search integration architecture for connecting internal knowledge systems to Glean's indexing and search platform. Designed for organizations needing unified search across Confluence, Google Drive, Notion, Slack, Jira, and custom internal tools. Key design drivers: connector reliability for continuous indexing, permission synchronization to enforce source-system ACLs, incremental vs bulk indexing tradeoffs, and low-latency search aggregation across heterogeneous document types. Architecture Diagram Service Layer Caching Strategy Event Pipeline Data Model Scaling Considerations Deploy one connector instance per datasource to isolate failures and rate limits Schedule bulk reindexing during off-peak hours — Glean indexing API has per-datasource throughput limits Use incremental sync (change cursors) for high-frequency sources (Slack, Jira) to minimize API calls Permission sync is the bottleneck — batch ACL updates and run as a separate queue consumer Monitor connector health per datasource; alert on sync lag > 15 minutes for critical sources Error Handling Component Failure Mode Recovery ----------- ------------- ---------- Connector sync Source API rate limit Per-datasource backoff, degrade to hourly bulk sync Document indexing Glean 429 throughput limit Queue retry with jitter, batch size reduction Permission sync ACL mismatch between source and Glean Reconciliation job flags discrepancies for admin review Bulk reindex Timeout on large datasource Checkpoint cursor, resume from last successful batch Search aggregation Stale index for one datasource Degrade gracefully — return results from healthy sources, flag staleness Prerequisites A named data owner for every source, a source-ACL inventory, and an approved classification/retention policy. Separate sandbox, staging, and production identities, credentials, queues, and destinations; production content must never be an architecture test fixture. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills jeremylongshore/glean-reference-architecture