Knowledge, x-ray and insights
The hub for the Knowledge and insights pages: where the platform learns your organization, and where you get evidence back out of it.
Concept · Updated 2026-10-19 · Describes release 0.1.164
What these pages are for
These pages are how the platform learns your organization, and how you get evidence back out of it. Use Agent quality → Knowledge base and Insights → Estate x-ray to build the context agents use, and Chat, Report, Lakehouse, Test Lab and Feedback to ask questions, review outcomes and test behaviour.
Discovery does not write to the connected estate. It can persist reports and a single-connection runbook inside AIOps. Reviewing and using context, enabling collection, or running live tests are distinct actions with their own permissions and consequences, described on each page.
The pages
This chapter became the pages below, in its old order. Every section kept its heading, so a bookmark into the old chapter now lands on one of them.
- Knowledge base — runbooks, reference documents, instructions and the estate model under Agent quality → Knowledge base; the skill catalog per pillar; the Review queue where a correction or an observation becomes a skill (
knowledge.view; decisions needskills.approve). - Estate harvest and x-ray — the deterministic harvest of an AWS, Azure or Kubernetes connection, the read-only Estate x-ray scan, the service graph, the Estate map and how far to trust it (
datasources.view; launching needsdatasources.manage). - Usage and billing — what the agents cost this workspace by calendar month, the five spend classes, the organization view and the CSV statement (
report.view). - Insights chat — questions over the evidence your role may read, with the tool behind each answer (
chat.ask). - Lakehouse — retained telemetry, collection specs and read-only SQL (
datasources.view; specs needdatasources.manage). - Artifacts — saved reports, certifications, pull requests and investigation outputs, append-only (
incidents.view). - Archive — incidents and changes past the retention window, read back read-only and downloadable; the owner's month batches with the fine-tuning export and back-test cases (
incidents.view; batches needmembers.manage). - Service delivery report — operations, remediation outcomes, posture and agent quality over 7, 30 or 90 days, with platform cost as an appendix (
report.view). - Benchmarks and scores — golden datasets, the builtin suite, back-tests on your own history, the gate and what a score means; live fault injection is in Test Lab (
benchmark.view; runs needbenchmark.manage). - Feedback — human ratings by pillar, model and week, and the finished incidents still waiting for a score (
report.view; rating needsfeedback.submit).
Terms are in the Glossary.