Indexterity

Index dexterity for MongoDB, PostgreSQL and SQL Server. A service that watches your indexes and manages them safely — drops the unused and redundant, merges overlapping, extends prefixes, creates the missing — and proves the result in freed bytes and latency.

Read-only by default. The one irreversible step, a drop, is gated behind an observe window, a pre-flight check and a read-latency regression test.

Pages

• Architecture — how the system is put together: the analysis engine, the apply pipeline, engine ports, security, the data model, the dashboard, deployment and metrics.
• Plans and policy — the per-cluster knobs, how the observe window scales to the index, how a recommendation is scored, and what each plan entitles an organization to.
• Connecting a cluster — the scoped user Indexterity needs on each engine, the exact createRole and GRANT snippets for analyze-only and live-manage, and which replicas get read.
• Running Indexterity — the development loop, how a release is cut, the images and the three topologies, every environment variable that matters and why, the metrics worth alerting on, and what error reporting sends.

Not here, on purpose

• README — what the product is, in about a minute. It stays deliberately short: it is the page GitHub shows first, and a wall of text there is a wall in front of the door. Everything it used to carry is on the pages above.
• Decision log — every load-bearing choice and whether it is still open. In the repository so a decision is reviewed alongside the change that makes it, and so it can be git blamed.
• Helm chart — its README ships inside the chart artifact, so it lives with the chart.
• Project board — what is planned. Not duplicated here: two roadmaps in one repo is one roadmap and one lie.