Skip to content
nowake.ai docs

Roadmap

Status date: 2026-05-28

This page is the consolidated roadmap for kube-insight. Detailed working notes remain in the linked documents; this page is the stable entry point for readers who need to understand what is done, what comes next, and what is intentionally deferred.

Current Position

Section titled “Current Position”

kube-insight is released as a local-first retained evidence tool. The default artifact uses SQLite so the normal install stays small and pure Go. ClickHouse is the primary central evidence backend candidate for long-running append-heavy history. A separate chDB-enabled artifact provides embedded ClickHouse-compatible local storage for users who accept the larger runtime dependency.

The MVP baseline is complete for SQLite default local mode, the chDB-enabled local variant, and the core MCP read surface. The agent-first Web UI foundation adds an embedded chat surface backed by a server-side autonomous agent, typed Kubernetes artifacts, streamed run events, and the existing read API. Kubernetes RBAC support remains the next major security milestone.

Completed Foundation

Section titled “Completed Foundation”
  • Kubernetes discovery, list/watch collection, and offline ingest.
  • Pre-storage filtering with auditable destructive and sensitive filter decisions.
  • Retained object versions, observations, facts, changes, topology edges, and resource health.
  • SQLite default storage with CLI, HTTP API, and MCP read surfaces.
  • ClickHouse append-only evidence schema and write path for central evidence history.
  • ClickHouse read paths for schema, read-only SQL, resource health, search, object history, topology, and service investigation.
  • chDB-enabled local variant that reuses the ClickHouse-compatible schema and read path when built with the optional chDB adapter.
  • MCP tools for backend-aware schema, read-only SQL, health, and retained object history.
  • Validation reports covering storage modes, agent-vs-kubectl workflows, live ClickHouse profiling, and chDB real-data smoke paths.

Next Milestones

Section titled “Next Milestones”

  1. Web UI hardening. Continue the agent-first human operator surface: richer tool-call visibility, dashboard/runtime health, artifact drill-down, resource and topology history travel, and frontend reliability work around retry and session projection.

  2. Kubernetes RBAC support. Add Kubernetes authn/authz-aware service access so API, MCP, and future UI reads can be constrained by the caller’s Kubernetes permissions. The first cut should cover request identity, SubjectAccessReview-style checks for requested resources, auditability, and predictable behavior when historical evidence refers to objects the caller can no longer read live.

  3. Backend hardening. Validate real S3/object-storage cold tiering, keep ClickHouse active and inactive footprint reporting visible, and add repair or materialized read models only where measured query cost requires them.

  4. Measured storage experiments. Benchmark native ClickHouse JSON, selected JSON/text skipping indexes, and any heavier path indexes before making them defaults.

  5. Production readiness. Add in-cluster deployment guidance, retention and redaction operator controls, rebuild/repair jobs, and collector observability suitable for long-running service mode.

Explicitly Deferred

Section titled “Explicitly Deferred”
  • chDB is not promoted to the default local backend until runtime packaging is small and reliable enough for normal installs.
  • chDB-enabled serve --metrics is deferred because the current libchDB runtime is not stable enough for the combined API plus metrics service shape.
  • PostgreSQL and CockroachDB remain possible metadata/control-plane or compatibility backends, not the next primary evidence store.
  • Broad synchronous cold JSON scans are not an MVP default; investigation should use facts, changes, topology edges, retained versions, and measured indexes.

Roadmap Documents

Section titled “Roadmap Documents”