Skip to content

Feature Parity

DasCondor edited this page Nov 15, 2026 · 2 revisions

Feature parity

Documentation snapshot from docs/FEATURE_PARITY.md at commit 20d7c81. Check the current source for updates.

KubeFactory · All docs · 2.0.0-beta.2 pre-release · Known limitations

Two questions: does 2.0 still do everything 1.0.1 did, and how much of the 2.0 handoff is in?

Status words, used the same way everywhere:

  • verified live — run against real infrastructure
  • verified on a real API server — run against the Kubernetes project's own kube-apiserver 1.32 and etcd, started for the test (tools\live-test.ps1). It is a real control plane with nothing behind it: no nodes that run containers, no controllers. API behaviour is real; nothing that needs a running container is covered
  • verified in simulation — run against the built-in simulation, a fake API server or a local test page
  • not tested — implemented, never run against the real thing
  • unsupported — not implemented

1.0.1 features in 2.0

Feature 2.0 Status
Simulated clusters, four demo worlds, browser demo unchanged verified in simulation
Tools 1-6, tiers, typed confirmations, protected namespaces unchanged; same code path on every backend verified in simulation
Scale, restart, delete, deploy, cordon, drain kubectl path unchanged; native path added native path: verified on a real API server; scale, restart, delete and drain with pods also verified on a kind cluster with running containers (29/29). kubectl path: not re-tested this cycle
Logs and events native path added events: verified on a real API server. Logs: verified on a kind cluster with running containers
Porter: app setups, Portainer templates unchanged; applies through server-side apply on the native path native apply: verified on a real API server (the game deployed and removed the postgres setup and a Portainer template)
Bob and Bob 2.0 log reading unchanged verified in simulation
Multi-user presence and action feed native path added verified on a real API server (one player; two players at once not tested)
Docker engine edition unchanged (docker CLI), plus usage figures backend verified live on a real container (logs, storage, usage); the edition's full tour not re-run
Graphics presets and benchmark unchanged verified in simulation
Audit log, log providers, sound, datapad, codex unchanged verified in simulation

Nothing was removed. --kubectl-cli selects the 1.0.1 Kubernetes path.

2.0 handoff

Section Item Status
00 Native connectivity Kubeconfig discovery, import by file / drop / paste, context list verified in simulation (real kubeconfig file, fake API server)
Live on a real cluster with running containers (kind 1.37): logs, storage, port-forward, scale, restart, drain with pods verified live (tools\live-test.ps1 -Kind, 29/29; the game also run on it)
Connect without kubectl (client-go) verified on a real API server and on kind with working nodes; not tested on any cloud provider
Plain failure screens (unreachable, TLS, auth, RBAC, timeout) RBAC and unreachable: verified on a real API server; TLS, auth: verified in simulation
Explicit trust for exec auth helpers verified in simulation
Watches instead of polling verified on a real API server
Clean Windows machine acceptance test not tested
00A Assets Model library integrated at three tiers verified in simulation
Clipping: V6 library, every display ray-tested clear verified in simulation; audit in MODEL_AUDIT.md. Solid-on-solid overlap: not tested
Type plate, real resource name and container list on the model's own surfaces verified in simulation
V6 handheld tools in first person, display facing the operator with real target data verified in simulation
Krew display surfaces; rigging and animation of tools and Krew unsupported
Asset reviewer hosted at model-viewer/ deployed to GitHub Pages; verified locally
Opening storage doors, stopped by obstacles verified in simulation
Read-only storage browser verified live on kind and on a real Docker container; links are never followed
Porter V3 as minimum standard machines, Porter and Bob use V3. One exception by the operator's choice: the Service pipelines use the V2 pipe kit (pipe_straight, conduit_elbow, pipe_valve); no other V2 model is placed
02 Docker Native Docker Engine API unsupported (docker CLI kept)
Several Docker hosts (CLI contexts), read-only by default for other machines, read-only proxy recipe verified live through a throw-away read-only proxy: state, docker inspect, the image scan, and the host refusing a removal with read-only off. The SSH tunnel and a separate physical host not tested
Komodo Core, read-only: servers and containers (2.3 request name, with a fallback for older Cores) verified against a mock Komodo built from the komodo_client request and entity shapes (agent tests, in-game test on both renderers). The read requests were checked against a real Core 2.3.4 by a contributor, with the old request name; the game has not been played against a real Core
02 Kubernetes Server-side apply, cordon, scale, rollout restart verified on a real API server
Drain with pods to evict verified live on the disposable kind cluster
Exec / terminal, ephemeral debug containers, Helm, CRD browsing unsupported
03 World and graphics Non-regression presets verified in simulation (BENCHMARK.md)
04 Assets and operator experience In-game settings: every key and mouse button rebindable, controller support, field of view, interface size, colour-blind palettes, minimap verified in simulation with simulated controller input (tests/live/settings_test.gd, minimap_test.gd, both renderers); not tested with a physical controller. Controller buttons cannot be rebound
05 LiveView Real page on an in-game screen, mouse and keyboard verified in simulation (local test page)
Endpoint discovery (Docker published ports, Kubernetes Services) verified in simulation (unit tests)
Port-forward for cluster Services verified live (HTTP through a loopback port); LiveView through it not run
Host allowlist, isolated profile verified in simulation
Machine screens step through dashboard, metrics, logs and storage; click to open full size; application icons verified in simulation (desktop and Compatibility renderers)
Metrics view: CPU and memory from the metrics API, all four from docker stats, unavailable shown as such Docker: verified live. Metrics API: only its "no metrics API" answer seen live; real figures not tested
Sort by column and CONTAINS filter on tables (under 10,000 rows) verified in simulation (unit tests and screenshots)
Touch controls for the browser demo (landscape and portrait) verified in simulation with touch events at phone sizes; not tested on a physical iPhone or Android device
Explicit desktop simulation starts without kubeconfig or Docker discovery; simulated actions skip webhooks and online catalogs verified with an instrumented Docker stub and the cluster switcher on both desktop renderers (tools\test-simulation.ps1)
Demo worlds: simulated Grafana dashboard and stream activity on machine screens, marked SIMULATED; nothing of the kind on a real cluster verified in simulation (desktop and Compatibility renderers)
Bundled browser engine (CEF), display builder, audio unsupported: uses installed Edge or Chrome
V8 providers Provider registry (Grafana, Tautulli), protected token storage, test connection verified live: Grafana OSS 10.4.10, 11.2.0, 12.3.0, 13.0.0 (tools\live-test.ps1 -Grafana) and Tautulli 2.19.1 (-Tautulli)
Grafana: dashboard search, panel list, time range, variables, credential-free links, sub-path verified live on Grafana OSS 10.4.10, 11.2.0, 12.3.0 and 13.0.0; not Enterprise or Cloud
Grafana: newer dashboard API and layouts, native charts from a data source unsupported
Tautulli: key handling, failures, missing values shown as unavailable verified live on Tautulli 2.19.1 (no Plex behind it)
Tautulli: real stream counts and bandwidth; the panel pinned to a machine verified in simulation (stand-in server); needs a Plex server to test live
Tautulli: history, home stats, libraries; control-room wall; Datapad pin unsupported
06-08 Krew Porter and Bob with V3 models verified in simulation
Gauge: optional world model, read-only node request and capacity table, on-demand workload usage UI verified in simulation; Docker usage verified live on a disposable container. Kubernetes metrics API figures not tested live
Krew framework and operations center: shared findings with evidence, proposal and risk; resolved history; Markdown and JSON export; approval through the guard rails verified in simulation (both renderers); verified live on Docker and a disposable kind cluster
Sentinel (security), Scout (network), Scrappy (cleanup) with stations verified live: Docker (disposable privileged container with the socket mounted) and kind (privileged host-root pod, a Service selecting nothing); approved removals verified in simulation only
Archie (storage), Forge (build), Blueprint (architect) with stations verified live on Docker with throw-away containers (database image without a volume, unnamed volume, Compose service with a stopped container); Forge's local-build check verified in simulation only, because Docker Desktop's containerd image store does not record image origin. Kubernetes in simulation and unit tests
Gauge forecasts and right-sizing; backups and restores (Archie); build analysis and publishing (Forge); Compose/YAML import and Terraform plans (Blueprint); incident coordination unsupported
11 Multiplayer Server mode: players connect to one KubeFactory server with tokens; roles and read-only enforced by the server verified in simulation: a headless server and client games on both renderers (sign-in, wrong token, viewer refused, operator restart). TLS (wss://) and a second machine not tested
08 Extensibility Backend plugin API 1 (out-of-process, trusted by hash, read-only unless allowed) with an example verified in simulation with the example plugin; input checking unit-tested
09 AI Providers (Ollama, LM Studio, OpenAI, Claude, OpenAI-compatible), redacted bounded context, typed proposals checked against a tool registry and the state, approval through the guard rails verified with a mock provider (agent unit tests for both API shapes; in-game test on both renderers). Not tested against any real provider
Read tools on demand, multi-step plans, streaming, cost limits beyond the per-request bounds, scheduled or autonomous actions unsupported
09A Foreman Six identities (rules, Ollama, LM Studio, OpenAI, Claude, custom) mapped to their models, drawn fallback in the browser build; rules-only incidents and brief verified in simulation (both renderers)
10 Autonomy Policy engine, self-healing, scheduling unsupported
11 Security No credentials in logs, Git or shipped files; browser demo isolated verified by review and by two tests (TestInspectListsContextsWithoutSecrets, TestBadKubeconfigErrorsDoNotEchoContents)
Multiplayer beyond 1.0.1 presence unsupported
12 Verification Unit tests (GDScript, Go), model audit, screenshot runs done; the tests are in tests/ and agent/*_test.go
Live acceptance tests, large-scale benchmarks kind and Docker tests completed; clean-machine, cloud and large-scale tests not completed
13 Distribution Windows, Linux and macOS packages with companion and models published in 2.0.0-beta.1 and 2.0.0-beta.2; packages have SHA-256 checksums and build-provenance attestations
Platform code signing macOS is ad-hoc signed; publisher signing and macOS notarization are unavailable

Remaining verification before a stable release

Tracked, with what blocks 2.0.0 and what can ship as a known limitation, in GO_LIVE_2.0.md.

Clone this wiki locally