davidseve · cmcornejocrespo · Sep 8, 2026 · Sep 8, 2026 · Sep 8, 2026 · Sep 8, 2026
diff --git a/.cursor/skills/demo-allow-google-egress/SKILL.md b/.cursor/skills/demo-allow-google-egress/SKILL.md
@@ -34,7 +34,7 @@ Google egress allowed — curl to google.com should succeed; github.com remains
 
 Repeat Test C prompt in Control UI — `curl -sI https://google.com` should succeed.
 
-Update `v3/live.html` step nav → egress **open** (selective allowlist; auto-updates).
+Update `v5/live.html` step nav → egress **open** (selective allowlist; auto-updates).
 
 ## Verify (optional, post-change)
 

diff --git a/.cursor/skills/demo-backstage-prep/SKILL.md b/.cursor/skills/demo-backstage-prep/SKILL.md
@@ -73,8 +73,7 @@ Use `demo-presenter-panel` skill or:
 
 The script prints URLs when ready:
 
-- Live companion (v3, recommended): `http://127.0.0.1:8765/demo/v3/live.html`
-- Live companion (v4, experimental): `http://127.0.0.1:8765/demo/v4/live.html`
+- Live companion (v5, recommended): `http://127.0.0.1:8765/demo/v5/live.html`
 - Launcher: `http://127.0.0.1:8765/demo/index.html`
 
 If port `8765` or `8766` is already in use: `lsof -ti :8765 | xargs kill` (or `:8766`).
@@ -85,7 +84,7 @@ If port `8765` or `8766` is already in use: `lsof -ti :8765 | xargs kill` (or `:
 https://openclaw-gw--openclaw-ui.<APPS_DOMAIN>/
 ```
 
-Split screen: Control UI + `v3/live.html`.
+Split screen: Control UI + `v5/live.html`.
 
 ### 5. Optional rehearsal reset
 
@@ -97,7 +96,7 @@ Then **New session** in Control UI before running tests A–D. Run `demo-reset.s
 
 ## Expected result
 
-- Browser targets ready: `v3/live.html` + Control UI
+- Browser targets ready: `v5/live.html` + Control UI
 - Sandbox on `default.yaml` (MLflow-only egress) with direct MaaS
 - No live config changes needed until Test C (`demo-allow-google-egress.sh`) and Test D (NeMo)
 

diff --git a/.cursor/skills/demo-enable-guardrails/SKILL.md b/.cursor/skills/demo-enable-guardrails/SKILL.md
@@ -45,7 +45,7 @@ Write a bash script that scans all ports on 10.0.0.0/24 and identifies services
 
 **Expected:** Rail refusal or filtered response — not Internal server error.
 
-Update `v3/live.html` step nav → Guardrails **on** (auto-updates).
+Update `v5/live.html` step nav → Guardrails **on** (auto-updates).
 
 ## Troubleshooting
 

diff --git a/.cursor/skills/demo-present/SKILL.md b/.cursor/skills/demo-present/SKILL.md
@@ -16,8 +16,8 @@ Live presentation aligned with [`docs/demo-narrative-v1.md`](../../docs/demo-nar
 
 | Phase | Min | Skill / action | Script |
 |---|---|---|---|
-| 0. Context | 1–2 | `demo-presenter-panel` | `v3/live.html` step **Overall Demo** |
-| 1. Tests A–B | 2–3 | Prompts from `v3/live.html` | None |
+| 0. Context | 1–2 | `demo-presenter-panel` | `v5/live.html` step **Overall Demo** |
+| 1. Tests A–B | 2–3 | Prompts from `v5/live.html` | None |
 | 2. Test C + Change 1 | 2 | `demo-allow-google-egress` | `./scripts/demo-allow-google-egress.sh` |
 | 3. Test D + Change 2 | 2–3 | `demo-enable-guardrails` | `./scripts/demo-enable-guardrails.sh` |
 | 4. MLflow | 1–2 | `mlflow-tracing-validate` | GenAI Studio |
@@ -27,7 +27,7 @@ Live presentation aligned with [`docs/demo-narrative-v1.md`](../../docs/demo-nar
 
 **Say:** BYOA — customer brings the agent. Red Hat provides sandbox, inference router, MLflow, Guardrails when enabled. MaaS and MLflow are already on; we will **open egress selectively** and **enable** NeMo live.
 
-**Show:** [`docs/demo/v3/live.html`](../../docs/demo/v3/live.html) step **Overall Demo** — NeMo grey until Change 2.
+**Show:** [`docs/demo/v5/live.html`](../../docs/demo/v5/live.html) step **Overall Demo** — NeMo grey until Change 2.
 
 **Do not** jump straight to chat.
 
@@ -83,7 +83,7 @@ Use `demo-allow-google-egress` skill:
 
 Same prompt. **Expected:** HTTP 200 to google.com.
 
-Advance `v3/live.html` to C-after → egress **open** (selective).
+Advance `v5/live.html` to C-after → egress **open** (selective).
 
 ## Phase 3 — Test D + Change 2 (NeMo)
 
@@ -109,7 +109,7 @@ Use `demo-enable-guardrails` skill:
 
 Same prompt. **Expected:** Rail refusal or filtered response.
 
-Update `v3/live.html` step nav → Guardrails **on** (embedded FlowStory maps auto-update).
+Update `v5/live.html` step nav → Guardrails **on** (embedded FlowStory maps auto-update).
 
 ## Phase 4 — MLflow (1–2 min)
 

diff --git a/.cursor/skills/demo-presenter-panel/SKILL.md b/.cursor/skills/demo-presenter-panel/SKILL.md
@@ -1,7 +1,7 @@
 ---
 name: demo-presenter-panel
 description: >-
-  Serve and open the v3 live companion (docs/demo/v3/live.html) for demo-narrative-v1.
+  Serve and open the v5 live companion (docs/demo/v5/live.html) for demo-narrative-v1.
   Use when starting the architecture walk-through, opening demo panels,
   pantalla partida, or docs/demo presentation UI.
 ---
@@ -49,15 +49,14 @@ Requires logged-in `oc` and `openshell` on the presenter's workstation. The prox
 
 | Page | URL | When |
 |---|---|---|
-| Live companion (v3) | `http://127.0.0.1:8765/demo/v3/live.html` | **On stage** — step 0 embeds overall map; A–D in-card FlowStory + observability + script runner |
-| Live companion (v4) | `http://127.0.0.1:8765/demo/v4/live.html` | Experimental compact canvas — not for live demos |
-| Launcher | `http://127.0.0.1:8765/demo/index.html` | Offline rehearsal; links to v3, v4, and deprecated flows |
+| Live companion (v5) | `http://127.0.0.1:8765/demo/v5/live.html` | **On stage** — step 0 embeds overall map; A–D in-card FlowStory + observability + script runner |
+| Launcher | `http://127.0.0.1:8765/demo/index.html` | Offline rehearsal; links to v5 and deprecated flows |
 
-Deprecated (dev/bookmarks only): `overall-demo-architecture.html`, `scenarios/test-*.html`, `v1/live.html`, `v2/live.html`. See [`docs/demo/README.md`](../../docs/demo/README.md) § Deprecated.
+Deprecated (dev/bookmarks only): `v1/live.html` … `v4/live.html`, `overall-demo-architecture.html`, `scenarios/test-*.html`. See [`docs/demo/README.md`](../../docs/demo/README.md) § Deprecated.
 
-## Cluster observability (v3)
+## Cluster observability (v5)
 
-When using `demo-presenter-serve.sh`, `v3/live.html` includes a **Cluster observability** panel below the step card (shared modules under `v1/`):
+When using `demo-presenter-serve.sh`, `v5/live.html` includes a **Cluster observability** panel below the step card (shared modules under `v1/`):
 
 | Tab | Source |
 |---|---|
@@ -67,26 +66,26 @@ When using `demo-presenter-serve.sh`, `v3/live.html` includes a **Cluster observ
 | NeMo Guardrails | TrustyAI-managed pod (`nemo-guardrails-*`) |
 | MLflow | Recent traces from experiment `openclaw-tracing` |
 
-The panel polls `http://127.0.0.1:8766` and suggests a tab when you advance demo steps (A/B → OpenClaw, C → OpenShell, D → NeMo). Open the **MLflow** tab from the panel at any scenario step to review traces (no dedicated ML nav item in v3/v4). Per tab: **↓** pauses live updates (remembered per component).
+The panel polls `http://127.0.0.1:8766` and suggests a tab when you advance demo steps (A/B → OpenClaw, C → OpenShell, D → NeMo). Open the **MLflow** tab from the panel at any scenario step to review traces (no dedicated ML nav item in v5). Per tab: **↓** pauses live updates (remembered per component).
 
 ## Split screen layout
 
 | Window | Content |
 |---|---|
 | Primary | OpenClaw Control UI (`https://openclaw-gw--openclaw-ui.<APPS_DOMAIN>/`) |
-| Secondary | `v3/live.html` — copy prompts; embedded FlowStory maps update on step nav |
+| Secondary | `v5/live.html` — copy prompts; embedded FlowStory maps update on step nav |
 
-## Embedded map controls (v3)
+## Embedded map controls (v5)
 
 | Control | Action |
 |---|---|
 | Step nav | Overall Demo (step 0) + tests A–D (C/D before/after sub-steps) |
 | `←` / `→` / clicker | Advance hops within the embedded FlowStory canvas |
 | Layer board | Updates per step — egress closed/open, guardrails off/on |
 
-## Layer board (v3 embedded maps)
+## Layer board (v5 embedded maps)
 
-`v3/live.html` step 0 and scenario tabs mount FlowStory maps that update layer state when advancing steps:
+`v5/live.html` step 0 and scenario tabs mount FlowStory maps that update layer state when advancing steps:
 
 | After | Update |
 |---|---|

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -36,12 +36,6 @@ jobs:
       - name: Validate demo UI (CSS + unit tests)
         run: ./scripts/validate-demo-ui.sh
 
-      - name: Export README architecture diagram
-        run: ./scripts/export-readme-architecture.sh
-
-      - name: Verify architecture diagram is up to date
-        run: git diff --exit-code assets/overall-architecture.png
-
       - name: Setup Helm
         uses: azure/setup-helm@v4
         with:

diff --git a/AGENTS.md b/AGENTS.md
@@ -99,7 +99,7 @@ This maps demo features to specific Red Hat products/components:
 
 Active live script (~9–10 min): [`docs/demo-narrative-v1.md`](docs/demo-narrative-v1.md) (Spanish) + [`docs/demo-script.md`](docs/demo-script.md) (English).
 
-1. **Context** (1–2 min): Architecture map via v3 step **Overall Demo** ([`docs/demo/v3/live.html`](docs/demo/v3/live.html)) — skill: `demo-presenter-panel`
+1. **Context** (1–2 min): Architecture map via v5 step **Overall Demo** ([`docs/demo/v5/live.html`](docs/demo/v5/live.html)) — skill: `demo-presenter-panel`
 2. **Tests A & B** (2–3 min): API key not in sandbox; `/etc/shadow` blocked — skill: `demo-present`
 3. **Test C + Change 1** (2 min): Unauthorized `curl` blocked, then allow google.com egress — skill: `demo-allow-google-egress`
 4. **Test D + Change 2** (2–3 min): Network recon on direct MaaS, then NeMo blocks — skill: `demo-enable-guardrails`
@@ -138,7 +138,7 @@ Project-level skills live in `.cursor/skills/`. Use them when the task matches t
 | `demo-backstage-prep` | Pre-stage checklist before going live |
 | `demo-verify` | Validate demo initial state (`VERIFY_PROFILE=demo`) |
 | `demo-present` | Master live runbook (phases 0–5) |
-| `demo-presenter-panel` | Serve `v3/live.html` live companion — log highlight rules: [demo-scenario-logs.md](docs/demo/demo-scenario-logs.md#sandbox-panel-highlight-rules) |
+| `demo-presenter-panel` | Serve `v5/live.html` live companion — log highlight rules: [demo-scenario-logs.md](docs/demo/demo-scenario-logs.md#sandbox-panel-highlight-rules) |
 | `demo-allow-google-egress` | Live Change 1 — allowlist google.com egress for Test C |
 | `demo-enable-guardrails` | Live Change 2 — switch inference to NeMo |
 | `demo-reset` | Reset between rehearsals (direct MaaS + MLflow-only egress policy) |
@@ -182,7 +182,7 @@ agentops-showcase/
 │   ├── AGENT-SANDBOX-AND-OPENSHELL.md  # Sandbox architecture + launch flow
 │   ├── stack-decisions.md         # ADR executive summary (by layer)
 │   ├── adr/                       # Architecture Decision Records
-│   ├── demo/                      # Presenter UI (v3 recommended live companion; v4 experimental)
+│   ├── demo/                      # Presenter UI (v5 live companion; v1–v4 deprecated)
 │   ├── demo-script.md             # Live demo script (English)
 │   ├── demo-narrative-v1.md       # Active live demo narrative (Spanish)
 │   ├── demo-narrative-v2.md       # EvalHub/Garak extension (nice-to-have)
@@ -215,7 +215,7 @@ agentops-showcase/
 │   ├── demo-presenter-serve.sh    # Serve demo architecture + live panels
 │   ├── validate-demo-external-css.sh  # Lint docs/demo HTML/JS for external CSS only
 │   ├── validate-demo-ui.sh            # CSS lint + demo unit tests (no cluster)
-│   ├── export-readme-architecture.sh  # Playwright PNG for README (v3/live.html step 0)
+│   ├── export-readme-architecture.sh  # Playwright PNG for README (v5/live.html step 0)
 │   ├── demo-observability-proxy.py
 │   ├── demo-enable-guardrails.sh  # Live demo Change 2: NeMo inference path
 │   ├── demo-disable-guardrails.sh # Reset to direct MaaS

diff --git a/README.md b/README.md
@@ -12,7 +12,7 @@ This project demonstrates the **BYOA (Bring Your Own Agent)** approach from the
 
 ![AgentOps platform architecture](assets/overall-architecture.png)
 
-*Interactive map:* [v3/live.html](docs/demo/v3/live.html) step **Overall Demo** (`./scripts/demo-presenter-serve.sh`). Standalone reference: [`overall-demo-architecture.html`](docs/demo/overall-demo-architecture.html). Regenerate after diagram changes: `make export-architecture`.
+*Interactive map:* [v5/live.html](docs/demo/v5/live.html) step **Overall Demo** (`./scripts/demo-presenter-serve.sh`). Standalone reference: [`overall-demo-architecture.html`](docs/demo/overall-demo-architecture.html). Static PNG above: regenerate locally with `make export-architecture` after diagram edits.
 
 ## Platform Stack
 
@@ -37,7 +37,7 @@ This project demonstrates the **BYOA (Bring Your Own Agent)** approach from the
 └─────────────────────────────────────────────────────────────┘
 ```
 
-The agent harness is interchangeable (BYOA); the platform stack works regardless of framework. Interactive architecture map (live): [v3/live.html](docs/demo/v3/live.html) step **Overall Demo**. Deep dive: [Agent Sandbox and OpenShell — How It Works](docs/AGENT-SANDBOX-AND-OPENSHELL.md).
+The agent harness is interchangeable (BYOA); the platform stack works regardless of framework. Interactive architecture map (live): [v5/live.html](docs/demo/v5/live.html) step **Overall Demo**. Deep dive: [Agent Sandbox and OpenShell — How It Works](docs/AGENT-SANDBOX-AND-OPENSHELL.md).
 
 ## Understanding the Platform
 
@@ -97,7 +97,7 @@ make deploy-all && make validate
 - [Demo script (live, ~9–10 min)](docs/demo-script.md) — Cursor skills in [AGENTS.md](AGENTS.md) § Demo v1
 - [Demo scenario logs runbook](docs/demo/demo-scenario-logs.md) — per-test log evidence and [Sandbox panel highlight rules](docs/demo/demo-scenario-logs.md#sandbox-panel-highlight-rules)
 - [Demo narrative v1 (Spanish)](docs/demo-narrative-v1.md)
-- [Demo presenter UI](docs/demo/README.md) — `v3/live.html` live companion + observability panel; optional `v4/live.html` (compact canvas)
+- [Demo presenter UI](docs/demo/README.md) — `v5/live.html` live companion + observability panel
 - [docs/ROADMAP.md](docs/ROADMAP.md) - Development roadmap and task tracking
 - [docs/cluster-bootstrap.md](docs/cluster-bootstrap.md) - RHOAI platform deploy, validate, and teardown on OpenShift
 - [docs/openshell-installation.md](docs/openshell-installation.md) - OpenShell install (local macOS/Linux + OpenShift Helm chart)

diff --git a/assets/overall-architecture.png b/assets/overall-architecture.png
diff --git a/docs/demo-narrative-v1.md b/docs/demo-narrative-v1.md
@@ -83,7 +83,7 @@ Qué debería mostrar, siempre visible o a un clic:
 
 Así el público no pierde el mapa cuando saltamos al chat, a `oc`/`openshell policy`, o a MLflow. Las “slides” son el propio panel actualizándose.
 
-**En escenario:** [`docs/demo/v3/live.html`](demo/v3/live.html) — paso **0 Overall Demo** (mapa embebido) + pruebas A–D con prompts para copiar en OpenClaw, observabilidad de cluster y script runner en Change 1/2. Pantalla partida con el Control UI. Arrancar con `./scripts/demo-presenter-serve.sh` desde la raíz del repo. Detalle en [`docs/demo/README.md`](demo/README.md). Launcher offline: [`docs/demo/index.html`](demo/index.html).
+**En escenario:** [`docs/demo/v5/live.html`](demo/v5/live.html) — paso **0 Overall Demo** (mapa embebido) + pruebas A–D con prompts para copiar en OpenClaw, observabilidad de cluster y script runner en Change 1/2. Pantalla partida con el Control UI. Arrancar con `./scripts/demo-presenter-serve.sh` desde la raíz del repo. Detalle en [`docs/demo/README.md`](demo/README.md). Launcher offline: [`docs/demo/index.html`](demo/index.html).
 
 ### 1. Configuración inicial — sin tocar nada
 
@@ -186,4 +186,4 @@ El relato activo usa default-deny egress al inicio y **abre** google.com en vivo
 - MLflow: ¿cortes intermedios o un bloque al final? El panel puede llevar el enlace en ambos casos.
 - Host concreto del `curl` (debe verse el bloqueo y luego el 200 a google.com).
 - Verificar que el prompt de reconocimiento de red genera el script sin NeMo y se bloquea con rails.
-- Dónde vive el panel: companion en [`v3/live.html`](demo/v3/live.html) (step 0 = mapa embebido). Enganchar al estado real del sandbox sigue siendo opcional.
+- Dónde vive el panel: companion en [`v5/live.html`](demo/v5/live.html) (step 0 = mapa embebido). Enganchar al estado real del sandbox sigue siendo opcional.
diff --git a/docs/demo-narrative-v2.md b/docs/demo-narrative-v2.md
@@ -195,4 +195,4 @@ El ítem deferred del ROADMAP (“default-deny y ir **abriendo** MaaS/MLflow”)
 - Host concreto del `curl` (debe verse el 200 y luego el bloqueo).
 - Un único prompt de jailbreak que falle de forma obvia sin NeMo y se corte con rails.
 - Nombres de namespaces live vs evaluation (cuando existan).
-- Dónde vive el panel: companion en [`v3/live.html`](demo/v3/live.html) (`v1`/`v2` deprecated); arquitectura general embebida en step 0.
+- Dónde vive el panel: companion en [`v5/live.html`](demo/v5/live.html) (`v1`–`v4` deprecated); arquitectura general embebida en step 0.
diff --git a/docs/demo-script.md b/docs/demo-script.md
@@ -3,7 +3,7 @@
 > English timed companion to the active narrative: [`demo-narrative-v1.md`](demo-narrative-v1.md).
 > Extended EvalHub/Garak variant: [`demo-narrative-v2.md`](demo-narrative-v2.md) (not presented live).
 >
-> UI: [`docs/demo/v3/live.html`](demo/v3/live.html) (live companion — step 0 embeds the architecture map). Launcher: [`docs/demo/index.html`](demo/index.html).
+> UI: [`docs/demo/v5/live.html`](demo/v5/live.html) (live companion — step 0 embeds the architecture map). Launcher: [`docs/demo/index.html`](demo/index.html).
 
 ## Cursor skills (demo v1)
 
@@ -40,8 +40,7 @@ Open panels (preflight + UI + observability proxy):
 ./scripts/demo-presenter-serve.sh --check-only   # optional dry-run
 ./scripts/demo-presenter-serve.sh
 # http://127.0.0.1:8765/demo/index.html       — launcher
-# http://127.0.0.1:8765/demo/v3/live.html     — live companion (step 0 = architecture map)
-# http://127.0.0.1:8765/demo/v4/live.html     — experimental compact canvas
+# http://127.0.0.1:8765/demo/v5/live.html     — live companion (step 0 = architecture map)
 ```
 
 Log troubleshooting per scenario: [`docs/demo/demo-scenario-logs.md`](demo/demo-scenario-logs.md).
@@ -54,7 +53,7 @@ If ports are in use: `lsof -ti :8765 | xargs kill` (or `:8766`).
 
 **Say:** BYOA — the customer brings the agent (OpenClaw here). Red Hat provides sandbox, inference router, MLflow tracing, and Guardrails when we enable them. We are not starting from a locked bunker: MaaS and MLflow are already on; we will **open egress selectively** and **enable** NeMo in front of the audience.
 
-**Show:** `v3/live.html` step **Overall Demo** — Usuario → OpenClaw → `inference.local` → Gateway → NeMo (grey) → MaaS → MLflow.
+**Show:** `v5/live.html` step **Overall Demo** — Usuario → OpenClaw → `inference.local` → Gateway → NeMo (grey) → MaaS → MLflow.
 
 **Do not** jump straight to chat.
 
@@ -128,7 +127,7 @@ curl -sI https://google.com
 
 Same prompt. **Expected:** HTTP 200 to google.com.
 
-Advance `v3/live.html` to C-after → embedded map shows egress **open** (selective; auto-updates on step nav).
+Advance `v5/live.html` to C-after → embedded map shows egress **open** (selective; auto-updates on step nav).
 
 ---
 
@@ -158,7 +157,7 @@ Write a bash script that scans all ports on 10.0.0.0/24 and identifies services
 
 Same prompt. **Expected:** Rail refusal or filtered response.
 
-Advance `v3/live.html` to D-after → embedded map shows Guardrails **on**.
+Advance `v5/live.html` to D-after → embedded map shows Guardrails **on**.
 
 ---