Add offline Workstation payload foundation

.gitignore

@@ -12,3 +12,4 @@ coverage/
 !.env.example
 test-results/
 playwright-report/
+autoinstall/offline-output/

autoinstall/offline/README.md

@@ -0,0 +1,33 @@
+# Nexus One AI Offline Server Payload
+
+This directory defines the payload embedded by `build-iso-offline.sh`. The
+offline installer is deliberately separate from the existing online bootstrap
+ISO until disconnected clean-install validation is complete.
+
+## Offline Workstation contents
+
+- the exact `aipackage` source revision used for the build
+- Ubuntu, Docker, NVIDIA and appliance `.deb` packages
+- Python wheels for the backend, RAG, ChromaDB and optional JupyterLab
+- Miniconda installer and an offline Python 3.11 environment package
+- Ollama runtime with `phi3:mini` and `nomic-embed-text`
+- vLLM and web/monitoring container images
+- the pinned Phi-3 Hugging Face model snapshot used by vLLM
+- K3s, MinIO and `mc` binaries for update/model packs (not enabled by Starter)
+- SHA-256 manifests and build provenance
+
+Server L/Max models are intentionally separate signed model packs. Optional external
+connectors cannot operate without a customer-approved route to their target.
+
+## Build stages
+
+1. On a connected Ubuntu 22.04 staging host with at least 80 GiB free, run
+   `sudo bash autoinstall/offline/build-bundle.sh`.
+2. Review `autoinstall/offline-output/manifest/` and sign
+   `SHA256SUMS` with the release key.
+3. Run `sudo bash autoinstall/build-iso-offline.sh` to inject the verified
+   payload into the Ubuntu Server ISO.
+4. Validate with all network interfaces disconnected.
+
+The bundle builder fails closed when an expected payload is absent. An ISO must
+not be labelled offline merely because the source tree was embedded.

autoinstall/offline/apt-packages.txt

@@ -0,0 +1,27 @@
+# Packages required in addition to the Ubuntu 22.04.5 Server media.
+ansible
+avahi-daemon
+build-essential
+ca-certificates
+curl
+ffmpeg
+git
+gnupg
+htop
+jq
+libmupdf-dev
+lsb-release
+mupdf-tools
+net-tools
+nginx
+openssh-server
+python3-pip
+python3-venv
+python3.11
+python3.11-venv
+rsync
+software-properties-common
+ttyd
+unzip
+wget
+whiptail

autoinstall/offline/build-bundle.sh

@@ -0,0 +1,115 @@
+#!/usr/bin/env bash
+# Stage all network-fetched Server Starter inputs for a disconnected install.
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PACKAGE_DIR="$(cd "$SCRIPT_DIR/../.." && pwd)"
+OUTPUT_DIR="${OFFLINE_OUTPUT_DIR:-$PACKAGE_DIR/autoinstall/offline-output}"
+PAYLOAD_DIR="$OUTPUT_DIR/payload"
+MANIFEST_DIR="$OUTPUT_DIR/manifest"
+
+# shellcheck source=payload.env
+source "$SCRIPT_DIR/payload.env"
+
+if [ "${EUID:-$(id -u)}" -ne 0 ]; then
+  echo "ERROR: run as root on a connected Ubuntu 22.04 staging host." >&2
+  exit 1
+fi
+. /etc/os-release
+if [ "${ID:-}" != ubuntu ] || [ "${VERSION_ID:-}" != 22.04 ]; then
+  echo "ERROR: Ubuntu 22.04 staging host required; found ${PRETTY_NAME:-unknown}." >&2
+  exit 1
+fi
+
+for command in apt-get curl docker git python3 rsync sha256sum tar; do
+  command -v "$command" >/dev/null || {
+    echo "ERROR: missing staging command: $command" >&2
+    exit 1
+  }
+done
+
+available_kb="$(df -Pk "$(dirname "$OUTPUT_DIR")" | awk 'NR==2 {print $4}')"
+if [ "$available_kb" -lt $((80 * 1024 * 1024)) ]; then
+  echo "ERROR: at least 80 GiB free is required for the Workstation offline payload." >&2
+  exit 1
+fi
+
+rm -rf "$OUTPUT_DIR"
+mkdir -p "$PAYLOAD_DIR"/{apt,containers,models,python,runtime,source} "$MANIFEST_DIR"
+
+echo "-> Capturing pinned appliance source"
+git -C "$PACKAGE_DIR" rev-parse HEAD > "$MANIFEST_DIR/source-commit.txt"
+rsync -a --delete \
+  --exclude '.git/' --exclude '*.iso' --exclude 'offline-output/' \
+  --exclude 'how-it-flows.gif' \
+  "$PACKAGE_DIR/" "$PAYLOAD_DIR/source/aipackage/"
+
+echo "-> Downloading Ubuntu package payload"
+export DEBIAN_FRONTEND=noninteractive
+mapfile -t apt_packages < <(sed -E '/^[[:space:]]*(#|$)/d' "$SCRIPT_DIR/apt-packages.txt")
+apt-get update
+apt-get install -y --download-only --reinstall \
+  -o "Dir::Cache::archives=$PAYLOAD_DIR/apt" "${apt_packages[@]}"
+
+echo "-> Downloading Python wheelhouse"
+python3 -m pip download --dest "$PAYLOAD_DIR/python" \
+  -r "$PACKAGE_DIR/ansible/roles/cezen-backend/files/requirements.txt" \
+  -r "$SCRIPT_DIR/python-packages.txt"
+
+echo "-> Capturing Miniconda installer"
+curl --fail --location --retry 3 "$MINICONDA_URL" \
+  --output "$PAYLOAD_DIR/runtime/miniconda.sh"
+chmod 0755 "$PAYLOAD_DIR/runtime/miniconda.sh"
+
+echo "-> Pulling and exporting container images"
+: > "$MANIFEST_DIR/container-images.txt"
+while IFS= read -r image; do
+  case "$image" in ''|'#'*) continue ;; esac
+  docker pull "$image"
+  safe_name="$(printf '%s' "$image" | tr '/:@' '___')"
+  docker save "$image" | gzip -1 > "$PAYLOAD_DIR/containers/$safe_name.tar.gz"
+  digest="$(docker image inspect "$image" --format '{{join .RepoDigests ","}}')"
+  printf '%s\t%s\n' "$image" "$digest" >> "$MANIFEST_DIR/container-images.txt"
+done < "$SCRIPT_DIR/container-images.txt"
+
+echo "-> Capturing Ollama runtime and Starter models"
+if ! command -v ollama >/dev/null; then
+  curl --fail --location --retry 3 https://ollama.com/install.sh | sh
+fi
+systemctl start ollama 2>/dev/null || true
+for model in $OLLAMA_MODELS; do ollama pull "$model"; done
+tar -C / -czf "$PAYLOAD_DIR/runtime/ollama-runtime.tar.gz" \
+  usr/local/bin/ollama usr/local/lib/ollama
+ollama_root="${OLLAMA_MODELS_DIR:-/usr/share/ollama/.ollama/models}"
+[ -d "$ollama_root" ] || ollama_root="/root/.ollama/models"
+[ -d "$ollama_root" ] || {
+  echo "ERROR: Ollama model store not found after pull." >&2
+  exit 1
+}
+tar -C "$(dirname "$ollama_root")" -czf "$PAYLOAD_DIR/models/ollama-models.tar.gz" models
+
+echo "-> Capturing vLLM Starter model snapshot"
+python3 -m pip install --quiet huggingface-hub
+python3 - "$VLLM_MODEL" "$PAYLOAD_DIR/models/huggingface" <<'PY'
+import sys
+from huggingface_hub import snapshot_download
+snapshot_download(repo_id=sys.argv[1], local_dir=sys.argv[2])
+PY
+
+echo "-> Writing provenance and integrity manifest"
+{
+  printf 'schema=%s\n' "$OFFLINE_SCHEMA"
+  printf 'tier=%s\n' "$OFFLINE_TIER"
+  printf 'ubuntu_release=%s\n' "$UBUNTU_RELEASE"
+  printf 'built_at=%s\n' "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
+  printf 'builder=%s\n' "$(hostname -f 2>/dev/null || hostname)"
+} > "$MANIFEST_DIR/provenance.env"
+(
+  cd "$OUTPUT_DIR"
+  find payload manifest -type f ! -name SHA256SUMS -print0 \
+    | sort -z | xargs -0 sha256sum > manifest/SHA256SUMS
+  sha256sum -c manifest/SHA256SUMS >/dev/null
+)
+
+echo "Offline payload staged at $OUTPUT_DIR"
+du -sh "$OUTPUT_DIR"

autoinstall/offline/container-images.txt

@@ -0,0 +1,7 @@
+# Bundle builder records the resolved digest for each image.
+ghcr.io/open-webui/open-webui:main
+vllm/vllm-openai:v0.10.2
+nvcr.io/nvidia/k8s/dcgm-exporter:3.3.0-3.2.0-ubuntu22.04
+prom/prometheus:latest
+prom/node-exporter:latest
+grafana/grafana:latest

autoinstall/offline/payload.env

@@ -0,0 +1,8 @@
+# Offline Server Starter release inputs. Mutable tags are resolved to immutable
+# image digests in the generated provenance file at bundle-build time.
+OFFLINE_SCHEMA="cezen.offline_payload.v1"
+OFFLINE_TIER="workstation"
+UBUNTU_RELEASE="22.04.5"
+MINICONDA_URL="https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh"
+OLLAMA_MODELS="phi3:mini nomic-embed-text"
+VLLM_MODEL="microsoft/Phi-3-mini-4k-instruct"

autoinstall/offline/python-packages.txt

@@ -0,0 +1,16 @@
+# Additional wheelhouse groups beyond backend requirements.txt.
+chromadb==0.5.23
+jupyterlab
+ipywidgets
+ipykernel
+notebook
+nbconvert
+langchain
+langchain-community
+llama-index
+transformers
+huggingface-hub
+peft
+bitsandbytes
+accelerate
+sentence-transformers

autoinstall/offline/verify-bundle.sh

@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PACKAGE_DIR="$(cd "$SCRIPT_DIR/../.." && pwd)"
+OUTPUT_DIR="${OFFLINE_OUTPUT_DIR:-$PACKAGE_DIR/autoinstall/offline-output}"
+
+required=(
+  manifest/provenance.env
+  manifest/source-commit.txt
+  manifest/container-images.txt
+  manifest/SHA256SUMS
+  payload/source/aipackage/install.sh
+  payload/runtime/miniconda.sh
+  payload/runtime/ollama-runtime.tar.gz
+  payload/models/ollama-models.tar.gz
+)
+
+for path in "${required[@]}"; do
+  [ -s "$OUTPUT_DIR/$path" ] || {
+    echo "ERROR: required offline payload is missing or empty: $path" >&2
+    exit 1
+  }
+done
+
+for directory in payload/apt payload/containers payload/python payload/models/huggingface; do
+  find "$OUTPUT_DIR/$directory" -type f -print -quit | grep -q . || {
+    echo "ERROR: required offline payload directory is empty: $directory" >&2
+    exit 1
+  }
+done
+
+(
+  cd "$OUTPUT_DIR"
+  sha256sum -c manifest/SHA256SUMS
+)
+
+echo "Offline payload integrity and completeness checks passed."