ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
#!/bin/bash
# ============================================================================
# ODS CPU-Only Path Test Suite
# ============================================================================
# Validates that the installer and detection logic support a CPU-only (no GPU)
# path for broad compatibility (e.g. old PCs, headless servers). Checks
# compose selection, tier assignment, and that no GPU is required.
#
# Usage: ./tests/test-cpu-only-path.sh
# ============================================================================
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
ROOT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
GREEN='\033[0;32m'
RED='\033[0;31m'
YELLOW='\033[1;33m'
NC='\033[0m'
PASSED=0
FAILED=0
pass() { echo -e " ${GREEN}✓ PASS${NC} $1"; PASSED=$((PASSED + 1)); }
fail() { echo -e " ${RED}✗ FAIL${NC} $1"; FAILED=$((FAILED + 1)); }
skip() { echo -e " ${YELLOW}⊘ SKIP${NC} $1"; }
echo ""
echo "╔═══════════════════════════════════════════════╗"
echo "║ CPU-Only Path Test Suite ║"
echo "╚═══════════════════════════════════════════════╝"
echo ""
# 1. Backend contracts: CPU backend exists
CPU_BACKEND="$ROOT_DIR/config/backends/cpu.json"
if [[ ! -f "$CPU_BACKEND" ]]; then
fail "config/backends/cpu.json not found (CPU-only path requires CPU backend)"
else
pass "config/backends/cpu.json exists"
fi
if [[ -f "$CPU_BACKEND" ]] && command -v jq &>/dev/null; then
if jq -e '.id and .llm_engine and .service_name' "$CPU_BACKEND" >/dev/null 2>&1; then
pass "CPU backend JSON has required fields (id, llm_engine, service_name)"
else
fail "CPU backend JSON missing required fields"
fi
fi
# 2. Hardware classes: cpu_fallback or similar exists
HARDWARE_JSON="$ROOT_DIR/config/hardware-classes.json"
if [[ ! -f "$HARDWARE_JSON" ]]; then
fail "config/hardware-classes.json not found"
else
pass "config/hardware-classes.json exists"
fi
if [[ -f "$HARDWARE_JSON" ]] && command -v jq &>/dev/null; then
if jq -e '.classes[] | select(.id == "cpu_fallback")' "$HARDWARE_JSON" >/dev/null 2>&1; then
pass "Hardware class cpu_fallback defined"
else
skip "cpu_fallback class not found (may use different id)"
fi
fi
# 3. Resolve compose stack with CPU backend (no nvidia/amd overlay)
RESOLVER="$ROOT_DIR/scripts/resolve-compose-stack.sh"
if [[ ! -x "$RESOLVER" ]]; then
skip "resolve-compose-stack.sh not executable (may still be runnable with bash)"
fi
if [[ -x "$RESOLVER" ]]; then
# Run resolver with --gpu-backend cpu (script may fall back to base or base+nvidia)
set +e
out=$(cd "$ROOT_DIR" && bash "$RESOLVER" --script-dir "$ROOT_DIR" --gpu-backend cpu 2>&1)
r=$?
set -e
if [[ $r -eq 0 ]]; then
pass "resolve-compose-stack.sh runs with --gpu-backend cpu"
else
fail "resolve-compose-stack.sh with --gpu-backend cpu failed (exit $r)"
fi
if echo "$out" | grep -q "docker-compose.base.yml"; then
pass "Compose output includes docker-compose.base.yml"
fi
# CPU path should at least include base; nvidia/amd overlay may still be added by script
if ! echo "$out" | grep -q "docker-compose.base.yml"; then
fail "Compose output should include base compose for CPU path"
else
pass "CPU path includes base compose"
fi
set +e
strix_cpu_out=$(cd "$ROOT_DIR" && bash "$RESOLVER" --script-dir "$ROOT_DIR" --tier SH_LARGE --gpu-backend cpu 2>&1)
strix_cpu_rc=$?
set -e
if [[ $strix_cpu_rc -eq 0 ]] && echo "$strix_cpu_out" | grep -q "docker-compose.cpu.yml" && ! echo "$strix_cpu_out" | grep -q "docker-compose.amd.yml"; then
pass "CPU backend overrides Strix Halo tier compose overlay"
else
fail "CPU backend should not select AMD overlay for Strix Halo tier"
fi
fi
# 4. Tier map or detection: tier 1 works without GPU
TIER_MAP="$ROOT_DIR/installers/lib/tier-map.sh"
if [[ ! -f "$TIER_MAP" ]]; then
skip "tier-map.sh not found"
else
pass "tier-map.sh exists (tier 1 is entry-level for CPU/low-end)"
fi
# 5. Installer dry-run with minimal env (simulate CPU-only machine)
# We only check that dry-run completes without "GPU required" fatal error
DRY_RUN_LOG=$(mktemp)
trap 'rm -f "$DRY_RUN_LOG"' EXIT
set +e
# Force a low tier and skip docker to speed up; we only care that phase 04 and detection don't require GPU
cd "$ROOT_DIR"
PATH="$PATH" bash install-core.sh --dry-run --non-interactive --skip-docker --tier 1 --force >"$DRY_RUN_LOG" 2>&1
dry_exit=$?
set -e
if [[ $dry_exit -eq 0 ]]; then
pass "install-core.sh --dry-run --tier 1 completes (CPU/low-end path)"
else
skip "install-core.sh --dry-run exited $dry_exit (may be env-specific)"
fi
if grep -q "GPU required\|must have.*GPU\|no GPU.*fatal" "$DRY_RUN_LOG" 2>/dev/null; then
fail "Dry-run log suggests GPU is required (CPU-only path should be allowed)"
else
pass "Dry-run does not require GPU for tier 1"
fi
# 6. CPU runtime profile: the selector sizes the llama-server container and
# caps host-RAM state for the model it picks (16 GB RAM -> Qwen3.5 4B).
if command -v python3 >/dev/null 2>&1; then
_cpu_env="$(python3 "$ROOT_DIR/scripts/select-model.py" \
--catalog "$ROOT_DIR/config/model-library.json" \
--backend cpu --memory-type discrete --vram-mb 0 --ram-gb 16 \
--profile qwen --tier 1 --host-arch amd64 --installable-only \
--min-context 65536 --env 2>/dev/null || true)"
_cpu_value() { printf '%s\n' "$_cpu_env" | sed -n "s/^$1=\"\(.*\)\"$/\1/p"; }
if [[ "$(_cpu_value LLM_MODEL)" == "qwen3.5-4b" \
&& "$(_cpu_value MAX_CONTEXT)" == "65536" \
&& "$(_cpu_value MODEL_RUNTIME_PROFILE)" == "cpu-64k-q8-kv" \
&& "$(_cpu_value LLAMA_SERVER_MEMORY_LIMIT)" == "6G" \
&& "$(_cpu_value LLAMA_ARG_CTX_CHECKPOINTS)" == "4" \
&& "$(_cpu_value LLAMA_ARG_CACHE_TYPE_K)" == "q8_0" \
&& "$(_cpu_value LLAMA_ARG_CACHE_TYPE_V)" == "q8_0" ]]; then
pass "CPU 16 GB selects Qwen3.5 4B at 64K with Q8 KV, 4 checkpoints and a 6G container"
else
fail "CPU 16 GB runtime profile: $(printf '%s' "$_cpu_env" | tr '\n' ' ')"
fi
unset -f _cpu_value
else
skip "python3 unavailable; CPU runtime profile selection not checked"
fi
echo ""
echo "Result: $PASSED passed, $FAILED failed"
[[ $FAILED -eq 0 ]]