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 Installer — Compose Selection
# ============================================================================
# Part of: installers/lib/
# Purpose: Resolve which docker-compose overlay files to use based on tier,
# GPU backend, and capability profile
#
# Expects: SCRIPT_DIR, TIER, GPU_BACKEND, CAP_COMPOSE_OVERLAYS, LOG_FILE,
# GPU_COUNT, log(), warn()
# Provides: resolve_compose_config() → sets COMPOSE_FILE, COMPOSE_FLAGS;
# the service-set checks ods_gateway_assert_no_managed_inference(),
# ods_host_native_assert_no_managed_llama(),
# ods_vllm_assert_engine_stack() and ods_compose_assert_no_webui()
#
# Modder notes:
# Add new compose overlay mappings or backends here.
# ============================================================================
[[ -f "${SCRIPT_DIR:-}/lib/safe-env.sh" ]] && . "${SCRIPT_DIR}/lib/safe-env.sh"
resolve_compose_config() {
COMPOSE_FILE="docker-compose.yml"
COMPOSE_FLAGS=""
if [[ -n "${CAP_COMPOSE_OVERLAYS:-}" ]]; then
IFS=',' read -r -a profile_overlays <<< "$CAP_COMPOSE_OVERLAYS"
compose_overlay_ok=true
for overlay in "${profile_overlays[@]}"; do
if [[ -f "$SCRIPT_DIR/$overlay" ]]; then
COMPOSE_FLAGS="$COMPOSE_FLAGS -f $overlay"
else
compose_overlay_ok=false
break
fi
done
if [[ "$compose_overlay_ok" == "true" && ${#profile_overlays[@]} -gt 0 ]]; then
COMPOSE_FLAGS="${COMPOSE_FLAGS# }"
COMPOSE_FILE="${profile_overlays[${#profile_overlays[@]}-1]}"
else
COMPOSE_FLAGS=""
fi
fi
# Backward compatibility default if no flags were set.
if [[ -z "$COMPOSE_FLAGS" ]]; then
if [[ "$TIER" == "NV_ULTRA" ]]; then
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.nvidia.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.nvidia.yml"
COMPOSE_FILE="docker-compose.nvidia.yml"
fi
elif [[ "$TIER" == "CLOUD" ]]; then
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml"
COMPOSE_FILE="docker-compose.base.yml"
fi
elif [[ "$GPU_BACKEND" == "cpu" ]]; then
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.cpu.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.cpu.yml"
COMPOSE_FILE="docker-compose.cpu.yml"
fi
elif [[ "$TIER" == "SH_LARGE" || "$TIER" == "SH_COMPACT" ]]; then
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.amd.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.amd.yml"
COMPOSE_FILE="docker-compose.amd.yml"
fi
elif [[ "$TIER" == "ARC" || "$TIER" == "ARC_LITE" || "$GPU_BACKEND" == "intel" || "$GPU_BACKEND" == "sycl" ]]; then
# Prefer docker-compose.arc.yml (oneAPI build-from-source) when present;
# fall back to docker-compose.intel.yml (pre-built image) if arc.yml is absent.
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.arc.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.arc.yml"
COMPOSE_FILE="docker-compose.arc.yml"
elif [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.intel.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.intel.yml"
COMPOSE_FILE="docker-compose.intel.yml"
fi
else
if [[ -f "$SCRIPT_DIR/docker-compose.base.yml" && -f "$SCRIPT_DIR/docker-compose.nvidia.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.base.yml -f docker-compose.nvidia.yml"
COMPOSE_FILE="docker-compose.nvidia.yml"
elif [[ -f "$SCRIPT_DIR/docker-compose.yml" ]]; then
COMPOSE_FLAGS="-f docker-compose.yml"
fi
fi
fi
if [[ -z "$COMPOSE_FLAGS" ]]; then
COMPOSE_FLAGS="-f $COMPOSE_FILE"
fi
if [[ -x "$SCRIPT_DIR/scripts/resolve-compose-stack.sh" ]]; then
COMPOSE_ENV="$("$SCRIPT_DIR/scripts/resolve-compose-stack.sh" \
--script-dir "$SCRIPT_DIR" \
--tier "$TIER" \
--gpu-backend "$GPU_BACKEND" \
--profile-overlays "${CAP_COMPOSE_OVERLAYS:-}" \
--gpu-count "${GPU_COUNT:-1}" \
--ods-mode "${ODS_MODE:-local}" \
--env 2>>"$LOG_FILE")"
load_env_from_output <<< "$COMPOSE_ENV"
fi
# Layer Tier 0 memory overlay for low-RAM machines
if [[ "$TIER" == "0" && -f "$SCRIPT_DIR/docker-compose.tier0.yml" ]]; then
COMPOSE_FLAGS="$COMPOSE_FLAGS -f docker-compose.tier0.yml"
log "Including docker-compose.tier0.yml (Tier 0 memory limits)"
fi
# Auto-include docker-compose.override.yml if present (standard Docker convention).
# This lets modders add services without editing core compose files.
if [[ -f "$SCRIPT_DIR/docker-compose.override.yml" ]]; then
COMPOSE_FLAGS="$COMPOSE_FLAGS -f docker-compose.override.yml"
log "Including docker-compose.override.yml (user overrides)"
fi
log "Compose selection: $COMPOSE_FLAGS"
}
# Compose overlays append profile lists. Verify the effective service set, not
# only the selected files, before an API-only gateway pulls or starts images.
ods_gateway_assert_no_managed_inference() {
local services compose_root="${INSTALL_DIR:-$PWD}"
# Phase 08 still runs from the source checkout during an upgrade. Resolve
# against the installed project, whose generated .env supplies Compose
# interpolation values and reflects the runtime selection being changed.
services="$(cd "$compose_root" && $DOCKER_COMPOSE_CMD "$@" config --services)" || return 1
if grep -Eq '^(llama-server|model-router)$' <<< "$services"; then
printf 'Gateway-only Compose includes ODS-managed inference. Clear COMPOSE_PROFILES and retry.\n' >&2
return 1
fi
return 0
}
# A host-native install (llama-server.exe on Windows, this stack in WSL) still
# runs the ODS model-router for Pixel, but it must never pull or launch the
# in-stack llama-server. Check the effective service set because inherited
# profiles can override an overlay.
ods_host_native_assert_no_managed_llama() {
local services compose_root="${INSTALL_DIR:-$PWD}"
services="$(cd "$compose_root" && $DOCKER_COMPOSE_CMD "$@" config --services)" || return 1
if grep -qx 'llama-server' <<< "$services"; then
printf 'Host-native llama-server Compose also starts the in-stack llama-server. Clear COMPOSE_PROFILES and retry.\n' >&2
return 1
fi
return 0
}
# The vLLM engine (LLM_BACKEND=vllm) serves the model in vllm-server, and the
# in-stack llama-server must stay off: both would claim the same GPU. Check the
# effective service set, because inherited profiles can override an overlay.
ods_vllm_assert_engine_stack() {
local services compose_root="${INSTALL_DIR:-$PWD}"
services="$(cd "$compose_root" && $DOCKER_COMPOSE_CMD "$@" config --services)" || return 1
if ! grep -qx 'vllm-server' <<< "$services"; then
printf 'The vLLM Compose stack has no vllm-server service; docker-compose.vllm.yml is not selected.\n' >&2
return 1
fi
if grep -qx 'llama-server' <<< "$services"; then
printf 'The vLLM Compose stack also starts the in-stack llama-server. Clear COMPOSE_PROFILES and retry.\n' >&2
return 1
fi
return 0
}
# Supply the early checks' placeholder GID to Compose. On a fresh install Phase
# 05 selects "sudo docker compose" because the new docker group does not apply
# to the running session yet, and sudo does not pass exported variables, so
# hand the value to Compose explicitly. Call only inside a subshell function.
_ods_compose_placeholder_pixel_gid() {
export PIXEL_INGRESS_GID="${PIXEL_INGRESS_GID:-1}"
# An inherited real GID also needs forwarding when the initial .env still
# has no installed identity. Keep the command-string assignment numeric.
[[ "$PIXEL_INGRESS_GID" =~ ^[1-9][0-9]*$ ]] || return 1
if [[ "${DOCKER_COMPOSE_CMD:-}" == "sudo "* ]]; then
DOCKER_COMPOSE_CMD="sudo env PIXEL_INGRESS_GID=$PIXEL_INGRESS_GID ${DOCKER_COMPOSE_CMD#sudo }"
fi
}
# Before image pulls, Pixel's ingress group has not been created yet. Supply
# an ephemeral numeric GID only for Compose's read-only service selection.
# Phase 11 still validates the installed identity and uses the strict helper.
ods_host_native_assert_no_managed_llama_before_pixel_identity() (
_ods_compose_placeholder_pixel_gid || return 1
ods_host_native_assert_no_managed_llama "$@"
)
ods_vllm_assert_engine_stack_before_pixel_identity() (
_ods_compose_placeholder_pixel_gid || return 1
ods_vllm_assert_engine_stack "$@"
)
# A caller can inherit COMPOSE_PROFILES=gateway-webui. Check the effective
# service set before pulling or starting a Portal-only stack.
ods_compose_assert_no_webui() {
local services compose_root="${INSTALL_DIR:-$PWD}"
services="$(cd "$compose_root" && $DOCKER_COMPOSE_CMD "$@" config --services)" || return 1
if grep -qx 'open-webui' <<< "$services"; then
printf 'No-WebUI Compose still enables Open WebUI. Clear COMPOSE_PROFILES and retry.\n' >&2
return 1
fi
return 0
}
# Phase 08 checks the selected service list before image pulls, while Pixel's
# private ingress group is created in Phase 11. Compose interpolates every
# selected service even for `config --services`, so supply a numeric GID only
# within this read-only early check. Phase 11 validates the real installed GID
# before starting containers and repeats the no-WebUI check.
ods_compose_assert_no_webui_before_pixel_identity() (
_ods_compose_placeholder_pixel_gid || return 1
ods_compose_assert_no_webui "$@"
)