Skip to main content

software catalog

Docs pulled from project reposUpdated Sep 24, 2026View on GitHub

CLI Reference

← Docs Index

All es_auto_installer.sh and model-manager commands for Intel® AI for Enterprise Solutions run from the repo root.

text 
./es_auto_installer.sh <action> [target] [--env <name>] [options]

--env defaults to local when not specified.

Actions

configure

One-time machine setup. Installs Python 3.11+, yq, kubectl, and helm into /usr/local/bin. Skips tools already present. Requires sudo. Run once per machine.

bash 
./es_auto_installer.sh configure

init <layer>

Create a new environment directory and seed it with configuration for the specified layer. The layer name (e.g., inference, erag) determines which repositories are cloned and which configs are seeded.

bash 
./es_auto_installer.sh init inference                    # standard inference stack
./es_auto_installer.sh init erag                         # inference + Intel AI for Enterprise RAG (default chatqna flavour)
./es_auto_installer.sh init erag --flavour docsum        # Intel AI for Enterprise RAG with docsum pipeline preset
./es_auto_installer.sh init erag --env prod              # seed env/prod/ instead of env/local/
./es_auto_installer.sh init inference --upgrade          # move already-cloned repos to pinned revs

Creates:

  • env/<name>/global_config.yaml — edit this before installing
  • env/<name>/nodes.yaml — node IPs and SSH credentials (edit for multi-node)
  • env/<name>/inventory/hosts.yaml — targets localhost by default; edit for multi-node
  • env/<name>/config.<layer>.yaml — layer-specific config (e.g., config.inference.yaml, config.erag.yaml)
  • env/<name>/models.yaml — model catalog, pre-seeded from the layer's model_catalog
  • env/<name>/.solutions.yaml — records which layers this env was inited for

install

Deploy components. Dependencies are resolved automatically.

bash 
# Infrastructure + platform + inference (via dependency auto-pull)
./es_auto_installer.sh install inference --env local

# A single layer
./es_auto_installer.sh install platform --env local
./es_auto_installer.sh install inference --env local

# A single component (with auto-pulled dependencies)
./es_auto_installer.sh install kserve --env local

# A single component, skipping dependencies
./es_auto_installer.sh install metallb --only --env local

# Opt-in Intel AI for Enterprise RAG layer (requires init erag first)
./es_auto_installer.sh install erag --env local

# Override a config value at runtime (no file edit needed)
./es_auto_installer.sh install kserve --env local -- -e kserve_version=0.15.0

# Dry run — show what would happen without making changes
./es_auto_installer.sh install inference --env local -- --check

# Pass additional Ansible flags (use -- to separate)
./es_auto_installer.sh install inference --env local -- -vvv

Targets for install / teardown:

TargetWhat it covers
infrastructurekubernetes, storage
platformcert_manager, istio, metallb, envoy_gateway, postgresql, keycloak, object_store, minio, observability
inferencekeycloak_config, envoy_ai_gateway, kserve, litellm, langfuse, llm_services, nri_cpu_balloons
eragapp_inference_models, app_pre_install, app_vector_databases, app_keycloak_config, app_apisix, app_chat_history, app_nats, app_fingerprint, app_hpa, app_pipeline, app_edp, app_mcp_gateway, app_ui, app_watcher, app_post_install (opt-in, from ext repo)
<component>Any individual component name (e.g. kserve, grafana, metallb)

teardown

Remove components in reverse dependency order. Configuration files and environment state are preserved.

bash 
# Remove everything (cluster included)
./es_auto_installer.sh teardown infrastructure --env local

# Remove a single component
./es_auto_installer.sh teardown keycloak --env local

# Remove the Intel AI for Enterprise RAG layer (keeps platform and inference)
./es_auto_installer.sh teardown erag --env local

# Remove a layer but skip dependencies (e.g., tear down cluster without uninstalling erag first)
./es_auto_installer.sh teardown infrastructure --skip erag --env local

validate

Run post-install health checks.

bash 
./es_auto_installer.sh validate inference --env local
./es_auto_installer.sh validate kserve --env local
./es_auto_installer.sh validate erag --env local

Checks are implemented per-component as tasks/validate.yaml (asserts, connectivity, replica counts).

status

Print a component status table.

bash 
./es_auto_installer.sh status --env local

show

List all available layers and components, including which are opt-in.

bash 
./es_auto_installer.sh show

Flags

FlagDescription
--env <name>Target environment (default: local)
--flavour <name>(init only) Select a pipeline preset for the layer being inited (e.g., chatqna, docsum, audioqna)
--upgrade(init only) Move already-cloned ext/ repos onto the revs the manifests pin now. Refuses on local changes; never rewrites your configs.
--onlySkip dependency auto-inclusion — run the named target alone
--skip <names>Comma-separated layers or components to leave out of the plan (e.g., --skip erag to tear down the cluster without uninstalling erag first)
--forceSkip the confirmation prompt (required in CI, where there is no terminal to answer it)
-- <ansible-flags>Pass remaining args directly to ansible-playbook (e.g. -- -vvv, -- --check, -- -e key=value)

Environment variables

VariablePurpose
HF_TOKENHugging Face token for gated models (Llama, Mistral, etc.)
KEYCLOAK_ADMIN_PASSWORDSet Keycloak admin password before install (auto-generated if unset)
GRAFANA_ADMIN_PASSWORDSet Grafana admin password before install (auto-generated if unset)
KUBECONFIGOverride kubeconfig path (auto-detected from env/<name>/kubeconfig.yaml)

model-manager

text 
./model-manager <command> [options] [--env <name>]
CommandDescription
listList models in the catalog
deploy <name>Download weights and start serving (from catalog)
deploy --id <hf/repo>Deploy any Hugging Face model ad-hoc
undeploy <name>Stop serving (weights stay on PVC)
undeploy allStop all models
statusShow running models and their endpoints

Key flags for deploy:

FlagDescription
--cpu <n>CPU cores
--memory Memory limit
--replicas <n>Number of serving replicas
--tp <n>Tensor parallelism width (1, 2, 4, or 8)
--runtime vllm|openvinoServing runtime
--waitBlock until the model is ready
--dry-runPrint manifest without applying
--skip-downloadSkip weight download (use existing PVC data)

See Deploy a Model for usage examples.

Logs

All Ansible output is captured per environment:

FileContents
env/<name>/logs/install-all-<timestamp>.logFull install output
env/<name>/logs/teardown-all-<timestamp>.logFull teardown output
env/<name>/logs/install-<component>-<timestamp>.logPer-component install output
env/<name>/logs/ai-solutions-ca.crtSelf-signed CA certificate (import to browser)

For more Ansible detail, append -- -vvv to any command.

Related Docs

If you want to…Go to
Look up what each global_config.yaml field controlsConfiguration Reference
Deploy, access, and manage LLM inference with model-managerDeploy an LLM
Set up multi-node or bring-your-own-cluster installsDeployment Guide
See the full command dispatch flow this CLI triggersArchitecture & Design Document
PreviousNetwork Architecture — External Access & Request FlowNextConfiguration Reference