AI Accelerator: A production-ready MLOps platform to deploy, monitor, and govern machine learning models with security, scalability, and auditability.
Project description
AIAC Console Command Reference
This document lists all currently available CLI commands and their options.
Run Forms
Use this form:
aiac <group> <command> [options]
Command Groups
authserverdeploymentmonitoringgovernanceadmin
Quick Help
aiac --help
aiac auth --help
aiac server --help
aiac deployment --help
aiac monitoring --help
aiac governance --help
aiac admin --help
server
server --help
aiac server --help
Usage: aiac server [OPTIONS] COMMAND [ARGS]...
Local API server commands
Options:
--help Show this message and exit.
Commands:
run Run the AI Accelerator Django API server.
server run
Run the AI Accelerator Django API server.
Options:
--host(default127.0.0.1)--port(default8000)--no-reload/--reload(default--no-reload)--background/--foreground(default--background)--migrate/--no-migrate(default--migrate)
Example:
aiac server run --host 127.0.0.1 --port 8000 --no-reload
Notes:
- By default, AIAC runs database migrations automatically before server startup.
server status
Show local background server status.
Example:
aiac server status
server stop
Stop local background server started by aiac server run.
Example:
aiac server stop
auth
auth register
Register a new user.
Options:
--email(prompted if omitted)--username(prompted if omitted)--password(prompted, hidden)--role(prompted)
Example:
aiac auth register --email user@example.com --username user1 --password secret --role client
auth login
Login and save access/refresh tokens.
Options:
--email(prompted)--password(prompted, hidden)
Example:
aiac auth login --email user@example.com --password secret
auth logout
Logout using refresh token (or saved token if blank).
Options:
--refresh-token(prompted, hidden)
Example:
aiac auth logout
auth me
Show current authenticated user info.
Example:
aiac auth me
auth password-reset
Request a secure password reset email.
Example:
aiac auth password-reset --email user@example.com
auth token-show
Verify credentials then show masked saved tokens.
Options:
--email(prompted)--password(prompted, hidden)
Example:
aiac auth token-show --email user@example.com --password secret
deployment
deployment create-project-deployment
Create a new project.
Options:
--owner-id(prompted)--name(prompted)--description(prompted)
Example:
aiac deployment create-project-deployment
deployment list-projects
List all projects.
Example:
aiac deployment list-projects
deployment delete-project
Delete a project.
Options:
--project-id(prompted)--confirm/--no-confirm(default confirm)
Example:
aiac deployment delete-project --project-id 2 --no-confirm
deployment create-model-version
Create model version from file upload.
Options:
--project-id(prompted)--description(prompted)--field-file-path(prompted)
Example:
aiac deployment create-model-version --project-id 1 --description "v1" --field-file-path model.pkl
deployment list-model-versions
List all model versions.
Options:
--status(filter by status)--project-id(filter by project)--deployed-only(only deployed)
Example:
aiac deployment list-model-versions
aiac deployment list-model-versions --status approved
deployment approve-model-version
Approve a model version for deployment. This marks the version as approved so it can be deployed.
Options:
--model-version-id(prompted)--note(optional note)
Example:
aiac deployment approve-model-version --model-version-id 4 --note "QA passed"
deployment reject-model-version
Reject a model version. This blocks the version from deployment until it is approved again.
Options:
--model-version-id(prompted)--note(optional note)
Example:
aiac deployment reject-model-version --model-version-id 4 --note "Missing bias checks"
deployment retire-model-version
Retire a model version. This keeps the version in history but marks it as retired so it should not be deployed again.
Options:
--model-version-id(prompted)--note(optional note)
Example:
aiac deployment retire-model-version --model-version-id 4 --note "Replaced by v5"
deployment list-model-approvals
List approval history for a model version.
Options:
--model-version-id(prompted)--format, -f(table|json, defaulttable)
Example:
aiac deployment list-model-approvals --model-version-id 4
deployment delete-model-version
Delete a model version.
Options:
--model-version-id(prompted)--confirm/--no-confirm(default confirm)
Example:
aiac deployment delete-model-version --model-version-id 4 --no-confirm
deployment deploy-model-version
Deploy a model version and optionally track progress.
Options:
--user-id(prompted)--model-version-id(prompted)--port(prompted)--auto-preflight/--no-auto-preflight(default--auto-preflight)--start-worker/--no-start-worker(default--start-worker)--wait/--no-wait(default--wait)--poll-interval-seconds(default2)--timeout-seconds(default900)--precheck/--no-precheck(default--precheck)--check-local-port/--no-check-local-port(default--check-local-port)--format, -f(text|json, defaulttext)--redis-host(defaultlocalhost)--redis-port(default6379)
Example:
aiac deployment deploy-model-version --user-id 14 --model-version-id 4 --port 6000 --format text
deployment preflight
Run pre-deployment checks before deploying a model.
Checks:
- API server reachability
- Docker CLI and Docker daemon
- Redis reachability
- Celery worker availability
- Optional model version existence
- Optional local port availability
Options:
--model-version-id(optional)--port(optional)--redis-host(defaultlocalhost)--redis-port(default6379)--check-local-port/--no-check-local-port(default--check-local-port)--format, -f(table|json, defaulttable)
Examples:
aiac deployment preflight --model-version-id 4 --port 6000
aiac deployment preflight --format json
deployment redeploy-model
Redeploy existing deployment.
Options:
--deployment-id(prompted)--start-worker/--no-start-worker(default--start-worker)--wait/--no-wait(default--wait)--timeout-seconds(default900)--redis-host(defaultlocalhost)--redis-port(default6379)
Example:
aiac deployment redeploy-model --deployment-id 20
deployment stop-deployment
Stop a deployment.
Options:
--deployment-id(prompted)
Example:
aiac deployment stop-deployment --deployment-id 20
deployment delete-deployment
Delete a deployment.
Options:
--deployment-id(prompted)--confirm/--no-confirm(default confirm)
Example:
aiac deployment delete-deployment --deployment-id 20 --no-confirm
deployment list-deployments
List all deployments.
Example:
aiac deployment list-deployments
deployment get-deployment-details
Show deployment details + runtime URLs.
Options:
--deployment-id(prompted)
Example:
aiac deployment get-deployment-details --deployment-id 20
deployment advisor
Advanced advisor report (risk/strategy/metrics).
Options:
--deployment-id(prompted)--format, -f(table|json, defaulttable)
Example:
aiac deployment advisor --deployment-id 20 --format table
deployment services
Show runtime service URLs and optional health probe.
Options:
--deployment-id(prompted)--probe/--no-probe(default--no-probe)--timeout(probe timeout, default2)--format, -f(table|json, defaulttable)
Example:
aiac deployment services --deployment-id 20 --probe --format table
Runtime capabilities:
- Runtime exposes
GET /capabilitiesfor NLP/CV support and dependency checks. - Example:
curl http://127.0.0.1:8003/capabilities
deployment traffic-shadow
Compare active model vs candidate model on recent samples.
Options:
--deployment-id(prompted)--candidate, -c(required candidate model version id)--samples, -s(default200, clamped10..1000)--format, -f(table|json, defaulttable)
Example:
aiac deployment traffic-shadow --deployment-id 20 --candidate 3 --samples 300 --format json
Result interpretation:
Deployment: live deployment being evaluated.Current Model: model version currently serving traffic.Candidate: model version tested in shadow mode.Samples: number of evaluation samples used.Recommendation:promote_candidate: candidate is safe/better to promote.keep_current: keep current model.review_required: results are mixed; validate manually.
Shadow metrics:
Match Rate: fraction of samples where current and candidate predictions match (matches / samples).Matches: count of matching predictions.Current Latency (ms): average inference latency of current model.Candidate Latency (ms): average inference latency of candidate model.Latency Ratio:candidate_latency / current_latency.< 1.0: candidate is faster.= 1.0: similar speed.> 1.0: candidate is slower.
Avg Abs Diff: average absolute prediction difference (mainly for numeric/regression outputs).0.0means no numeric deviation on tested samples.
deployment k8s-deploy
Deploy AI runtime image to Kubernetes and expose a Service.
Options:
--deployment-id(prompted)--image, -i(required container image)--namespace, -n(defaultdefault)--replicas, -r(default1)--container-port(default8000)--service-port(default80)--service-type(ClusterIP|NodePort|LoadBalancer, defaultClusterIP)--rollout-strategy(RollingUpdate|Recreate, defaultRollingUpdate)--max-unavailable(default25%, RollingUpdate only)--max-surge(default25%, RollingUpdate only)--kubeconfig(optional)--context(optional)
Notes:
- If kubeconfig is missing, AIAC tries automatic Minikube context refresh (
minikube update-context) before failing. - On PowerShell, explicit kubeconfig path format:
--kubeconfig $env:USERPROFILE\\.kube\\config
Example:
aiac deployment k8s-deploy --deployment-id 20 --image ghcr.io/acme/model-runtime:1.0 --namespace ml --replicas 2 --service-type LoadBalancer
aiac deployment k8s-deploy --deployment-id 20 --image ghcr.io/acme/model-runtime:1.0 --rollout-strategy RollingUpdate --max-unavailable 1 --max-surge 1
deployment k8s-preflight
Run Kubernetes readiness checks before running other k8s commands.
Checks include:
- Kubernetes SDK installed
- kubeconfig/context load
- API reachability
- namespace existence
- optional deployment resource existence
Options:
--namespace, -n(defaultdefault)--deployment-id(optional)--kubeconfig(optional)--context(optional)
Examples:
aiac deployment k8s-preflight
aiac deployment k8s-preflight --namespace ml --deployment-id 20
deployment k8s-bootstrap
Install Minikube (Windows/winget) and start a local Kubernetes cluster.
Options:
--install/--no-install(default--install)--driver(defaultdocker)--profile(optional Minikube profile)
Examples:
aiac deployment k8s-bootstrap
aiac deployment k8s-bootstrap --no-install --driver docker
aiac deployment k8s-bootstrap --profile aiac-local
deployment k8s-doctor
Diagnose Kubernetes environment and print exact recovery commands.
Checks include:
- Docker CLI availability
kubectlandminikubeavailability- Python Kubernetes SDK installation
- Minikube cluster status
- kubeconfig/API reachability
Options:
--kubeconfig(optional)--context(optional)
Examples:
aiac deployment k8s-doctor
aiac deployment k8s-doctor --kubeconfig $env:USERPROFILE\.kube\config
Local setup (Minikube + Docker):
- Prerequisite: Docker Desktop/Engine must be running.
- Install Minikube:
winget install Kubernetes.minikube
- Start a local cluster using Docker driver:
minikube start --driver=docker
- Verify cluster:
kubectl cluster-info
kubectl get nodes
- Then run AIAC preflight:
aiac deployment k8s-preflight
deployment k8s-status
Show Kubernetes deployment/service status for an AIAC deployment ID.
Options:
--deployment-id(prompted)--namespace, -n(defaultdefault)--kubeconfig(optional)--context(optional)
Example:
aiac deployment k8s-status --deployment-id 20 --namespace ml
deployment k8s-scale
Scale Kubernetes replicas for an AIAC deployment ID.
Options:
--deployment-id(prompted)--replicas, -r(prompted)--namespace, -n(defaultdefault)--kubeconfig(optional)--context(optional)
Example:
aiac deployment k8s-scale --deployment-id 20 --replicas 4 --namespace ml
deployment k8s-hpa
Create or update Horizontal Pod Autoscaler for an AIAC deployment.
Options:
--deployment-id(prompted)--min(default1)--max(prompted)--cpu-percent(default70)--namespace, -n(defaultdefault)--kubeconfig(optional)--context(optional)
Example:
aiac deployment k8s-hpa --deployment-id 20 --min 2 --max 8 --cpu-percent 65 --namespace ml
deployment k8s-delete
Delete Kubernetes Deployment and Service created by k8s-deploy.
Options:
--deployment-id(prompted)--namespace, -n(defaultdefault)--yes(skip confirmation)--kubeconfig(optional)--context(optional)
Behavior:
- Deletes HPA (
<name>-hpa) if present. - Deletes Service and Deployment.
Example:
aiac deployment k8s-delete --deployment-id 20 --namespace ml --yes
deployment k8s-rollback
Rollback Kubernetes deployment image to a previous ReplicaSet revision.
Options:
--deployment-id(prompted)--namespace, -n(defaultdefault)--to-revision(default0: auto-select previous revision)--yes(skip confirmation)--kubeconfig(optional)--context(optional)
Examples:
aiac deployment k8s-rollback --deployment-id 20 --namespace ml
aiac deployment k8s-rollback --deployment-id 20 --namespace ml --to-revision 3 --yes
deployment explain-decision
Run explainable decision inference with configurable refusal checks.
Options:
--deployment-id(prompted)--features, -x(JSON numeric feature array)--min-confidence(optional refusal threshold)--min-margin(optional refusal threshold)--blocked-labels(comma-separated labels to refuse)--fallback/--no-fallback(default--fallback; use/predictif/predict-decisionis missing)--timeout(request timeout seconds, default10)--format, -f(table|json, defaulttable)
Example:
aiac deployment explain-decision --deployment-id 20 --features "[0.1, 0.2, 0.3]" --min-confidence 0.7 --min-margin 0.15 --blocked-labels "denied,blocked" --format table
Result interpretation:
decision=approved: runtime policy checks passed.decision=refused: runtime policy checks rejected the request.decision=approved_with_fallback:/predict-decisionwas unavailable, so CLI used/predict; refusal checks were not enforced.confidence=Noneandmargin=None: model/runtime did not provide probability scores, so confidence-margin checks could not be evaluated.
Output includes:
- Decision summary table
- Interpretation block (plain-language explanation)
- Reasons list
- Top probabilities (if available)
- Linear feature contributions and contribution summary (if available)
Troubleshooting:
- If you see
runtime request failed: 404or fallback messages, your deployment runtime is likely outdated. Redeploy the model to enable/predict-decision. - If compliance/safety requires strict refusal enforcement, run with
--no-fallbackso the command fails instead of using/predict. - If
confidence/marginremainNoneafter redeploy, your model may not expose probabilities; use models/pipelines withpredict_probafor confidence-based refusal checks.
monitoring
monitoring deploy-stats
Live or one-shot deployment stats + health warnings.
Options:
--deployment-id(prompted)--format, -f(table|json, defaulttable)--watch, -w(continuous polling)--interval, -i(seconds, default5)--iterations(0= infinite)--cpu-warn(default85.0)--ram-warn(default85.0)--latency-warn(default500.0)--error-rate-warn(default5.0)
Example:
aiac monitoring deploy-stats --deployment-id 4 --watch --interval 10
monitoring deploy-records
Show monitoring records with summary; supports table/json/csv.
Options:
--deployment-id(prompted)--format, -f(table|json|csv, defaulttable)--limit, -l(default50)--watch, -w--interval, -i(default5)--iterations(0= infinite)--cpu-warn(default85.0)--ram-warn(default85.0)--latency-warn(default500.0)
Example:
aiac monitoring deploy-records --deployment-id 10 --format table
monitoring alert
List alerts for deployment.
Options:
--deployment-id(prompted)--include-resolved(include resolved alerts)--limit, -l(max alerts displayed)--only-type(filter by alert type)--only-severity(filter by severity: low|medium|high)--since-hours(only alerts in last N hours)--since(only alerts after ISO timestamp)--format, -f(table|json, defaulttable)
Example:
aiac monitoring alert --deployment-id 18
aiac monitoring alert --deployment-id 18 --only-severity high --since-hours 24
aiac monitoring alert --deployment-id 18 --only-type latency --since 2026-02-25T10:00:00Z
monitoring resolve-alert
Resolve alert by id, or choose from deployment alerts.
Options:
--alert-id, -a(alert UUID)--deployment-id, -d(to select alert interactively)--include-resolved(include already resolved alerts when selecting)--yes, -y(skip confirmation)
Notes:
- After resolving interactively, CLI prints the resolved alert summary (type/message/deployment).
Examples:
aiac monitoring resolve-alert --alert-id <uuid>
aiac monitoring resolve-alert --deployment-id 18
monitoring health-report
Advanced health report with trends/recommendations.
Options:
--deployment-id(prompted)--window, -w(records window, default50)--format, -f(table|json, defaulttable)
Example:
aiac monitoring health-report --deployment-id 4 --window 100
monitoring cost-intelligence
Advanced FinOps report with monthly/annual cost estimation, efficiency scoring, budget variance, risk flags, and scenario projections.
Options:
--deployment-id(prompted)--window, -w(default200)--cpu-hour-rate(default0.05)--gb-ram-hour-rate(default0.01)--request-million-rate(default1.0)--ram-reference-gb(default4.0)--budget(optional monthly budget for variance analysis)--target-cpu-utilization(default65.0)--target-ram-utilization(default70.0)--scenarios/--no-scenarios(default--scenarios)--format, -f(table|json, defaulttable)
Example:
aiac monitoring cost-intelligence --deployment-id 4 --window 300 --budget 250 --target-cpu-utilization 65 --target-ram-utilization 70 --scenarios --format table
monitoring detect-drift
Drift detection with profiles, thresholds, history, and watch mode.
Options:
--model-version-id(prompted)--profile, -p(sensitive|balanced|conservative, defaultbalanced)--kl-threshold(override)--wasserstein-threshold(override)--ks-threshold(override)--chi-square-threshold(override)--explain/--no-explain(default explain)--history(show previous scans)--history-limit(default5)--watch, -w--interval, -i(default30)--iterations(0= infinite)
Example:
aiac monitoring detect-drift --model-version-id 4 --profile balanced --history
monitoring samples
Upload/validate sample data for monitoring and drift.
Options:
--model-version-id(prompted)--data-samples(JSON array string)--samples-file(JSON or CSV file)--csv-file(CSV file)--drop-last-column(CSV only; removes last column per row, useful for label/target column)--format, -f(auto|json|csv, for--samples-file, defaultauto)--chunk-size(0single request)--dry-run(validate only)--preview(show first N samples, default3)--strict-shape(require same vector dimension)
Examples:
aiac monitoring samples --model-version-id 4 --data-samples "[0.1, 0.2, 0.3]"
aiac monitoring samples --model-version-id 4 --samples-file samples.json
aiac monitoring samples --model-version-id 4 --csv-file samples.csv --chunk-size 100
aiac monitoring samples --model-version-id 4 --csv-file samples.csv --drop-last-column --preview 20
governance
governance create-policy
Create policy. If --rules is omitted, CLI prompts interactive metadata template (deployment|monitoring|both) and rules.
Arguments:
name(required)policy_type(required)
Options:
--description, -d(optional text)--rules, -r(JSON object string)
Examples:
aiac governance create-policy model deployment
aiac governance create-policy model deployment -d "prod policy" -r '{"max_latency_ms": 500}'
governance list-policies
List policies with owner, creator, created_at, description, and rules preview.
Example:
aiac governance list-policies
governance delete-policy
Delete policy by id.
Arguments:
policy_id(required)
Example:
aiac governance delete-policy 3
governance apply-policy
Apply policy to deployment.
Arguments:
policy_id(required)deployment_id(optional; if omitted, prompt asks)
Examples:
aiac governance apply-policy 3 14
aiac governance apply-policy 3
governance view-violations
View policy violations.
Example:
aiac governance view-violations
governance resolve-violation
Mark a violation as resolved.
Options:
--violation-id(prompted)
Example:
aiac governance resolve-violation --violation-id 12
governance reopen-violation
Reopen a previously resolved violation.
Options:
--violation-id(prompted)
Example:
aiac governance reopen-violation --violation-id 12
governance resolve-all-violations
Resolve all unresolved violations. You can filter by deployment and/or policy.
Options:
--deployment-id(optional)--policy-id(optional)
Examples:
aiac governance resolve-all-violations
aiac governance resolve-all-violations --deployment-id 14
aiac governance resolve-all-violations --policy-id 3
aiac governance resolve-all-violations --deployment-id 14 --policy-id 3
governance metrics
Show aggregated violation metrics.
Options:
--deployment-id(optional filter)--severity(low|medium|high, optional)--resolved(all|true|false, defaultall)--format, -f(table|json, defaulttable)
Example:
aiac governance metrics
aiac governance metrics --deployment-id 14 --severity high --resolved false
governance run-policy-engine
Trigger policy engine execution immediately.
Example:
aiac governance run-policy-engine
governance debug-policy-engine
Debug why violations are/aren't generated for a policy.
Options:
--policy, -p(required policy id)--deployment, -d(optional deployment filter)--limit, -l(default50, clamped1..200)
Example:
aiac governance debug-policy-engine --policy 3 --deployment 14 --limit 100
governance policy-insights
Advanced coverage/risk insights with recommendations.
Options:
--policy, -p(optional)--days, -d(default7, clamped1..90)--format, -f(table|csv, defaulttable)
Examples:
aiac governance policy-insights
aiac governance policy-insights --policy 3 --days 30 --format csv
governance alert-logs
View governance alert logs.
Example:
aiac governance alert-logs
admin
admin list-users
List all users (admin only).
Options:
--role(filter by role)--active(only active users)--inactive(only inactive users)--staff(only staff users)--email(filter by email substring)--username(filter by username substring)--limit, -l(limit results)--offset(skip first N results)--format, -f(table|json, defaulttable)
Example:
aiac admin list-users
aiac admin list-users --role admin --active
aiac admin list-users --email "@gmail.com" --limit 20
aiac admin list-users --username dad --offset 10
admin user
Show details for one user.
Options:
--user-id(prompted)
Example:
aiac admin user --user-id 7
admin promote-admin
Promote a user to admin.
Options:
--user-id(prompted)
Example:
aiac admin promote-admin --user-id 7
admin demote-admin
Demote a user to client.
Options:
--user-id(prompted)
Example:
aiac admin demote-admin --user-id 7
admin activate-user
Activate a user account.
Options:
--user-id(prompted)
Example:
aiac admin activate-user --user-id 7
admin deactivate-user
Deactivate a user account.
Options:
--user-id(prompted)
Example:
aiac admin deactivate-user --user-id 7
admin soft-delete-user
Soft delete (deactivate) a user account.
Options:
--user-id(prompted)
Example:
aiac admin soft-delete-user --user-id 7
admin reset-password
Reset a user password.
Options:
--user-id(prompted)--password(prompted, hidden)
Example:
aiac admin reset-password --user-id 7
admin list-audits
List audit logs (admin only).
Options:
--limit, -l(default50)--format, -f(table|json, defaulttable)
Example:
aiac admin list-audits --limit 100
admin export-audits
Export audit logs to CSV.
Options:
--out(CSV file path, defaultaudit_logs.csv)
Example:
aiac admin export-audits --out audit_logs.csv
admin export-audits-json
Export audit logs to JSON.
Options:
--out(JSON file path, defaultaudit_logs.json)
Example:
aiac admin export-audits-json --out audit_logs.json
Notes
- Most required values are prompted interactively if omitted.
- Booleans can be toggled with
--flag/--no-flagfor Typer bool options. - For the most accurate local signature at any time:
aiac <group> <command> --help
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file ai_accelerator-1.1.2.tar.gz.
File metadata
- Download URL: ai_accelerator-1.1.2.tar.gz
- Upload date:
- Size: 303.2 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
ecbe56445ef54a293ca8a586a3bc26c81d33fea1e6a970437d9b31bbfc90391c
|
|
| MD5 |
3a31f0e09007c3f99051348dab9a15a6
|
|
| BLAKE2b-256 |
a26beb41c10e6079ca06b09fc67eeb0a4a75b166218e2e0013c993883bbf3dd9
|
File details
Details for the file ai_accelerator-1.1.2-py3-none-any.whl.
File metadata
- Download URL: ai_accelerator-1.1.2-py3-none-any.whl
- Upload date:
- Size: 220.4 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.5
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
40c895cf78a311eed2f3574bfc87cfeadaf72203937fe8bdca624bc6f2ebe1e5
|
|
| MD5 |
92a498f6264832524d7f446adbcfa96b
|
|
| BLAKE2b-256 |
8dfc7d78aaaa5fff4d5de1aaa6fcd62ef86a8a4312c41fff0c593f49045ddf7e
|