LiteLLM on Google Cloud Run

LiteLLM is an open-source LLM proxy and AI gateway that provides a unified OpenAI-compatible API across 100+ providers including OpenAI, Anthropic, Google Gemini, Azure OpenAI, AWS Bedrock, and Ollama. Organizations use it to centralize AI spend tracking, manage virtual API keys, enforce rate limits, and gain full visibility over model usage. This module deploys LiteLLM on Cloud Run v2 on top of the App_CloudRun foundation, which provisions and manages the shared Google Cloud infrastructure.

This guide focuses on the cloud services LiteLLM uses and how to explore and operate them from the Google Cloud Console and the command line. For the mechanics common to every Cloud Run application — service identity, ingress and load balancing, scaling and concurrency, CI/CD, Cloud Armor, IAP, Binary Authorization, VPC Service Controls, backups, and the deployment lifecycle — refer to the App_CloudRun foundation guide rather than repeating them here.

---

What LiteLLM costs on RAD, and how that compares

LiteLLM on RAD's Cloud Run module costs about US$77 a month in your own project, with no per-request metering. RAD charges 75 credits once per deployment in your own project (67.5 credits, 10% lower, in a project RAD manages), plus build time. LiteLLM also deploys on GKE Autopilot for about US$129 a month — see the GKE guide if you need it to run there instead. If you only need it occasionally, a RAD-managed project can be deleted and restored within 30 days for a few credits, so that monthly figure becomes a few dollars instead — see Pause it for free, below.

What you pay on RAD

	In a Google Cloud project you own	In a project RAD manages for you
Module fee, once per deployment	75 credits (US$7.50 at the top-up price)	67.5 credits (10% lower)
Build time	A few credits per build	The same
Google Cloud running cost	Billed by Google to your own billing account, about US$77 a month for the default configuration (table below)	Metered hourly in credits; RAD publishes 34 credits a day, about 1,020 a month (about US$102.00 at the top-up price, less on a plan)
Who owns what	You own the project and its billing; RAD deploys and updates it	RAD owns the project, with guardrails, quotas and budget alerts; you get console access, and a minimum purchased balance is held in reserve

10 credits cost US$1 on a one-off top-up, and less on a monthly plan. In a project RAD manages, the database, file server and network are shared by every application in the project, so a second application does not add a second database.

Default running cost in your own project (us-central1, Google list prices):

Resource (module default)	Per month
Cloud Run, 1x vCPU / 2 GiB (warm)	US$20
Cloud SQL for PostgreSQL, 1 vCPU / 3.75 GB, zonal, 10 GB SSD	US$51
Cloud NAT and networking	US$5
Cloud Storage (add-ons, backups)	US$1
Total	about US$77

How it compares

• Portkey, a hosted LLM gateway offering a similar unified-API/cost-tracking feature set, costs US$49 a month on its Production plan, covering 100,000 logged requests a month, with usage above that billed in US$9-per-100k-request increments.
• RAD's LiteLLM module costs about US$77 a month in your own project — more than Portkey at light volume, but there is no per-request metering, and every prompt, response and provider API key stays inside infrastructure you control rather than passing through a third party's gateway.
• Self-managed comparison: its 1 vCPU / 2 GiB Cloud Run service is close to a Hetzner CPX22 (2 vCPU/4 GB, ~US$24/month) or a DigitalOcean 2 vCPU/4 GB droplet (US$24/month) — but then you run PostgreSQL, security patches and the gateway's own uptime yourself; RAD's managed Cloud SQL, Secret Manager and monitoring cover that.

Pause it for free: delete a RAD-managed project, restore it when you need it

If LiteLLM runs in a project RAD manages for you, you have a second option that goes well beyond scaling to zero: delete the whole project, and restore it within 30 days for close to nothing. This suits LiteLLM you only need occasionally — evaluating it, a demo environment, a seasonal or intermittent workload — far better than running it continuously.

• How it works. Deleting a RAD-managed project unlinks its billing first, then asks Google to delete the project. Google does not remove the project immediately: it keeps it, recoverable, for 30 days. Because billing is already unlinked, nothing is charged while it waits. Unlike deleting one module, this does not tear down the database, any VM or the compute resource one by one — the whole project, and everything in it, simply stops.
• Restoring costs a few credits, not a rebuild. Within 30 days, the project's owner can restore it. RAD asks Google to undelete the project and reattaches its billing account, then asks you to run Update on each deployment to confirm everything came back. Because nothing was individually destroyed, that Update finds the same resources already there — it is a check, not a rebuild, and an Update never charges the module fee again. For LiteLLM behind Services_GCP (a Postgres database) and Project_GCP, that is roughly a handful of credits (under US$1) in total, against the about 80 credits (US$8.00) a full redeploy costs.
• So a month of occasional use can cost a few dollars, not US$77. Deploy LiteLLM, use it for a while, delete the project. Restore it next time you need it, confirm with Update, and delete it again when you're done. You pay only for the module fee once, the builds, and whatever time it was actually live.
• What this needs. You must own the project (not one RAD only manages billing for), and you restore it yourself within the 30 days — after that, Google deletes it for good. Restoring is admitted like creating a new project: your purchased credit balance must still clear the tier's floor (100 credits for the sandbox tier most evaluation use fits). Google says most services are fully working again within 36 hours of a restore.
• One real gap: nightly backups do not survive. Backups and other generated files are written to a bucket inside the project, and that bucket is not protected by Cloud Storage's soft-delete, so it is very likely gone as soon as you delete the project — even though the project itself is recoverable for 30 days. If you have customised LiteLLM and want to keep that work, copy a backup out (to Google Drive, or a bucket outside the project) before you delete. For a default installation with nothing irreplaceable in it, this does not matter.

Pay only while you use it, the other way: delete and redeploy

The option above only applies to a RAD-managed project; in your own project, or once the 30-day window has passed, the way to stop paying is to delete the deployment and deploy it again when you need it.

• What a redeploy costs. The module fee again, plus the builds — roughly 80 credits (US$8.00) in total. Deleting saves money only once LiteLLM would otherwise sit unused long enough to clear that redeploy cost against its own running cost — about 4 days or more, both in your own project (about US$2.56 a day) and in a RAD-managed one (34 credits a day).
• Most of the running cost is usually shared. For LiteLLM that is the database. They stop only when nothing else in the project uses them, so deleting this app while something else shares the project saves only this app's own compute part.
• Keep your data first. Nightly backups go to a bucket inside the deployment and are deleted with it, so copy the latest backup out before deleting if you want to keep it.

Lab sessions and Managed Environments

• Lab sessions, for training. A trainer runs a session for a class. Each participant gets the app in their own Google Cloud project for 15 minutes to 24 hours, within an allowance the trainer sets. Either the trainer funds every place, or each participant pays for their own. Everything is deleted when the session ends and unused credits go back to the trainer.
• Managed Environments, for consultancies. A partner runs the app for a client from a ring-fenced wallet it funds, and settles with the client directly. If the wallet runs low, billing pauses and the data is kept, so nobody receives an unexpected charge. At the end the partner hands the project over and the deployments become the client's own.

Credits can be bought in more than 20 currencies, including XAF, XOF, NGN, GHS, KES and ZAR, by card, bank transfer or mobile money.

Sources (8 October 2026): Google Cloud list prices from the Cloud Billing Catalog API; RAD fees and the daily-credit estimate from radmodules.dev/pricing; Delete and restore projects for the pause/restore mechanics and what it says about Cloud Storage objects without soft delete; Portkey. Prices change; check each source before relying on a figure.

1. Overview

LiteLLM runs as a Python-based proxy container on Cloud Run v2. The deployment wires together a focused set of Google Cloud services:

Capability	Google Cloud service	Notes
Compute	Cloud Run v2	Python proxy service, 1 vCPU / 2 GiB by default, request-based billing
Database	Cloud SQL for PostgreSQL 15	Required — LiteLLM's Prisma ORM uses PostgreSQL for virtual keys and spend tracking
Object storage	Cloud Storage	Optional — no buckets created by default
Cache	Redis	Optional — reduces latency and cost for repeated identical LLM requests
Secrets	Secret Manager	Auto-generated master key and salt key; LLM provider API keys injected at runtime
Ingress	Cloud Run URL / Cloud Load Balancing	Default run.app URL, optional external HTTPS load balancer + custom domain

Sensible defaults worth knowing up front:

• PostgreSQL 15 is mandatory. LiteLLM's Prisma ORM requires PostgreSQL for virtual key management and spend tracking; changing the engine breaks startup.
• A custom container image is built by Cloud Build. The image embeds an entrypoint.sh that assembles DATABASE_URL from the DB_* environment variables injected by the foundation at runtime.
• LITELLM_MASTER_KEY and LITELLM_SALT_KEY are auto-generated and stored in Secret Manager. The salt key must never be rotated after virtual keys have been issued — all existing virtual keys would become permanently invalid.
• STORE_MODEL_IN_DB = "true" is set automatically, enabling runtime model management and the Admin UI without container restarts.
• Redis is disabled by default. Enable it for multi-instance deployments to share rate-limit counters and response caches.
• The startup probe targets /health/readiness, which validates database connectivity and confirms Prisma migrations have completed before the service is marked ready.

---

2. Google Cloud Services & How to Explore Them

All commands assume PROJECT and REGION are set. Service and resource names are reported in the deployment Outputs.

A. Cloud Run — the LiteLLM service

LiteLLM runs as a Cloud Run v2 service that autoscales by request load between the minimum and maximum instance counts. Each deployment creates an immutable revision; traffic can be split across revisions for safe rollouts.

• Console: Cloud Run → select the service for revisions, traffic, logs, and metrics.
• CLI:

  gcloud run services list --project "$PROJECT" --region "$REGION"
  gcloud run services describe <service-name> --project "$PROJECT" --region "$REGION"
  gcloud run revisions list --service <service-name> --project "$PROJECT" --region "$REGION"

See App_CloudRun for scaling, concurrency, execution environment, and traffic splitting.

B. Cloud SQL for PostgreSQL 15

LiteLLM stores all virtual keys, usage logs, cost records, and model routing rules in a managed Cloud SQL for PostgreSQL 15 instance. The service connects privately through the Cloud SQL Auth Proxy over a Unix socket (no public IP). On first deploy an initialization job creates the application database and user.

• Console: SQL → select the instance for connections, backups, flags, metrics.
• CLI:

  gcloud sql instances list --project "$PROJECT"
  gcloud sql instances describe <instance-name> --project "$PROJECT"
  gcloud sql connect <instance-name> --user=<db-user> --database=litellm_db --project "$PROJECT"

The instance name, database, user, and password secret are in the Outputs. See App_CloudRun for the connection model, backups, and password rotation.

C. Cloud Storage

No storage buckets are created by default. Buckets can be declared via storage_buckets and mounted via GCS Fuse using gcs_volumes — for example, to deliver a config.yaml to the container without rebuilding the image.

• Console: Cloud Storage → Buckets.
• CLI:

  gcloud storage buckets list --project "$PROJECT"
  gcloud storage ls gs://<bucket-name>/

See App_CloudRun for GCS Fuse and CMEK options.

D. Redis cache

Redis backs LiteLLM's optional response caching and shared rate-limit counters. When enable_redis = true, the REDIS_HOST, REDIS_PORT, and (optionally) REDIS_PASSWORD environment variables are injected into the service automatically.

• Console: Memorystore → Redis (if using a managed instance).
• CLI:

  redis-cli -h <redis-host> ping
  redis-cli -h <redis-host> info keyspace

E. Secret Manager

LITELLM_MASTER_KEY (the primary admin API key, prefixed sk-) and LITELLM_SALT_KEY (used to hash virtual keys) are generated automatically and stored in Secret Manager. LLM provider API keys are injected by referencing pre-existing secrets via secret_environment_variables.

• Console: Security → Secret Manager.
• CLI:

  gcloud secrets list --project "$PROJECT"
  # Retrieve the master key:
  gcloud secrets versions access latest --secret=<master-key-secret> --project "$PROJECT"

See App_CloudRun for injection and rotation details.

F. Networking & ingress

The service is reachable at its run.app URL by default. An external HTTPS load balancer with a custom domain, Cloud CDN, and Cloud Armor can be layered on; ingress settings and VPC egress control connectivity.

• Console: Cloud Run (service URL); Network services → Load balancing.
• CLI:

  gcloud run services describe <service-name> --region "$REGION" --format='value(status.url)'
  gcloud compute addresses list --project "$PROJECT"

See App_CloudRun.

G. Cloud Logging & Monitoring

Container logs flow to Cloud Logging; Cloud Run and Cloud SQL metrics flow to Cloud Monitoring, with optional uptime checks and alert policies.

• Console: Logging → Logs Explorer; Monitoring → Dashboards / Alerting.
• CLI:

  gcloud run services logs read <service-name> --project "$PROJECT" --region "$REGION" --limit 50

---

3. LiteLLM Application Behaviour

• First-deploy database setup. An initialization job creates the LiteLLM database and user before the service starts. It is idempotent and connects to Cloud SQL through the Auth Proxy Unix socket.

• Prisma migrations on start. LiteLLM runs its Prisma ORM migrations on each instance start, so upgrading the version applies schema changes automatically. The startup probe waits until /health/readiness returns 200, confirming migrations are complete before routing traffic to the service.

• Admin UI. The LiteLLM Admin UI is available at /ui on the service URL. Authenticate with the LITELLM_MASTER_KEY (retrieve it from Secret Manager). From the UI you can add models, create virtual keys, set budgets, and view usage dashboards — all without redeploying the service.

• Adding LLM provider keys. Provider API keys are not managed by this module. Supply them at deploy time via secret_environment_variables (mapping each env var to a pre-existing Secret Manager secret), or add them after deployment via the Admin UI or the /model/new API endpoint using the master key.

• Virtual key management. Use the /key/generate API with the master key to issue per-team or per-user virtual keys with rate limits and spend budgets. These keys are stored in PostgreSQL and salted with LITELLM_SALT_KEY.

  # Retrieve the master key then create a virtual key:
  MASTER_KEY=$(gcloud secrets versions access latest --secret=<master-key-secret> --project "$PROJECT")
  curl -X POST "https://<service-url>/key/generate" \
    -H "Authorization: Bearer $MASTER_KEY" \
    -H "Content-Type: application/json" \
    -d '{"key_alias": "team-a", "max_budget": 10.0}'

• IAP and programmatic API calls. IAP (enable_iap = true) requires a browser OAuth flow and blocks direct LLM API calls. Use IAP only if restricting access to the Admin UI; for API gateway use, prefer ingress_settings = "internal" with VPN or mutual authentication.

• Health endpoints. /health/readiness validates database connectivity and Prisma migration completion; /health/liveliness confirms the proxy process is running. These are used as the startup and liveness probes respectively.

---

4. Configuration Variables

Variables are grouped exactly as they appear on the deployment platform. Only settings specific to or notable for LiteLLM are listed; every other input is inherited from App_CloudRun with its standard behaviour.

Group 1 — Project & Identity

Variable	Default	Description
project_id	(required)	Target Google Cloud project.
region	us-central1	Region for the service and regional resources.

Group 2 — Deployment Environment

Variable	Default	Description
tenant_id	demo	Short suffix that makes resource names unique per environment.
support_users	[]	Emails granted project access and monitoring alerts.
resource_labels	{}	Labels applied to all resources.

Group 3 — Application Identity

Variable	Default	Description
application_name	litellm	Base name for resources. Do not change after first deploy.
display_name	LiteLLM AI Gateway	Friendly name shown in the Console.
description	(set)	Service description.
application_version	main-stable	LiteLLM image version tag; pin to a specific release for production stability.

Group 4 — Runtime & Scaling

Variable	Default	Description
deploy_application	true	Set false to provision infrastructure only.
cpu_limit	1000m	CPU per instance.
memory_limit	2Gi	Memory per instance; do not shrink below 2Gi — LiteLLM OOM-crashes on startup below that.
cpu_always_allocated	false	Request-based billing by default — LiteLLM is a stateless proxy with no in-process background work.
min_instance_count	1	Must stay ≥ 1. LiteLLM's scheduled jobs (including the budget reset) run only while an instance is alive; at 0 they stop when Cloud Run scales the service down.
max_instance_count	3	Maximum instances.
container_port	4000	LiteLLM's native port.
execution_environment	gen2	Cloud Run execution generation; gen2 is required for NFS and Direct VPC Egress.
timeout_seconds	600	Max request duration; increase for long-running LLM inference calls (max 3600).
enable_cloudsql_volume	true	Cloud SQL Auth Proxy for Unix socket connections.
traffic_split	[]	Split traffic across revisions for staged rollouts.

Group 5 — Access & Ingress Control

Variable	Default	Description
ingress_settings	all	Which networks may reach the service; set to internal for VPC-only API gateway use.
vpc_egress_setting	PRIVATE_RANGES_ONLY	How outbound traffic is routed through the VPC connector.
enable_iap	false	Require Google sign-in via Identity-Aware Proxy (blocks direct API calls).
iap_authorized_users / iap_authorized_groups	[]	Who may access through IAP.

Group 6 — Environment Variables & Secrets

Variable	Default	Description
environment_variables	{ LITELLM_LOG="INFO", NUM_WORKERS="1" }	Extra non-secret settings. Core LiteLLM vars are set automatically.
secret_environment_variables	{}	Map of env var → Secret Manager secret name. Use to inject LLM provider API keys.
secret_propagation_delay / secret_rotation_period	(set)	Replication wait / rotation cadence.

Group 7 — Backup & Restore

Variable	Default	Description
backup_schedule	0 2 * * *	Automated backup cron (UTC).
backup_retention_days	7	Retention; raise for production.
enable_backup_import / backup_source / backup_uri / backup_format	restore options	Restore from a backup on deploy.

Group 8 — CI/CD & Binary Authorization

Standard App_CloudRun Cloud Build / Cloud Deploy integration — see App_CloudRun. Key inputs: enable_cicd_trigger, github_repository_url, github_token, enable_cloud_deploy, enable_binary_authorization.

Group 9 — Custom SQL

enable_custom_sql_scripts, custom_sql_scripts_bucket, custom_sql_scripts_path, custom_sql_scripts_use_root — run SQL from a GCS bucket after provisioning. See App_CloudRun.

Group 10 — Domain, CDN, Cloud Armor & Image Retention

Variable	Default	Description
enable_cloud_armor	false	Provision Global HTTPS LB + Cloud Armor WAF.
admin_ip_ranges	[]	CIDR ranges exempted from WAF rules.
application_domains	[]	Custom hostnames for the external load balancer.
enable_cdn	false	Enable Cloud CDN on the LB backend.
max_images_to_retain / delete_untagged_images / image_retention_days	(set)	Artifact Registry cleanup policy.

Group 11 — Storage & Filesystem

Variable	Default	Description
create_cloud_storage	true	Provision buckets declared in storage_buckets.
storage_buckets	[]	No buckets created by default.
enable_nfs	false	NFS is not required for LiteLLM; enable only for shared config file delivery.
nfs_mount_path	/mnt/nfs	Mount path inside the container.
gcs_volumes	[]	GCS Fuse volume mounts for config file delivery.
manage_storage_kms_iam / enable_artifact_registry_cmek	false	CMEK options.

Group 12 — Database Backend

Variable	Default	Description
database_type	POSTGRES_15	Fixed — do not change; LiteLLM requires PostgreSQL 15.
db_name	litellm_db	Database name. Immutable after first deploy.
db_user	litellm_user	Application user. Immutable after first deploy.
database_password_length	32	Generated password length (16–64).
enable_auto_password_rotation / rotation_propagation_delay_sec	off	DB password rotation.
db_host_env_var_name / db_name_env_var_name / db_user_env_var_name / db_port_env_var_name / service_url_env_var_name	(set)	Additional env var names under which connection details are injected.

Group 13 — Jobs & Scheduled Tasks

Variable	Default	Description
initialization_jobs	[]	Leave empty to use the built-in database setup job from LiteLLM_Common.
cron_jobs	[]	Recurring Cloud Run Jobs for maintenance or housekeeping tasks.

Group 14 — Observability & Health

Variable	Default	Description
startup_probe	/health/readiness	HTTP probe; validates DB connectivity and Prisma migrations before marking the service ready.
liveness_probe	/health/liveliness	HTTP probe; confirms the proxy process is running.
uptime_check_config	disabled	Cloud Monitoring uptime check against /health/liveliness; enable explicitly to activate.
alert_policies	[]	Metric alert policies.

Group 21 — Redis Cache

Variable	Default	Description
enable_redis	false	Enable Redis for response caching and shared rate-limit counters.
redis_host	""	Redis endpoint; required when enable_redis = true.
redis_port	6379	Redis port.
redis_auth	""	Optional Redis auth password (sensitive).

Group 22 — VPC Service Controls & Audit Logging

Variable	Default	Description
enable_vpc_sc	false	Enforce a VPC-SC perimeter (requires organization_id).
vpc_cidr_ranges / vpc_sc_dry_run	(set)	Access level CIDRs / dry-run mode.
enable_audit_logging	false	Detailed Cloud Audit Logs.

---

5. Outputs

Returned on a successful deployment — the quickest way to locate and explore the running resources.

Output	Description
service_name	Cloud Run service name.
service_url	Default run.app URL of the service.
service_location	Region the service runs in.
stage_services	Stage-specific service URLs (Cloud Deploy).
load_balancer_ip / load_balancer_url	External HTTPS load balancer IP / URL (when enabled).
database_instance_name	Cloud SQL instance name.
database_name / database_user	Application database name / user.
database_password_secret	Secret Manager secret holding the DB password.
database_host / database_port	DB endpoint / port.
storage_buckets	Created Cloud Storage buckets.
network_name / network_exists / regions	VPC network, presence, regions.
container_image / container_registry	Deployed image and Artifact Registry repo.
monitoring_enabled / monitoring_notification_channels / uptime_check_names	Monitoring status, channels, uptime checks.
initialization_jobs	Names of the setup jobs.
deployment_id / tenant_id / resource_prefix	Naming identifiers.
project_id / project_number	Project identifiers.
cicd_enabled / github_repository_url / github_repository_owner / github_repository_name / cicd_configuration	CI/CD status and details.
artifact_registry_repository / cloudbuild_trigger_name / cloudbuild_trigger_id	Registry and build trigger.
vpc_sc_enabled / vpc_sc_perimeter_name / vpc_sc_dry_run_mode	VPC-SC status.
audit_logging_enabled / artifact_registry_cmek_enabled	Audit logging and CMEK status.

---

6. Configuration Pitfalls & Sensible Defaults

Risk: Critical (data loss / outage / security) — High (service degraded) — Medium (cost or partial degradation) — Low (minor).

Setting	Sensible value	Risk	Consequence if wrong
database_type	POSTGRES_15	Critical	LiteLLM requires PostgreSQL; changing the engine breaks the Prisma ORM and prevents startup.
enable_cloudsql_volume	true	Critical	The Auth Proxy sidecar is required for database connectivity; disabling it causes Prisma to fail at startup.
LITELLM_SALT_KEY	auto-generated, never rotated	Critical	Rotating the salt key invalidates every previously issued virtual key; all API consumers lose access immediately.
db_name / db_user	set once	Critical	Immutable after first deploy; renaming recreates the DB/user and destroys all virtual keys and spend data.
enable_backup_import	false unless restoring	Critical	Enabling without a valid backup_uri fails the import job.
ingress_settings	restrict for production	Critical	"all" exposes the master key endpoint publicly; use "internal" for VPC-only API gateway deployments.
LITELLM_MASTER_KEY	auto-generated	High	Treat as a credential; rotating it breaks all existing integrations holding the key until they are updated.
enable_redis	true for multi-instance	High	Without Redis, rate-limit counters are per-instance and not shared; quotas are not enforced across replicas.
redis_host	set when Redis enabled	High	An empty host with enable_redis = true causes connection errors on every request.
min_instance_count	1	High	At 0, budget resets and LiteLLM's other scheduled jobs stop when the instance is reaped; cold starts also add 20–40 s latency for dependent services.
timeout_seconds	600	High	Large language model inference can take minutes; too-short timeout causes 504 errors on slow models.
enable_iap	false for API endpoints	High	IAP blocks all direct programmatic API calls; only use IAP if accessing the Admin UI exclusively.
execution_environment	gen2	High	NFS mounts and Direct VPC Egress are gen2-only; downgrading breaks networking.
application_version	pin for production	Medium	LiteLLM releases frequently; unpinned versions may change the Prisma schema or break virtual key formats.
NUM_WORKERS	1 (raise for throughput)	Medium	A single worker serialises all requests; increase to 2–4 and scale cpu_limit proportionally for high-traffic gateways.
backup_retention_days	7 (raise for prod)	Medium	Too short for compliance retention.
enable_auto_password_rotation	false unless ready	Medium	Enabling without adequate rotation_propagation_delay_sec can cause a race condition where the service restarts before the new password propagates.

---

For the foundation behaviour referenced throughout — service identity, scaling and concurrency, ingress and load balancing, CI/CD, Cloud Armor, IAP, Binary Authorization, VPC-SC, backups, and image mirroring — see App_CloudRun. LiteLLM-specific application configuration shared with the GKE variant is described in LiteLLM_Common.

Related guides

• Hands-on lab: LiteLLM on Cloud Run — deploy it step by step, with the console screens and commands at each stage.
• LiteLLM on GKE Autopilot — the same application on Kubernetes, for when you need the other deployment target.
• LiteLLM Common — Shared Application Configuration — the configuration shared by both deployment targets.
• Deployed alongside Ollama on Google Cloud Run, Qdrant on Google Cloud Run, Open WebUI on Google Cloud Run, SearXNG on Google Cloud Run in the Private AI Assistant solution.