AnythingLLM on Google Cloud Run

This document provides a comprehensive reference for the modules/AnythingLLM_CloudRun Terraform module. It covers architecture, IAM, configuration variables, AnythingLLM-specific behaviours, and operational patterns for deploying AnythingLLM on Google Cloud Run (v2).

---

1. Module Overview

AnythingLLM is a private AI workspace and Retrieval-Augmented Generation (RAG) platform. It allows teams to chat with documents, connect to any LLM provider (OpenAI, Anthropic, Ollama, and others), and build AI-powered knowledge assistants — without sending data to third-party services. AnythingLLM CloudRun is a wrapper module built on top of App CloudRun. It uses App CloudRun for all GCP infrastructure provisioning and injects AnythingLLM-specific application configuration, secrets, and storage configuration via AnythingLLM Common.

Key Capabilities:

• Compute: Cloud Run v2 (Gen2 recommended), 2 vCPU / 4 Gi by default. min_instance_count = 1 to keep AnythingLLM warm for AI workloads.
• Data Persistence: Cloud SQL PostgreSQL 15 (required by AnythingLLM's Prisma ORM). GCS document storage bucket auto-provisioned by AnythingLLM Common. Optional NFS for persistent document/vector store data.
• Security: Four application-level secrets are auto-generated by AnythingLLM Common — JWT_SECRET, AUTH_TOKEN, SIG_KEY, and SIG_SALT — and injected into the container via Secret Manager. Inherits Cloud Armor WAF, IAP, Binary Authorization, and VPC Service Controls from App CloudRun.
• Caching: Redis is disabled by default (enable_redis = false). Enable for session or cache workloads if required.
• CI/CD: Cloud Build custom image pipeline by default; Cloud Deploy progressive delivery optional.
• Reliability: Health probes target /api/ping with a 60-second initial delay to allow model loading and database migration on first boot.

Project & Application Identity

Variable	Group	Type	Default	Description
project_id	1	string	—	GCP project ID. Required.
tenant_deployment_id	2	string	'demo'	Short suffix appended to all resource names.
support_users	2	list(string)	[]	Email recipients for monitoring alerts.
resource_labels	2	map(string)	{}	Labels applied to all provisioned resources.
application_name	3	string	'anythingllm'	Base resource name. Do not change after initial deployment.
application_display_name	3	string	'AnythingLLM'	Human-readable name shown in dashboards.
application_description	3	string	'AnythingLLM Private AI Workspace on Cloud Run'	Service description.
application_version	3	string	'latest'	Container image version tag.

Wrapper architecture: AnythingLLM CloudRun calls AnythingLLM Common to build an application_config object containing AnythingLLM-specific environment variables, probe configuration, and the db-init job definition. AnythingLLM Common generates and stores JWT_SECRET, AUTH_TOKEN, SIG_KEY, and SIG_SALT in Secret Manager and returns their IDs via module.anythingllm_app.secret_ids. The GOOGLE_CLOUD_STORAGE_BUCKET_NAME environment variable is automatically set from module.anythingllm_app.storage_buckets[0].name. The scripts_dir is resolved to abspath("${module.anythingllm_app.path}/scripts") at apply time.

PostgreSQL note: AnythingLLM uses Prisma ORM and requires PostgreSQL. database_type = "POSTGRES_15" is the default. Do not set this to a MySQL or SQL Server variant.

---

2. IAM & Access Control

AnythingLLM CloudRun delegates all IAM provisioning to App CloudRun. The Cloud Run SA, Cloud Build SA, IAP service agent, and password rotation role sets are identical to those in App CloudRun.

Application secrets: AnythingLLM Common auto-generates four secrets on first apply:

• JWT_SECRET — signs AnythingLLM authentication tokens.
• AUTH_TOKEN — optional API bearer token for programmatic access.
• SIG_KEY — HMAC signing key for request signatures (32+ characters).
• SIG_SALT — salt used alongside SIG_KEY for HMAC signatures (32+ characters).

These secrets are stored in Secret Manager and injected natively at Cloud Run revision start. Plaintext is never written to Terraform state.

Database initialisation identity: The db-init Cloud Run Job runs under the Cloud Run SA. It connects to Cloud SQL PostgreSQL via the Auth Proxy Unix socket (since enable_cloudsql_volume = true by default), using DB_HOST (the socket path under /cloudsql), DB_USER, and ROOT_PASSWORD (from Secret Manager).

120-second IAM propagation delay: Inherited from App CloudRun — the AnythingLLM service is not deployed until the delay completes, preventing secret-read failures on the first revision start.

---

3. Core Service Configuration

A. Compute (Cloud Run)

AnythingLLM is a Node.js application that performs AI embedding and LLM inference operations. It has higher resource requirements than typical web applications.

min_instance_count = 1 is recommended to keep AnythingLLM warm and avoid cold starts for AI document ingestion and chat operations. max_instance_count = 1 is the default — AnythingLLM's local storage model means multiple instances require NFS for consistent document access.

Startup CPU Boost is always enabled (hardcoded in App CloudRun).

Container image: container_image_source defaults to 'custom', meaning Cloud Build compiles a custom AnythingLLM image using AnythingLLM_Common's Dockerfile. Set container_image_source = 'prebuilt' and container_image = '<image-uri>' to skip the build and deploy an existing image.

Variable	Group	Default	Description
deploy_application	4	true	Set false for infrastructure-only deployment (SQL, storage, secrets).
container_image_source	4	'custom'	'custom' builds via Cloud Build. 'prebuilt' deploys an existing image URI.
container_image	4	""	Override image URI. Leave empty for Cloud Build to manage the image.
container_build_config	4	{ enabled=true }	Cloud Build configuration: dockerfile_path, context_path, build_args, artifact_repo_name.
enable_image_mirroring	4	true	Mirrors the container image into Artifact Registry before deployment.
cpu_limit	4	'2000m'	CPU per instance. 2 vCPU minimum for AI workloads.
memory_limit	4	'4Gi'	Memory per instance. 4 Gi minimum for embedding and inference.
min_instance_count	4	1	Minimum instances. Set to 1 to avoid cold starts.
max_instance_count	4	1	Maximum instances. Increase with NFS for shared document access.
container_port	4	3001	AnythingLLM's native HTTP port.
container_protocol	4	'http1'	'http1' or 'h2c'.
execution_environment	4	'gen2'	Gen2 required for NFS mounts and GCS Fuse.
timeout_seconds	4	300	Max request duration. Increase for long-running document ingestion.
enable_cloudsql_volume	4	true	Injects Cloud SQL Auth Proxy sidecar for database connectivity.
cloudsql_volume_mount_path	4	'/cloudsql'	Container path for the Auth Proxy Unix socket.
traffic_split	4	[]	Percentage-based canary/blue-green traffic allocation.
service_annotations	4	{}	Advanced Cloud Run annotations.
service_labels	4	{}	Labels applied to the Cloud Run service.

Differences from App CloudRun defaults:

Variable	App CloudRun	AnythingLLM CloudRun	Reason
container_port	8080	3001	AnythingLLM's native port.
cpu_limit	'1000m'	'2000m'	AI workloads require more CPU for embedding and inference.
memory_limit	'512Mi'	'4Gi'	LLM context, document vectors, and Prisma ORM require more RAM.
min_instance_count	0	1	Keep warm for AI operations; cold starts are expensive.
health probe path	'/healthz'	'/api/ping'	AnythingLLM's health endpoint.

B. Database (Cloud SQL — PostgreSQL 15)

AnythingLLM uses Prisma ORM and requires PostgreSQL. database_type = "POSTGRES_15" is the default set by AnythingLLM Common. The entrypoint script constructs the DATABASE_URL Prisma connection string from the platform-injected DB_* variables at container start time, working correctly on both Unix socket (Cloud Run) and TCP (GKE) connections.

Variable	Group	Default	Description
database_type	12	'POSTGRES_15'	Cloud SQL engine. AnythingLLM requires PostgreSQL.
application_database_name	12	'anythingllmdb'	PostgreSQL database name. Do not change after initial deployment.
application_database_user	12	'anythingllmuser'	Database user. Password auto-generated and stored in Secret Manager.
database_password_length	12	32	Auto-generated password length. Range: 16–64.
enable_auto_password_rotation	12	false	Automated zero-downtime password rotation.
rotation_propagation_delay_sec	12	90	Seconds to wait after rotation before restarting the service.
db_host_env_var_name	12	""	Additional env var name to expose DB_HOST. Leave empty to inject only DB_HOST.
db_user_env_var_name	12	""	Additional env var name to expose DB_USER.
db_name_env_var_name	12	""	Additional env var name to expose DB_NAME.
db_port_env_var_name	12	""	Additional env var name to expose DB_PORT.
service_url_env_var_name	12	""	Additional env var name to expose CLOUDRUN_SERVICE_URL.

C. Storage (NFS & GCS)

NFS is disabled by default (enable_nfs = false). For multi-instance deployments or persistent document/vector store access, enable NFS. Requires execution_environment = 'gen2'.

GCS document bucket: AnythingLLM Common automatically provisions a dedicated anythingllm-docs GCS bucket. The GOOGLE_CLOUD_STORAGE_BUCKET_NAME environment variable is set to this bucket name automatically. This bucket is separate from any additional buckets in storage_buckets.

Variable	Group	Default	Description
create_cloud_storage	11	true	Set false to skip bucket creation. The anythingllm-docs bucket from AnythingLLM Common is always provisioned.
storage_buckets	11	[{ name_suffix = "data" }]	Additional GCS buckets beyond the auto-provisioned docs bucket.
enable_nfs	11	false	Provisions a Cloud Filestore NFS instance. Required for multi-instance AnythingLLM deployments.
nfs_mount_path	11	'/mnt/nfs'	Container path where the NFS share is mounted.
nfs_instance_name	11	""	Name of an existing NFS GCE VM. Leave empty to auto-discover.
nfs_instance_base_name	11	'app-nfs'	Base name for an inline NFS GCE VM when none exists.
gcs_volumes	11	[]	GCS buckets to mount via GCS Fuse (requires gen2).
manage_storage_kms_iam	11	false	Creates a CMEK KMS keyring/key and enables CMEK on all storage buckets.
enable_artifact_registry_cmek	11	false	Creates an Artifact Registry KMS key and enables at-rest CMEK encryption.

D. Networking

Cloud Run uses Direct VPC Egress to reach Cloud SQL's internal IP. With enable_cloudsql_volume = true, the Auth Proxy sidecar handles the Cloud SQL connection via Unix socket.

Variable	Group	Default	Description
ingress_settings	5	'all'	'all' — public internet; 'internal' — VPC only; 'internal-and-cloud-load-balancing' — forces traffic through the HTTPS LB.
vpc_egress_setting	5	'PRIVATE_RANGES_ONLY'	'PRIVATE_RANGES_ONLY' routes only RFC 1918 traffic via VPC. 'ALL_TRAFFIC' routes all egress via VPC.

E. Initialization & Bootstrap

A db-init Cloud Run Job is automatically provisioned by AnythingLLM Common when initialization_jobs is left as the default empty list ([]). It uses the postgres:15-alpine image and executes AnythingLLM_Common/scripts/create-db-and-user.sh, which performs the following idempotent operations:

1. Connects to Cloud SQL PostgreSQL via the Auth Proxy Unix socket.
2. Creates the AnythingLLM database user with the password from Secret Manager.
3. Creates the AnythingLLM database if it does not exist.
4. Grants the user full privileges on the database.

Override initialization_jobs with a non-empty list to replace this default with custom jobs.

Variable	Group	Default	Description
initialization_jobs	13	[]	One-shot Cloud Run Jobs. Leave empty for AnythingLLM Common to supply the default db-init job. Each entry: name, description, image, command, args, env_vars, secret_env_vars, cpu_limit, memory_limit, timeout_seconds, max_retries, task_count, execution_mode, mount_nfs, mount_gcs_volumes, depends_on_jobs, execute_on_apply, script_path.
cron_jobs	13	[]	Recurring jobs triggered by Cloud Scheduler. Each entry: name, schedule (cron UTC), image, command, args, env_vars, secret_env_vars, cpu_limit, memory_limit, timeout_seconds, max_retries, task_count, parallelism, mount_nfs, mount_gcs_volumes, script_path, paused.
additional_services	13	[]	Additional Cloud Run services deployed alongside AnythingLLM.

---

4. Advanced Security

A. Cloud Armor WAF

When enable_cloud_armor = true, a Global HTTPS Load Balancer with a Cloud Armor WAF policy (OWASP Top 10, adaptive DDoS, 500 req/min rate limiting) is provisioned in front of Cloud Run.

Variable	Group	Default	Description
enable_cloud_armor	10	false	Provisions Global HTTPS LB + Cloud Armor WAF. Required for custom domains and DDoS protection.
admin_ip_ranges	10	[]	CIDR ranges exempted from WAF rules (e.g., office VPN, CI/CD egress IPs).

B. Identity-Aware Proxy (IAP)

When enable_iap = true, Cloud Run's native IAP integration is enabled. Google identity authentication is required before requests reach AnythingLLM. Recommended for internal AI workspace deployments.

Variable	Group	Default	Description
enable_iap	5	false	Enables IAP natively on the Cloud Run service.
iap_authorized_users	5	[]	Users/service accounts granted access. Format: 'user:email' or 'serviceAccount:sa@...'.
iap_authorized_groups	5	[]	Google Groups granted access. Format: 'group:name@example.com'.

C. Binary Authorization

When enable_binary_authorization = true, Cloud Run enforces that deployed images carry a valid cryptographic attestation.

Variable	Group	Default	Description
enable_binary_authorization	8	false	Enforces image attestation. Requires a Binary Authorization policy and attestor pre-configured in the project.

D. VPC Service Controls

When enable_vpc_sc = true, all GCP API calls from this module are bound within an existing VPC-SC perimeter.

Variable	Group	Default	Description
enable_vpc_sc	22	false	Registers module API calls within the project's VPC-SC perimeter.
vpc_cidr_ranges	22	[]	VPC subnet CIDR ranges for the VPC-SC network access level.
vpc_sc_dry_run	22	true	Logs VPC-SC violations without blocking. Set false to enforce.
organization_id	22	""	GCP Organization ID for VPC-SC. Required when enable_vpc_sc = true.
enable_audit_logging	22	false	Enables detailed Cloud Audit Logs.

E. Secret Manager Integration

Four AnythingLLM application secrets are auto-generated by AnythingLLM Common and stored in Secret Manager:

Secret	Purpose
JWT_SECRET	Signs AnythingLLM authentication tokens.
AUTH_TOKEN	Optional API bearer token for programmatic access.
SIG_KEY	HMAC signing key for request signatures (32+ characters).
SIG_SALT	Salt used alongside SIG_KEY for HMAC request signatures.

Additional user-defined secrets can be added via secret_environment_variables (e.g., LLM API keys).

Variable	Group	Default	Description
secret_environment_variables	6	{}	Map of env var name → Secret Manager secret ID. (e.g., { OPENAI_API_KEY = "anythingllm-openai-key" })
secret_rotation_period	6	'2592000s'	Frequency at which Secret Manager emits rotation notifications. Default: 30 days.
secret_propagation_delay	6	30	Seconds to wait after secret creation before dependent resources proceed.

---

5. Traffic & Ingress

A. HTTPS Load Balancer

When enable_cloud_armor = true, a Global HTTPS Load Balancer backed by a Serverless NEG is provisioned. Traffic flows: Internet → Cloud Armor → Global HTTPS LB → Serverless NEG → Cloud Run.

Setting ingress_settings = 'internal-and-cloud-load-balancing' forces all AnythingLLM traffic through the LB, preventing direct *.run.app URL access.

B. Cloud CDN

When enable_cdn = true (requires enable_cloud_armor = true), Cloud CDN is attached to the HTTPS Load Balancer backend.

AnythingLLM consideration: AnythingLLM is primarily an API-driven application with authenticated sessions. CDN caching is only appropriate for static assets. Ensure proper Cache-Control headers are in place before enabling CDN on the AnythingLLM API endpoints.

Variable	Group	Default	Description
enable_cdn	10	false	Enables Cloud CDN on the HTTPS LB backend. Only effective when enable_cloud_armor = true.
max_images_to_retain	10	7	Maximum container images to keep in Artifact Registry.
delete_untagged_images	10	true	Automatically deletes untagged images from Artifact Registry.
image_retention_days	10	30	Days after which images are eligible for deletion.

C. Custom Domains

Variable	Group	Default	Description
application_domains	10	[]	Custom domain names for the HTTPS LB. Google-managed SSL certificates provisioned per domain.

---

6. CI/CD & Delivery

A. Cloud Build Triggers

When enable_cicd_trigger = true, a Cloud Build GitHub connection and push trigger are provisioned.

Variable	Group	Default	Description
enable_cicd_trigger	8	false	Provisions a Cloud Build GitHub trigger. Requires github_repository_url and credentials.
github_repository_url	8	""	Full HTTPS URL of the GitHub repository.
github_token	8	""	GitHub PAT (repo, admin:repo_hook scopes). Required on first apply. Sensitive.
github_app_installation_id	8	""	GitHub App installation ID (preferred for organisation repos).
cicd_trigger_config	8	{ branch_pattern = "^main$" }	Advanced trigger config: branch_pattern, included_files, ignored_files, trigger_name, substitutions.

B. Cloud Deploy Pipeline

Variable	Group	Default	Description
enable_cloud_deploy	8	false	Provisions a Cloud Deploy pipeline. Requires enable_cicd_trigger = true.
cloud_deploy_stages	8	[dev, staging, prod(approval)]	Ordered promotion stages. Each: name, target_name, service_name, require_approval, auto_promote.

---

7. Reliability & Scheduling

A. Health Probes & Uptime Monitoring

AnythingLLM exposes /api/ping as its health endpoint. Both the startup and liveness probes target this path. AnythingLLM requires additional startup time for AI model loading and database migration.

Variable	Group	Default	Description
startup_probe	14	{ enabled=true, path="/api/ping", initial_delay_seconds=60, timeout_seconds=5, period_seconds=10, failure_threshold=30 }	Startup readiness probe. Container receives no traffic until this succeeds.
liveness_probe	14	{ enabled=true, path="/api/ping", initial_delay_seconds=30, timeout_seconds=5, period_seconds=30, failure_threshold=3 }	Liveness probe. Container is restarted after failure_threshold consecutive failures.
startup_probe_config	14	Same as startup_probe	Alias passed to App CloudRun health probe configuration.
health_check_config	14	Same as liveness_probe	Alias passed to App CloudRun liveness probe configuration.
uptime_check_config	14	{ enabled=true, path="/" }	Cloud Monitoring uptime check. Alerts notify support_users if unreachable.
alert_policies	14	[]	Cloud Monitoring metric alert policies.

B. Auto Password Rotation

When enable_auto_password_rotation = true, a zero-downtime password rotation pipeline is provisioned.

Variable	Group	Default	Description
enable_auto_password_rotation	12	false	Enables automated password rotation.
rotation_propagation_delay_sec	12	90	Seconds to wait after writing the new secret before restarting the service.

---

8. Integrations

A. LLM Provider Configuration

AnythingLLM connects to LLM providers via environment variables. Use environment_variables for non-sensitive provider configuration and secret_environment_variables for API keys.

Common AnythingLLM environment variables:

Variable	Purpose	Example
LLM_PROVIDER	LLM backend	openai, anthropic, ollama, native
EMBEDDING_ENGINE	Embedding provider	native, openai, ollama
VECTOR_DB	Vector database backend	lancedb, chroma, qdrant, pinecone
SERVER_PORT	HTTP port (fixed by AnythingLLM Common)	3001
STORAGE_DIR	Document storage path (fixed)	/app/server/storage

Store provider API keys in Secret Manager and reference them via secret_environment_variables:

secret_environment_variables = {
  OPENAI_API_KEY      = "anythingllm-openai-key"
  ANTHROPIC_API_KEY   = "anythingllm-anthropic-key"
}

B. Redis Cache

Redis is disabled by default for AnythingLLM. Enable it if your deployment requires session caching or a shared cache layer.

Variable	Group	Default	Description
enable_redis	21	false	Enables Redis. Not required for AnythingLLM core functionality.
redis_host	21	null	Redis server hostname or IP. Required when enable_redis = true.
redis_port	21	'6379'	Redis server TCP port (string).
redis_auth	21	""	Redis AUTH password. Sensitive — leave empty if not required.

C. Backup Import & Recovery

Variable	Group	Default	Description
backup_schedule	7	'0 2 * * *'	Cron expression (UTC) for automated daily backups.
backup_retention_days	7	7	Days to retain backup files in GCS.
enable_backup_import	7	false	Triggers a one-time restore on apply. Set false after a successful import.
backup_source	7	'gcs'	'gcs' or 'gdrive'.
backup_file	7	'backup.sql'	Filename of the backup to import.
backup_format	7	'sql'	Backup file format. Options: sql, tar, gz, tgz, tar.gz, zip, auto.

D. Custom SQL Scripts

Variable	Group	Default	Description
enable_custom_sql_scripts	9	false	Runs custom SQL scripts from a GCS bucket against the application database.
custom_sql_scripts_bucket	9	""	GCS bucket containing SQL scripts.
custom_sql_scripts_path	9	""	Path prefix within the bucket.
custom_sql_scripts_use_root	9	false	Run scripts as the root DB user.

---

9. Platform-Managed Behaviours

The following behaviours are applied automatically by AnythingLLM CloudRun regardless of variable values. They cannot be overridden via tfvars.

Behaviour	Implementation	Detail
PostgreSQL 15 required	database_type = "POSTGRES_15" fixed by AnythingLLM Common	AnythingLLM's Prisma ORM requires PostgreSQL. MySQL is not supported.
Prisma DATABASE_URL	Constructed by anythingllm-entrypoint.sh at container start	The entrypoint script builds the PostgreSQL connection string from DB_* vars injected by App CloudRun.
Application secrets auto-generated	JWT_SECRET, AUTH_TOKEN, SIG_KEY, SIG_SALT provisioned by AnythingLLM Common	Generated with random_password (32 chars, alphanumeric). Secret IDs are forwarded to App CloudRun via module_secret_env_vars.
GCS document bucket	anythingllm-docs bucket provisioned by AnythingLLM Common	The GOOGLE_CLOUD_STORAGE_BUCKET_NAME env var is set automatically in module_env_vars.
Fixed environment variables	SERVER_PORT=3001, STORAGE_DIR=/app/server/storage, UID=1000, GID=1000	Set by AnythingLLM Common. Do not override these in environment_variables.
Unix socket by default	enable_cloudsql_volume = true default	AnythingLLM connects to Cloud SQL via the Auth Proxy Unix socket.
Image mirroring enabled	enable_image_mirroring = true default	Mirrors the AnythingLLM image into Artifact Registry to avoid Docker Hub rate limits.
Default db-init job	Supplied by AnythingLLM Common when initialization_jobs = []	PostgreSQL database and user are created automatically. Override with a non-empty initialization_jobs list to replace.
Scripts directory	scripts_dir = abspath("${module.anythingllm_app.path}/scripts")	Initialization scripts are sourced from AnythingLLM Common.

---

10. Variable Reference

All user-configurable variables exposed by AnythingLLM CloudRun, sorted by UI group then order. Group 0 variables are reserved for platform metadata — leave them at their defaults for standard deployments.

Variable	Group	Default	Description
module_description	0	(AnythingLLM platform text)	Platform metadata: module description.
module_documentation	0	(docs URL)	Platform metadata: documentation URL.
module_dependency	0	['Services GCP']	Platform metadata: required modules.
module_services	0	(GCP service list)	Platform metadata: GCP services consumed.
credit_cost	0	50	Platform metadata: deployment credit cost.
require_credit_purchases	0	false	Platform metadata: enforces credit balance check.
enable_purge	0	true	Permits full deletion of module resources on destroy.
public_access	0	false	Platform catalogue visibility.
deployment_id	0	""	Deployment ID suffix. Auto-generated if empty.
resource_creator_identity	0	(platform SA)	Service account used by Terraform to manage resources.
impersonation_service_account	0	""	Service account to impersonate for GCP API calls.
project_id	1	—	GCP project ID. Required.
region	1	'us-central1'	GCP region for all resources.
tenant_deployment_id	2	'demo'	Short suffix appended to all resource names.
support_users	2	[]	Email addresses for monitoring alerts.
resource_labels	2	{}	Labels applied to all provisioned resources.
application_name	3	'anythingllm'	Base resource name. Do not change after initial deployment.
application_display_name	3	'AnythingLLM'	Human-readable name.
application_description	3	'AnythingLLM Private AI Workspace on Cloud Run'	Service description.
application_version	3	'latest'	Container image version tag.
deploy_application	4	true	Set false for infrastructure-only deployment.
container_image_source	4	'custom'	'custom' (Cloud Build) or 'prebuilt' (existing image).
container_image	4	""	Container image URI. Leave empty for Cloud Build to manage.
container_build_config	4	{ enabled=true }	Cloud Build configuration object.
enable_image_mirroring	4	true	Mirrors the container image into Artifact Registry.
cpu_limit	4	'2000m'	CPU per instance. 2 vCPU minimum for AI workloads.
memory_limit	4	'4Gi'	Memory per instance. 4 Gi minimum for AnythingLLM.
min_instance_count	4	1	Minimum instances. Set to 1 to keep AnythingLLM warm.
max_instance_count	4	1	Maximum instances.
container_port	4	3001	AnythingLLM's native port.
container_protocol	4	'http1'	'http1' or 'h2c'.
execution_environment	4	'gen2'	Gen2 required for NFS mounts and GCS Fuse.
timeout_seconds	4	300	Max request duration. Increase for document ingestion.
enable_cloudsql_volume	4	true	Injects Cloud SQL Auth Proxy sidecar.
cloudsql_volume_mount_path	4	'/cloudsql'	Container path for the Auth Proxy socket.
traffic_split	4	[]	Canary/blue-green traffic allocation.
service_annotations	4	{}	Advanced Cloud Run annotations.
service_labels	4	{}	Labels applied to the Cloud Run service.
ingress_settings	5	'all'	'all', 'internal', or 'internal-and-cloud-load-balancing'.
vpc_egress_setting	5	'PRIVATE_RANGES_ONLY'	'PRIVATE_RANGES_ONLY' or 'ALL_TRAFFIC'.
enable_iap	5	false	Enables IAP on the Cloud Run service.
iap_authorized_users	5	[]	Users/SAs granted IAP access.
iap_authorized_groups	5	[]	Google Groups granted IAP access.
environment_variables	6	{}	Plain-text env vars. Do not include secret values.
secret_environment_variables	6	{}	Secret Manager references (e.g., { OPENAI_API_KEY = "anythingllm-openai-key" }).
secret_rotation_period	6	'2592000s'	Secret rotation notification frequency.
secret_propagation_delay	6	30	Seconds to wait after secret creation.
backup_schedule	7	'0 2 * * *'	Cron expression (UTC) for automated backups.
backup_retention_days	7	7	Days to retain backup files in GCS.
enable_backup_import	7	false	Triggers a one-time restore on apply.
backup_source	7	'gcs'	'gcs' or 'gdrive'.
backup_file	7	'backup.sql'	Backup filename to import.
backup_format	7	'sql'	Backup format. Options: sql, tar, gz, tgz, tar.gz, zip, auto.
enable_cicd_trigger	8	false	Provisions a Cloud Build GitHub trigger.
github_repository_url	8	""	Full HTTPS URL of the GitHub repository.
github_token	8	""	GitHub PAT. Sensitive.
github_app_installation_id	8	""	GitHub App installation ID.
cicd_trigger_config	8	{ branch_pattern = "^main$" }	Advanced Cloud Build trigger config.
enable_cloud_deploy	8	false	Provisions a Cloud Deploy pipeline.
cloud_deploy_stages	8	[dev, staging, prod(approval)]	Ordered Cloud Deploy promotion stages.
enable_binary_authorization	8	false	Enforces image attestation on deployment.
enable_custom_sql_scripts	9	false	Runs SQL scripts from GCS after provisioning.
custom_sql_scripts_bucket	9	""	GCS bucket containing SQL scripts.
custom_sql_scripts_path	9	""	Path prefix within the bucket.
custom_sql_scripts_use_root	9	false	Run scripts as the root DB user.
enable_cloud_armor	10	false	Provisions Global HTTPS LB + Cloud Armor WAF.
admin_ip_ranges	10	[]	CIDR ranges exempted from WAF rules.
application_domains	10	[]	Custom domains with Google-managed SSL certificates.
enable_cdn	10	false	Enables Cloud CDN on the HTTPS LB backend.
max_images_to_retain	10	7	Maximum container images to keep in Artifact Registry.
delete_untagged_images	10	true	Automatically deletes untagged images.
image_retention_days	10	30	Days after which images are eligible for deletion.
create_cloud_storage	11	true	Set false to skip GCS bucket creation.
storage_buckets	11	[{ name_suffix = "data" }]	Additional GCS buckets to provision.
enable_nfs	11	false	Provisions NFS shared storage. Requires gen2.
nfs_mount_path	11	'/mnt/nfs'	Container path where NFS is mounted.
nfs_instance_name	11	""	Name of an existing NFS GCE VM.
nfs_instance_base_name	11	'app-nfs'	Base name for inline NFS VM.
gcs_volumes	11	[]	GCS buckets to mount via GCS Fuse (requires gen2).
manage_storage_kms_iam	11	false	Creates CMEK KMS key and enables CMEK on storage buckets.
enable_artifact_registry_cmek	11	false	Creates Artifact Registry KMS key for at-rest image encryption.
database_type	12	'POSTGRES_15'	Cloud SQL engine. AnythingLLM requires PostgreSQL.
application_database_name	12	'anythingllmdb'	PostgreSQL database name.
application_database_user	12	'anythingllmuser'	Database user.
database_password_length	12	32	Auto-generated password length. Range: 16–64.
enable_auto_password_rotation	12	false	Automated zero-downtime password rotation.
rotation_propagation_delay_sec	12	90	Seconds to wait after rotation before restarting.
db_host_env_var_name	12	""	Additional env var name for DB_HOST.
db_user_env_var_name	12	""	Additional env var name for DB_USER.
db_name_env_var_name	12	""	Additional env var name for DB_NAME.
db_port_env_var_name	12	""	Additional env var name for DB_PORT.
service_url_env_var_name	12	""	Additional env var name for CLOUDRUN_SERVICE_URL.
initialization_jobs	13	[]	One-shot Cloud Run Jobs. Leave empty for default db-init.
cron_jobs	13	[]	Recurring scheduled Cloud Run Jobs.
additional_services	13	[]	Additional Cloud Run services alongside AnythingLLM.
startup_probe	14	{ path="/api/ping", initial_delay_seconds=60, failure_threshold=30, ... }	Startup probe. Long delay for AI model loading.
liveness_probe	14	{ path="/api/ping", initial_delay_seconds=30, failure_threshold=3, ... }	Liveness probe.
startup_probe_config	14	(same as startup_probe)	Alias passed to App CloudRun.
health_check_config	14	(same as liveness_probe)	Alias passed to App CloudRun.
uptime_check_config	14	{ enabled=true, path="/" }	Cloud Monitoring uptime check.
alert_policies	14	[]	Cloud Monitoring metric alert policies.
enable_redis	21	false	Enables Redis. Not required for AnythingLLM core functionality.
redis_host	21	null	Redis hostname or IP. Required when enable_redis = true.
redis_port	21	'6379'	Redis TCP port (string).
redis_auth	21	""	Redis AUTH password. Sensitive.
enable_vpc_sc	22	false	Registers API calls within the project's VPC-SC perimeter.
vpc_cidr_ranges	22	[]	VPC subnet CIDR ranges for VPC-SC network access level.
vpc_sc_dry_run	22	true	Logs VPC-SC violations without blocking.
organization_id	22	""	GCP Organization ID for VPC-SC.
enable_audit_logging	22	false	Enables detailed Cloud Audit Logs.

---

11. Outputs

Output	Description
service_name	Name of the Cloud Run service.
service_url	Public URL of the Cloud Run service.
service_location	GCP region where the Cloud Run service is deployed.
project_id	GCP project ID.
deployment_id	Deployment ID suffix used in resource names.
database_instance_name	Name of the Cloud SQL PostgreSQL instance.
database_name	Name of the application database.
database_user	Name of the application database user.
database_password_secret	Secret Manager secret name for the database password.
database_host	Database host IP address (sensitive).
database_port	Database port.
storage_buckets	Created GCS storage buckets.
nfs_server_ip	NFS server internal IP (sensitive).
nfs_mount_path	NFS mount path inside containers.
container_image	Container image used for the deployment.
container_registry	Artifact Registry repository name.
deployment_summary	Summary of the deployment.
initialization_jobs	Created initialization job names.
cicd_enabled	Whether the CI/CD pipeline is enabled.
github_repository_url	GitHub repository URL connected for CI/CD.
artifact_registry_repository	Artifact Registry repository for container images.
cloudbuild_trigger_name	Cloud Build trigger name for CI/CD.
monitoring_enabled	Whether monitoring is configured.

Configuration Pitfalls & Sensible Defaults

Risk levels: Critical (data loss, full outage, security breach) — High (service unavailable or significant degradation) — Medium (degraded function or increased cost) — Low (minor impact).

Variable	Sensible Default	Risk	Consequence of Incorrect Value
JWT_SECRET (auto-generated)	Random secret stored in Secret Manager	Critical	AnythingLLM uses JWT_SECRET to sign all user authentication tokens. Rotating or changing it immediately invalidates all active sessions and logs out every user. Treat as immutable after first login.
AUTH_TOKEN (optional)	"" (no token required)	High	AUTH_TOKEN acts as a global API bearer token for AnythingLLM's REST API. If left empty, the API is unauthenticated (only protected by network-level controls). Set a strong value and distribute it only to authorised callers.
STORAGE_DIR / document storage	GCS Fuse mount	Critical	AnythingLLM stores all workspace documents, vector embeddings, and LLM configuration under STORAGE_DIR. If this directory is not on a persistent volume (GCS Fuse or NFS), all data is lost on every container restart. Never point STORAGE_DIR to the container's local filesystem without a persistent backend.
enable_nfs	false	High	Without NFS or GCS Fuse, vector indices and uploaded documents are ephemeral. For multi-instance deployments (max_instance_count > 1) shared storage is required or each pod has an inconsistent view of the workspace data.
nfs_mount_path	"/mnt/nfs"	High	Must match the STORAGE_DIR environment variable passed to AnythingLLM. A mismatch means the app writes to a local path while the NFS mount is unused, and data is lost on restart.
LLM_PROVIDER (via environment_variables)	"native" (built-in)	Critical	Without setting the correct LLM_PROVIDER value and corresponding API key, AnythingLLM falls back to its built-in model or errors on all AI requests. When using OpenAI, set LLM_PROVIDER = "openai" and provide OPENAI_API_KEY via secret_environment_variables.
EMBEDDING_ENGINE (via environment_variables)	"native"	High	The embedding engine must be consistent across all documents in a workspace. Changing EMBEDDING_ENGINE after ingesting documents makes existing vector indices incompatible — re-ingest all documents after any change.
secret_environment_variables (API keys)	{} (none)	Critical	AI provider API keys (OPENAI_API_KEY, ANTHROPIC_API_KEY, etc.) must be provided via secret_environment_variables pointing to existing Secret Manager secrets. Providing them as plain environment_variables exposes them in Cloud Run revision metadata.
memory_limit	4Gi	High	AnythingLLM loads embedding models into memory during document ingestion. Running the native embedding engine with less than 4 Gi causes OOM-kills during ingestion of large documents.
cpu_limit	2000m	Medium	AnythingLLM AI workloads are CPU-intensive. With less than 2 vCPU, embedding and inference operations run slowly and Cloud Run requests may time out.
min_instance_count	1	High	Scale-to-zero causes 30–60 s cold starts. For a knowledge workspace used interactively, set to 1 to keep a warm instance. With scale-to-zero, in-memory vector index state is rebuilt on each cold start from the persistent storage.
timeout_seconds	300	High	Document ingestion for large files (PDFs, long text) can take several minutes. Requests time out with a 504 if timeout_seconds is too low. Increase to 600–3600 for document-heavy workloads.
database_type	"POSTGRES"	Critical	AnythingLLM requires PostgreSQL for workspace metadata, user accounts, and conversation history. Using "NONE" causes the application to fall back to SQLite which is not persistent on Cloud Run.
enable_cloudsql_volume	true	Critical	Disabling the Cloud SQL Auth Proxy while relying on Cloud SQL causes all database connections to fail at startup.
ingress_settings	"all"	High	"all" exposes the AnythingLLM workspace publicly. Only the login form protects it. Enable IAP (enable_iap = true) for production deployments, especially if AUTH_TOKEN is not set.
enable_iap	false	High	Without IAP, access control relies solely on the application login. Enabling IAP provides an additional authentication layer before the application is reached.
gcs_volumes	Auto-provisioned for document storage	High	The GCS volume stores all workspace documents. Without implicit-dirs in the mount options, AnythingLLM cannot list files stored in subdirectories and document browsing fails.
execution_environment	"gen2"	High	NFS mounts require gen2. If enable_nfs = true with gen1, the NFS volume silently fails to mount and all document writes go to the ephemeral container filesystem.
backup_schedule	"" (disabled)	High	Without automated backups, the PostgreSQL metadata database (workspaces, users, settings) is unprotected. Enable and set backup_retention_days for production.
enable_redis	false	Low	Redis is optional for AnythingLLM core functionality. If enabled, redis_host must be set or the container will fail to start with a connection error on Redis initialisation.
VECTOR_DB (via environment_variables)	"lancedb" (built-in)	High	Changing the vector database engine after workspaces have been populated requires re-ingesting all documents. Existing vectors are stored in the previous engine's format and are not automatically migrated.
application_version	"latest"	Medium	AnythingLLM releases may change the database schema. An unplanned upgrade can run migrations that are incompatible with the current schema, crashing startup. Pin to a specific release in production.

Destroying Resources

Known Deletion Issue: Serverless IPv4 Address Release

When destroying a Cloud Run deployment, you may encounter an error similar to:

Error: Error waiting for Subnetwork to be deleted: The following serverless IPv4 address(es) on subnet ... are still in use.

Cause: GCP holds serverless IPv4 addresses on the VPC subnet asynchronously after a Cloud Run service is deleted. These addresses are released by GCP approximately 20–30 minutes after the Cloud Run service is removed.

Resolution: Wait 20–30 minutes after the initial destroy attempt, then re-run:

tofu destroy

The second run will succeed once GCP has released the reserved addresses.