Skip to content

jeremylongshore/tons-of-skills-marketplace / plugins/devops/jeremy-vertex-terraform/skills/vertex-infra-expert

vertex-infra-expert

Verified

Execute use when provisioning Vertex AI infrastructure with Terraform. Trigger with phrases like "vertex ai terraform", "deploy gemini terraform", "model garden infrastructure", "vertex ai endpoints terraform", or "vector search terraform". Provisions Model Garden models, Gemini endpoints, vector search indices, ML pipelines, and production AI services with encryption and auto-scaling.

FileKindSize
SKILL.mdskill md2.8 KB
references/ARD.mdmarkdown4.8 KB
references/PRD.mdmarkdown3.8 KB
references/errors.mdmarkdown862 B
references/examples.mdmarkdown4.6 KB
references/implementation.mdmarkdown4.7 KB
scripts/init-terraform.shexecutablescript939 B

references/ARD.md

View on GitHub
# ARD: Vertex Infra Expert

> Part of [Tons of Skills](https://tonsofskills.com) by [Intent Solutions](https://intentsolutions.io) | [jeremylongshore.com](https://jeremylongshore.com)

## System Context

The Vertex Infra Expert generates Terraform modules for provisioning Vertex AI services: endpoints, deployed models, vector search indices, and ML pipelines with production guardrails.

```
Infrastructure Requirements
       ↓
[Vertex Infra Expert]
  ├── Generates: Terraform modules (.tf files)
  ├── Configures: Endpoints, Vector Search, Pipelines, IAM, CMEK, Monitoring
  └── Validates: terraform init/validate/plan
       ↓
GCP Vertex AI Infrastructure
  ├── Model Endpoints (auto-scaled)
  ├── Deployed Models (traffic-split)
  ├── Vector Search Indices
  ├── ML Pipelines
  ├── CMEK Encryption
  ├── IAM (least-privilege)
  └── Cloud Monitoring
```

## Data Flow

1. **Input**: GCP project ID, region, Vertex AI service requirements (endpoint type, model artifact URI, vector dimensions, scaling config), encryption requirements (KMS key ID), and monitoring preferences from user request
2. **Processing**: Generate Terraform module files for each service layer (endpoints, vector search, pipelines, monitoring). Configure CMEK encryption references, auto-scaling parameters, traffic splitting weights, and IAM bindings. Validate with `terraform init` and `terraform plan`.
3. **Output**: Directory of `.tf` files organized by service, `terraform.tfvars.example` with documented variables, validation output, and a deployment runbook with apply commands and verification steps

## Key Design Decisions

| Decision | Choice | Rationale |
|----------|--------|-----------|
| Resource-per-file organization | Separate .tf files for endpoints, models, vector search, monitoring | Clear ownership; teams can adopt individual resources without the full module |
| CMEK-ready by default | Encryption block included with optional KMS key variable | Prod environments require CMEK; easier to configure upfront than retrofit |
| Auto-scaling with scale-down delay | Default 10-minute scale-down delay | Prevents thrashing during bursty traffic; configurable via variable |
| Traffic splitting support | Percentage-based routing across model versions | Enables canary deployments and A/B testing for model updates |
| Vector search dimensions as variable | No default dimension; must be explicitly set | Prevents silently deploying with wrong dimensions that cause retrieval failures |

## Tool Usage Pattern

| Tool | Purpose |
|------|---------|
| Read | Inspect existing Terraform configs, model artifacts, and embedding configurations |
| Write | Create new `.tf` files for Vertex AI resources (endpoints, vector search, pipelines) |
| Edit | Update existing Terraform to add resources, adjust scaling, or fix validation errors |
| Grep | Search for existing Vertex AI resource definitions, KMS key references, and variable usage |
| Glob | Discover existing Terraform module structure and related configuration files |
| Bash(terraform:*) | Run `terraform init`, `terraform validate`, `terraform plan`, `terraform fmt` |
| Bash(gcloud:*) | Verify Vertex AI API status, list models in registry, check KMS key availability |

## Error Handling Strategy

| Error Class | Detection | Recovery |
|------------|-----------|----------|
| Model not found in registry | `terraform plan` fails with model resource not found | Verify model URI; list available models with `gcloud ai models list --region=REGION` |
| KMS key permission denied | Plan fails with cryptoKeyEncrypterDecrypter error | Grant `roles/cloudkms.cryptoKeyEncrypterDecrypter` to the Vertex AI service agent |
| Quota exceeded for endpoints | Plan or apply fails with quota error | Request quota increase via Cloud Console; or reduce max_replicas to fit within quota |
| Vector search dimension mismatch | Apply succeeds but queries return empty results | Verify embedding model output dimension matches the index configuration |
| Region not supported | Vertex AI resource type not available in specified region | Use `us-central1` for widest Vertex AI service availability |

## Extension Points

- Multi-model endpoints: extend traffic splitting module for A/B testing across model versions
- Batch prediction pipelines: add Terraform for scheduled batch prediction jobs
- Feature Store integration: add Terraform for Vertex AI Feature Store for ML feature management
- Custom monitoring: extend dashboards with model-specific metrics (drift detection, prediction quality)
- Module composition: combine with adk-infra-expert for Agent Engine + Vertex AI unified deployments
- Cost estimation: integrate `infracost` for pre-apply cost projections based on endpoint configuration
- Drift detection: scheduled `terraform plan` to catch manual console changes that deviate from code
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall