terradev-cli 2.1.0

BYOAPI Parallel provisioning and orchestration for cross-cloud cost optimization with real-time analytics

Terradev

BYOAPI Parallel provisioning and orchestration for cross-cloud cost optimization.

Developers overpay for compute by only accessing single-cloud workflows or using slow sequential provisioning with inefficient egress and rate-limiting.

Terradev is a cross-cloud provisioning CLI that saves developers 60% on end-to-end compute provisioning costs. It works faster than any sequential tool to compress + stage datasets, parallel-provision optimal instances + nodes, and deploy 5x faster than baseline provisions, at an average of 3 minutes, with egress optimization.

Terradev is integrated with Kubernetes and Karpenter: We generate Helm templates.

🚀 HuggingFace Spaces One-Click Deployment

Deploy any HuggingFace model to Spaces in 2-5 minutes with one command:

# Install HF Spaces support
pip install terradev-cli[hf]

# Set your HF token
export HF_TOKEN=your_huggingface_token

# Deploy Llama 2 with one click
terradev hf-space my-llama --model-id meta-llama/Llama-2-7b-hf --template llm

# Deploy custom model with GPU
terradev hf-space my-model --model-id microsoft/DialoGPT-medium \
  --hardware a10g-large --sdk gradio

# Result:
# 🌐 Space URL: https://huggingface.co/spaces/username/my-llama
# 📊 100k+ researchers can now access your model!
# ⏱️  Ready in 2-5 minutes

🎯 HF Spaces Features

• 🌐 One-Click Deployment: No manual configuration required
• 📋 Template-Based: LLM, embedding, and image model templates
• 🔧 Multi-Hardware: CPU-basic to A100-large GPU tiers
• ⚡ Auto-Generated Apps: Gradio, Streamlit, and Docker support
• 💰 Revenue Streams: Hardware upgrades, private spaces, template licensing

📊 Available Templates

# LLM Template (A10G GPU)
terradev hf-space my-llama --model-id meta-llama/Llama-2-7b-hf --template llm

# Embedding Template (CPU-upgrade)
terradev hf-space my-embeddings --model-id sentence-transformers/all-MiniLM-L6-v2 --template embedding

# Image Model Template (T4 GPU)
terradev hf-space my-image --model-id runwayml/stable-diffusion-v1-5 --template image

Installation

pip install terradev-cli

With HF Spaces support:

pip install terradev-cli[hf]        # HuggingFace Spaces deployment
pip install terradev-cli[all]        # All cloud providers + ML services + HF Spaces

Quick Start

# 1. Get setup instructions for any provider
terradev setup runpod --quick
terradev setup aws --quick

# 2. Configure your cloud credentials (BYOAPI — you own your keys)
terradev configure --provider runpod
terradev configure --provider aws
terradev configure --provider vastai

# 3. Deploy to HuggingFace Spaces (NEW!)
terradev hf-space my-llama --model-id meta-llama/Llama-2-7b-hf --template llm
terradev hf-space my-embeddings --model-id sentence-transformers/all-MiniLM-L6-v2 --template embedding
terradev hf-space my-image --model-id runwayml/stable-diffusion-v1-5 --template image

# 4. Get enhanced quotes with conversion prompts
terradev quote -g A100
terradev quote -g A100 --quick  # Quick provision best quote

# 5. Provision the cheapest instance (real API call)
terradev provision -g A100

# 6. Configure ML services
terradev configure --provider wandb --dashboard-enabled true
terradev configure --provider langchain --tracing-enabled true

# 7. Use ML services
terradev ml wandb --test
terradev ml langchain --create-workflow my-workflow

# 8. View analytics
python user_analytics.py

# 9. Provision 4x H100s in parallel across multiple clouds
terradev provision -g H100 -n 4 --parallel 6

# 10. Dry-run to see the allocation plan without launching
terradev provision -g A100 -n 2 --dry-run

# 11. Manage running instances
terradev status --live
terradev manage -i <instance-id> -a stop
terradev manage -i <instance-id> -a start
terradev manage -i <instance-id> -a terminate

# 12. Execute commands on provisioned instances
terradev execute -i <instance-id> -c "python train.py"

# 13. Stage datasets near compute (compress + chunk + upload)
terradev stage -d ./my-dataset --target-regions us-east-1,eu-west-1

# 14. View cost analytics from the tracking database
terradev analytics --days 30

# 15. Find cheaper alternatives for running instances
terradev optimize

# 16. One-command Docker workload (provision + deploy + run)
terradev run --gpu A100 --image pytorch/pytorch:latest -c "python train.py"

# 17. Keep an inference server alive
terradev run --gpu H100 --image vllm/vllm-openai:latest --keep-alive --port 8000

BYOAuth — Bring Your Own Authentication

Terradev never touches, stores, or proxies your cloud credentials through a third party. Your API keys stay on your machine in ~/.terradev/credentials.json — encrypted at rest, never transmitted.

How it works:

1. You run terradev configure --provider <name> and enter your API key
2. Credentials are stored locally in your home directory — never sent to Terradev servers
3. Every API call goes directly from your machine to the cloud provider
4. No middleman account, no shared credentials, no markup on provider pricing

Why this matters:

• Zero trust exposure — No third party holds your AWS/GCP/Azure keys
• No vendor lock-in — If you stop using Terradev, your cloud accounts are untouched
• Enterprise-ready — Compliant with SOC2, HIPAA, and internal security policies that prohibit sharing credentials with SaaS vendors
• Full audit trail — Every provision is logged locally with provider, cost, and timestamp

CLI Commands

Command	Description
terradev configure	Set up API credentials for any provider
terradev quote	Get real-time GPU pricing across all clouds
terradev provision	Provision instances with parallel multi-cloud arbitrage
terradev manage	Stop, start, terminate, or check instance status
terradev status	View all instances and cost summary
terradev execute	Run commands on provisioned instances
terradev stage	Compress, chunk, and stage datasets near compute
terradev analytics	Cost analytics with daily spend trends
terradev optimize	Find cheaper alternatives for running instances
terradev run	Provision + deploy Docker container + execute in one command
terradev hf-space	NEW: One-click HuggingFace Spaces deployment
terradev up	NEW: Manifest cache + drift detection
terradev rollback	NEW: Versioned rollback to any deployment
terradev manifests	NEW: List cached deployment manifests
terradev integrations	Show status of W&B, Prometheus, and infra hooks

🚀 HF Spaces Commands (NEW!)

# Deploy Llama 2 to HF Spaces
terradev hf-space my-llama --model-id meta-llama/Llama-2-7b-hf --template llm

# Deploy with custom hardware
terradev hf-space my-model --model-id microsoft/DialoGPT-medium \
  --hardware a10g-large --sdk gradio --private

# Deploy embedding model
terradev hf-space my-embeddings --model-id sentence-transformers/all-MiniLM-L6-v2 \
  --template embedding --env BATCH_SIZE=64

📊 Manifest Cache Commands (NEW!)

# Provision with manifest cache
terradev up --job my-training --gpu-type A100 --gpu-count 4

# Fix drift automatically
terradev up --job my-training --fix-drift

# Rollback to previous version
terradev rollback my-training@v2

# List all cached manifests
terradev manifests --job my-training

Observability & ML Integrations

Terradev facilitates connections to your existing tools via BYOAPI — your keys stay local, all data flows directly from your instances to your services.

Integration	What Terradev Does	Setup
Weights & Biases	Auto-injects WANDB_* env vars into provisioned containers	terradev configure --provider wandb --api-key YOUR_KEY
Prometheus	Pushes provision/terminate metrics to your Pushgateway	terradev configure --provider prometheus --api-key PUSHGATEWAY_URL
Grafana	Exports a ready-to-import dashboard JSON	terradev integrations --export-grafana

Infrastructure Hooks

Terradev integrates with end-user managed infrastructure when available:

• Kubernetes — Deploy to existing K8s clusters
• Karpenter — Multi-cloud node provisioning via Karpenter CRDs
• Grafana — Dashboard creation for training job monitoring
• Open Policy Agent — Policy-as-code compliance enforcement

Cost Savings Examples

GPU Type	On-Demand (AWS)	Best Spot Price	Terradev Best	Savings
A100	$4.06/hr	$1.22/hr	$0.84/hr	79%
H100	$7.20/hr	$2.16/hr	$1.50/hr	79%
RTX 4090	$1.21/hr	$0.36/hr	$0.25/hr	79%

Prices queried in real-time from all 10+ providers. Actual savings vary by availability.

Project Structure

terradev_cli/
├── cli.py                   # Main CLI (quote, provision, manage, status, etc.)
├── core/
│   ├── cost_tracker.py      # SQLite cost tracking database
│   ├── dataset_stager.py    # Compression, chunking, parallel staging
│   ├── egress_optimizer.py  # Cross-cloud data transfer cost optimization
│   ├── parallel_provisioner.py  # Concurrent multi-cloud deployment engine
│   ├── tier_manager.py      # Research / Research+ / Enterprise tier logic
│   └── rate_limiter.py      # API rate limiting
├── providers/
│   ├── base_provider.py     # Abstract provider interface
│   ├── aws_provider.py      # AWS EC2 (boto3)
│   ├── gcp_provider.py      # Google Cloud Compute
│   ├── azure_provider.py    # Azure Compute
│   ├── runpod_provider.py   # RunPod GraphQL API
│   ├── vastai_provider.py   # Vast.ai REST API
│   ├── lambda_labs_provider.py  # Lambda Cloud API
│   ├── coreweave_provider.py    # CoreWeave API
│   ├── tensordock_provider.py   # TensorDock API
│   ├── crusoe_provider.py       # Crusoe Cloud API
│   └── provider_factory.py  # Dynamic provider instantiation
├── integrations/
│   ├── wandb_integration.py     # W&B BYOAPI facilitation
│   └── prometheus_integration.py # Prometheus Pushgateway metrics
└── utils/
    └── formatters.py        # Output formatting

Pricing Tiers

	Research (Free)	Research+ ($49.99/mo)	Enterprise ($299.99/mo)
Provisions/month	10	100	Unlimited
Max concurrent instances	1	4	32
Providers	All 11	All 11	All 11 + priority
Cost tracking	Yes	Yes	Yes
Dataset staging	Yes	Yes	Yes
Egress optimization	Basic	Full	Full + custom routes

Integrations

Jupyter / Colab / VS Code Notebooks

pip install terradev-jupyter

%load_ext terradev_jupyter

%terradev quote -g A100
%terradev provision -g H100 --dry-run
%terradev run --gpu A100 --image pytorch/pytorch:latest --dry-run

GitHub Actions

- uses: theoddden/terradev-action@v1
  with:
    gpu-type: A100
    max-price: "1.50"
  env:
    TERRADEV_RUNPOD_KEY: ${{ secrets.RUNPOD_API_KEY }}

Docker (One-Command Workloads)

terradev run --gpu A100 --image pytorch/pytorch:latest -c "python train.py"
terradev run --gpu H100 --image vllm/vllm-openai:latest --keep-alive --port 8000

Requirements

• Python >= 3.9
• Cloud provider API keys (configured via terradev configure)

License

MIT License - see LICENSE file for details