VibOps Python SDK

Typed Python SDK for the VibOps API -- The AI Infrastructure Engine.

Install

pip install vibops

Quick start

from vibops import VibOps

v = VibOps(url="https://vibops.example.com", token="vib_...")

# List GPU clusters
clusters = v.clusters.list()

# Deploy a model
job = v.models.deploy("llama3", "gpu-prod", replicas=2)

# Check FinOps budget
budget = v.finops.budget()

# List proactive insights
insights = v.insights.list(severity="critical")

Async usage

import asyncio
from vibops import AsyncVibOps

async def main():
    async with AsyncVibOps(url="https://vibops.example.com", token="vib_...") as v:
        clusters = await v.clusters.list()
        jobs = await v.jobs.list(status="running")
        waste = await v.finops.waste(sustained_hours=6)

asyncio.run(main())

Resources

Resource	Methods
clusters	list(), deployments(cluster), gpu_metrics(cluster)
jobs	list(), get(id), create(action, payload), cancel(id)
models	deploy(model, cluster), scale(name, namespace, replicas, cluster)
gateways	list(), register(name), delete(id)
finops	budget(), spend_trend(), chargeback(), waste(), live_cost()
agents	usage(), usage_detail(id), budget(id), set_budget(id, limit), list_budgets()
security	scan(), findings()
compliance	check(), results()
insights	list(), acknowledge(id)

Error handling

from vibops.exceptions import AuthenticationError, NotFoundError, RateLimitError

try:
    job = v.jobs.get("nonexistent")
except NotFoundError:
    print("Job not found")
except AuthenticationError:
    print("Check your token")

Auto-retry is built in for 502/503 responses (1 retry).

License

MIT