bricqs-sdk 0.1.0

BRICQS Python SDK — deploy and manage AI models programmatically

bricqs-sdk

Official Python SDK for BRICQS — deploy AI models, manage GPU deployments, and build AI infrastructure programmatically.

Installation

pip install bricqs-sdk

Quick start

from bricqs_sdk import BricqsClient

client = BricqsClient(api_key="<your-api-key>")

# Deploy a model (blocks until running)
dep = client.deployments.deploy(
    name="my-llama-api",
    model_id="meta-llama/Llama-3-8B-Instruct",
    environment="production",
)

print(dep.endpoint)  # https://<app>.azurecontainerapps.io

Deployments

# List all deployments
deps = client.deployments.list()
deps = client.deployments.list(environment="preview")

# Get a single deployment
dep = client.deployments.get("deployment-id")

# Deploy without waiting (non-blocking)
dep = client.deployments.deploy(
    name="my-api",
    model_id="mistralai/Mistral-7B-Instruct-v0.3",
    wait=False,
)

# Poll manually (useful for progress bars)
for snapshot in client.deployments.poll(dep.id):
    print(f"Stage: {snapshot.stage}")

# Stop, delete, promote
client.deployments.stop("deployment-id")
client.deployments.delete("deployment-id")
prod = client.deployments.promote("preview-deployment-id")

# Logs and metrics
lines = client.deployments.logs("deployment-id", lines=200)
metrics = client.deployments.metrics("deployment-id", hours=6)
print(metrics.cpu_pct)       # list of {time, avg, total, max}
print(metrics.response_time_ms)

# Summary stats
summary = client.deployments.summary()
print(summary.active_deployments, summary.gpu_hours_30d)

Models

models = client.models.list()
for m in models:
    print(m.id, m.task, m.gpu)

Support

ticket = client.support.create_ticket(
    subject="Deployment stuck in provisioning",
    body="My deployment bricqs-abc123 has been in 'creating_container' for 20 minutes.",
    category="deployments",
    priority="high",
)
print(ticket.id)

tickets = client.support.list_tickets()

Context manager

with BricqsClient(api_key="<token>") as client:
    dep = client.deployments.deploy(name="my-api", model_id="meta-llama/Llama-3-8B-Instruct")

Error handling

from bricqs_sdk import BricqsClient, AuthError, NotFoundError, RateLimitError, BricqsError

try:
    dep = client.deployments.get("nonexistent-id")
except NotFoundError:
    print("Deployment not found")
except AuthError:
    print("Invalid API key")
except RateLimitError:
    print("Rate limited — back off and retry")
except BricqsError as e:
    print(f"API error {e.status_code}: {e}")

Types

All methods return typed dataclasses:

Type	Fields
Deployment	id, name, status, model_id, environment, endpoint, min_replicas, max_replicas, region, stage, created_at
Model	id, label, task, gpu, cpu, memory, description
DeploymentSummary	active_deployments, gpu_hours_30d, active_regions, failure_rate, requests_per_sec
Metrics	cpu_pct, memory_pct, requests, response_time_ms, replicas, gpu_pct
Ticket	id, subject, category, status, priority, created_at

Every type also exposes a .raw dict with the full API response.

Links

• Documentation
• API Reference
• CLI (pip install bricqs)