Getting Started
Quickstart
Deploy a live LLaMA 3 API endpoint in under 60 seconds.
1
Install the CLI
pip install bricqsRequires Python 3.9+. The CLI talks directly to the BRICQS API over HTTPS.
2
Log in to your account
bricqs loginYour credentials are stored in ~/.bricqs/config.json with 600 permissions. Don't have an account? Sign up free →
3
Deploy LLaMA 3
bricqs deploy my-llama-api --model meta-llama/Llama-3-8B-InstructThe CLI polls in real-time and prints the live endpoint once provisioning completes (~2–5 min for first GPU pull).
textDeploying my-llama-api
Model: meta-llama/Llama-3-8B-Instruct
Environment: production
Replicas: 0–1
████░░ waiting_for_model
✓ Running!
╭─ Deployment Live ───────────────────────────────────────────────╮
│ Endpoint: https://bricqs-my-llama-api.bricqs.run │
│ │
│ Test it: │
│ curl .../v1/chat/completions -d '{"model":"llama3:8b",...}' │
│ │
│ View logs: bricqs logs <id> │
│ Stop: bricqs stop <id> │
╰──────────────────────────────────────────────────────────────────╯4
Call your model
Your endpoint is fully OpenAI-compatible. Use curl or the OpenAI SDK:
bashcurl https://<your-endpoint>/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "llama3:8b",
"messages": [{"role": "user", "content": "Hello!"}]
}'Or with the OpenAI Python SDK:
pythonfrom openai import OpenAI
client = OpenAI(
base_url="https://<your-endpoint>/v1",
api_key="not-needed",
)
response = client.chat.completions.create(
model="llama3:8b",
messages=[{"role": "user", "content": "What is quantum computing?"}],
)
print(response.choices[0].message.content)5
Stop when done
bashbricqs stop <deployment-id>
# Or permanently delete:
bricqs delete <deployment-id>Stopped deployments scale to zero — you're only billed while running.
What just happened?
- BRICQS provisioned a GPU runtime container with a T4 GPU in
westus2 - The Ollama runtime pulled LLaMA 3 8B weights (~4.7 GB) into GPU memory
- An OpenAI-compatible HTTPS endpoint went live with auto-scaling 0→2 replicas
- Real-time metrics (CPU, memory, GPU %, requests, latency) are collected automatically