Quickstart Guide

Connect your first model endpoint to MLSrvyn in under 30 minutes. This guide walks through installation, configuration, and verifying that traffic profiling is running.

Prerequisites

• An active MLSrvyn account (request access at mlsrvyn.com/login/signup.html)
• An inference endpoint serving at a stable URL (HTTP/gRPC)
• Python 3.9+ or access to your Kubernetes cluster
• Your API key from the MLSrvyn dashboard

Step 1: Install the SDK

$ pip install mlsrvyn
Successfully installed mlsrvyn-1.4.2
$ mlsrvyn --version
mlsrvyn 1.4.2

Step 2: Authenticate

$ mlsrvyn auth login --api-key sk-srv-xxxxxxxx
Authenticated as workspace: your-org
Region: us-west-2

Step 3: Create your first config

Create an mlsrvyn.yaml file in your project directory. Start with a minimal config — MLSrvyn will auto-detect the archetype after 24 hours of traffic observation.

workspace: your-org
models:
  - name: my-first-model
    endpoint: http://inference-svc.prod:8080
    archetype: auto     # profiler will detect
    slo:
      p99_ms: 100
    scaling:
      min_replicas: 1
      max_replicas: 10

Step 4: Deploy the agent

$ mlsrvyn apply -f mlsrvyn.yaml
Connecting to endpoint http://inference-svc.prod:8080... OK
Agent registered: my-first-model (id: mdl-8f3a9c)
Traffic profiling started. Archetype will be detected in ~24h.
Dashboard: https://mlsrvyn.com/login/index.html → Fleet → my-first-model

Step 5: Verify in the dashboard

Log into the MLSrvyn dashboard and navigate to Fleet → your model name. You should see the telemetry stream populating within 2 minutes. After 24 hours, the archetype field will auto-populate with the detected traffic shape.

Next steps

• Full configuration reference — all yaml fields documented
• API reference — programmatic fleet management
• Use case guides — archetype-specific configuration examples