BUILD & OPERATE
Docs / Getting started
Quickstart: connect a first workload
A provider-neutral checklist for taking one model-powered task from local experiment to an observable integration.
↗
quickstart
Docs / Configuration concepts
Routing policies
Model route selection as a versioned policy with explicit eligibility, preference, and fallback behavior.
↗
routing-policies
Docs / Operations
Observability and request traces
Capture enough metadata to troubleshoot model requests while keeping prompt and response content out of routine telemetry.
↗
observability
Docs / Architecture
Deployment patterns
Choose an integration shape by considering trust boundaries, network paths, operational ownership, and failure handling.
↗
deployment-options
Docs / Trust and data
Data handling for model requests
A practical review framework for deciding what data may enter a model workflow, where it travels, and what gets retained.
↗
data-handling
DOCUMENTATION
Build with Strateline.
Start with the request path. Explore routing, traces and deployment boundaries, then adapt the examples to your platform.
Start with one workload.
Choose a model-powered feature, define its constraints and follow one request from input to result. Keep credentials on your server and separate development from production.
Understand the request boundary.
A familiar request surface keeps application code independent of routing decisions. This provider-neutral example shows the structure to adapt to your own endpoint and credentials.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.your-domain.com/v1",
api_key=os.environ["API_KEY"]
)
response = client.chat.completions.create(
model="balanced",
messages=[{
"role": "user",
"content": "What should we build next?"
}]
)Make the system observable.
Use the guides to choose a routing policy, inspect a request trace and define your deployment boundary. Collect only the operational metadata your team needs.