The control plane
for LLM traffic.
Olyx gives beta teams one governed path for model calls: create a trace, execute through the SDK, complete the workflow, then inspect cost, latency, and routing behavior.
TypeScript · Python · Ruby · OpenAI-compatible migration path · optional private agent routes.
Route model traffic through project policy instead of duplicating provider logic across call sites.
Record traces for model calls, tool steps, latency, cost, and routing decisions that developers can inspect.
Use completed traces and replays to compare models before changing production routing.
From zero to first traced request.
Start with a project-scoped API key and an SDK call. If you are migrating an existing app, the OpenAI-compatible gateway is the quickest swap; the SDK pages cover the full trace-first workflow.
- 01
STEP / 01
Create a project
Projects isolate model registry settings, API keys, trace history, and beta configuration.
Open dashboard - 02
STEP / 02
Generate an API key
Use one project-scoped key per service or environment so staging, production, and workers are easy to separate.
API key docs - 03
STEP / 03
Send an SDK request
Create a trace, call `execute`, and complete the trace when the workflow is done.
See SDK example - 04
STEP / 04
Inspect and tune
Use traces, cost intelligence, and replays to understand model behavior before changing routing.
Trace docs
Keep the same client. Change only the base URL.
This is the fastest migration path for an existing app. You keep your current OpenAI-style client and point it at Olyx. The SDK pages below cover the governed trace-first path.
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.OLYX_API_KEY,
baseURL: "https://olyx.ai/v1",
});
const response = await client.chat.completions.create({
model: "gpt-4o",
messages: [{ role: "user", content: "Translate to French: Hello, world." }],
}); import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["OLYX_API_KEY"],
base_url="https://olyx.ai/v1",
)
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": "Translate to French: Hello, world."}],
) require "openai"
client = OpenAI::Client.new(
access_token: ENV["OLYX_API_KEY"],
uri_base: "https://olyx.ai/v1",
)
response = client.chat(
parameters: {
model: "gpt-4o",
messages: [{ role: "user", content: "Translate to French: Hello, world." }],
}
) These examples show the same shape across the supported SDKs. Use the language your service already runs.
import Olyx from "@olyx-labs/olyx";
const client = new Olyx({ apiKey: process.env.OLYX_API_KEY! });
const trace = await client.traces.create({
metadata: { feature: "quickstart" },
});
const result = await client.execute({
traceId: trace.data.id,
input: "Translate to French: Hello world.",
});
await client.traces.complete(trace.data.id);
console.log(result.data.output); import os
import olyx
client = olyx.Olyx(
api_key=os.environ["OLYX_API_KEY"],
mock=False,
)
trace = client.traces.create(metadata={"feature": "quickstart"})
result = client.execute(
trace_id=trace.id,
input="Translate to French: Hello world.",
)
client.traces.complete(trace.id)
print(result.output) client = Olyx.new(api_key: ENV.fetch("OLYX_API_KEY"))
trace = client.traces.create(
metadata: { feature: "quickstart" }
)
result = client.execute(
trace_id: trace.id,
input: "Translate to French: Hello world."
)
client.traces.complete(trace.id)
puts result.output
The working map.
ARCHITECTURE
How SDK calls move through traces, routing, providers, and optional private agent routes.
MODEL REGISTRY
Register public or private models and assign them to routing tiers.
MCP
Connect scoped MCP tools while keeping execution in your application.
TRACES
Inspect execution steps, summaries, routing decisions, and tool activity.
COST INTELLIGENCE
Read cost summaries, margin fields, infrastructure breakdowns, and optimization grades.
REPLAYS
Compare a recorded trace against another model before changing live routing.
ALERTS
Send beta webhook alerts for selected trace and spend-cap events.
PERFORMANCE
Measure trace latency, MCP overhead, and closed-beta load-test behavior.
TYPESCRIPT SDK
TypeScript client for traces, execution, replays, and OpenAI-compatible migration.
PYTHON SDK
Python client for trace-bound execution and backend service workflows.
RUBY SDK
Ruby client for Rails services, trace orchestration, and gateway calls.
API REFERENCE
Low-level endpoint reference for SDK authors and control-plane automation.
Backend services use API keys. Dashboard users use sessions. Keep model-calling traffic on project-scoped API keys and keep those keys out of browser code.
OLYX_API_KEY=ak_<key_id>.<secret>