EN
Open developer docs

OCTORYN ROUTER · SYDNEY PRODUCTION

One API. Every model. Your rules.

A governed AI routing layer for Octopus customers and the Octoryn product family. Keep the OpenAI SDK you already use, then choose providers by policy—not lock-in.

Read the quickstartExplore capabilities
A familiar requestSydney production
from openai import OpenAI

client = OpenAI({
  base_url="https://api.octoryn.dev/v1",
  api_key="oct_••••••••"
)

response = client.chat.completions.create({
  model="policy/frontier",
  messages=[...]
)

Change the base URL, not your application architecture.

01CompatibleOpenAI client patterns
02PortableProviders stay replaceable
03GovernedPolicy before dispatch
04ObservableOne audit trail

MODEL DIRECTORY

A model catalogue designed around decisions.

Compare capabilities, context, price class and deployment policy. Publish stable aliases so applications can evolve without shipping a new build.

WorkloadStable aliasOptimise
01Frontier reasoningpolicy/frontierquality

Prioritises the strongest permitted reasoning route, with health-aware fallback.

complex analysis · planning · fallback
02Balanced reasoningpolicy/reasoningvalue

Balances reasoning quality, response time and cost for everyday production work.

analysis · production · cost control
03Private reasoningpolicy/reasoning-privategoverned

Keeps sensitive reasoning traffic within approved providers and regions.

residency · allowlists · audit
WorkloadStable aliasOptimise
01Reliable tool usepolicy/toolsreliability

Routes tool-calling workloads to models selected for consistent schema adherence.

tools · JSON schema · retries
02Fast functionspolicy/functionslatency

Optimises short function-calling loops where time to first action matters.

functions · low latency · streaming
03Private toolspolicy/tools-privategoverned

Applies regional and provider controls before a tool-enabled request is dispatched.

policy gates · residency · audit
WorkloadStable aliasOptimise
01Visual reasoningpolicy/visionquality

Selects a high-capability multimodal route for detailed image understanding.

images · documents · analysis
02Fast visionpolicy/vision-fastlatency

Targets responsive visual classification and extraction workloads.

classification · extraction · speed
03Private visionpolicy/vision-privategoverned

Constrains image traffic to approved data-handling and regional policies.

residency · privacy · evidence
WorkloadStable aliasOptimise
01Fast generalpolicy/fastlatency

Chooses the lowest-latency healthy route that satisfies the workload policy.

chat · streaming · health routing
02Realtime responsepolicy/realtimespeed

Optimises interactive applications for a responsive first token and steady stream.

realtime · first token · streaming
03Sydney regionalpolicy/sydneyproximity

Prefers approved capacity close to Australian production workloads.

Sydney · proximity · failover
WorkloadStable aliasOptimise
01Private workloadpolicy/residencygoverned

Routes only through providers and regions allowed by the workload policy.

residency · provider policy · audit
02Australia onlypolicy/auresidency

Restricts eligible routes to approved Australian processing locations.

Australia · regional controls · evidence
03BYOK routepolicy/byokownership

Uses customer-managed provider credentials while preserving central routing controls.

BYOK · portability · spend control

ROUTING

Turn provider choice into policy.

Route for quality, latency, cost or residency. Define permitted providers, set fallback order and keep the same application contract as infrastructure changes.

01APPLICATIONOpenAI SDK
02OCTORYN ROUTERpolicy · budget · health
PROVIDER APROVIDER BBYOK / PRIVATE
01

Fallbacks

Move around provider incidents without rewriting clients.

02

Policy gates

Apply approved models, regions and data-handling rules before dispatch.

03

Provider portability

Bring provider credentials and move workloads without re-platforming.

OPERATING CONTROL

Guardrails your finance and security teams can verify.

Separate teams and environments. Assign keys and budgets. Trace each request from application to provider with the policy decision attached.

01

Budget

Set hard limits by workspace, key or environment.

02

Audit

Export request, policy, usage and cost evidence.

03

Observe

See latency, failures, token usage and provider outcomes.

04

Govern

Keep model access and routing rules centrally managed.

DEVELOPERS

Two lines between your code and a portable model layer.

Use the OpenAI SDK, point it to Octoryn, and select a model or governed alias. Streaming, tool calls and structured outputs stay in the workflow developers already know.

Open developer docs
Choose code language
Python · OpenAI SDKTypeScript · OpenAI SDK
from openai import OpenAI

client = OpenAI({
  base_url="https://api.octoryn.dev/v1",
  api_key=os.environ["OCTORYN_API_KEY"]
)

stream = client.chat.completions.create({
  model="policy/frontier",
  messages=[{"role": "user", "content": prompt}],
  stream=True
)
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.octoryn.dev/v1",
  apiKey: process.env.OCTORYN_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "policy/frontier",
  messages: [{ role: "user", content: prompt }],
  stream: true,
});

BUILT FOR THE OCTOPUS FAMILY

One routing layer across the Octoryn runtime.

Octoryn Router connects model supply to the wider Octopus system: knowledge, agents, voice, evaluation and managed operations. Customers get one governed path to models without coupling every product to every provider.

01Octoryn Runtime

The governed execution layer for AI applications, workflows and model calls.

API · orchestration · observability
02Octoryn Knowledge

Connect trusted organisational knowledge to models with retrieval controls and evidence.

retrieval · permissions · citations
03Octoryn Agents

Run tool-using agents with policy boundaries, durable execution and operational visibility.

tools · workflows · guardrails
04Octoryn Voice

Build low-latency voice experiences on the same governed model and identity layer.

realtime · speech · telephony
05Octoryn Evaluation

Measure model, prompt and routing quality before changes reach production traffic.

datasets · scoring · regression
06Octoryn Cloud

Managed deployment and operations for the complete Octoryn production stack.

hosting · security · operations

Build once. Route with intent.

Start with the interface your team already knows. Add policy, portability and production control when you need them.

Start with the docsMeet Octopus OS