Use case

Migrate OpenAI SDK apps to a multi-model gateway

Many AI products start with a single OpenAI SDK integration. The next problem is model choice: teams want Claude for some tasks, DeepSeek for cost control, Gemini for experiments and Chinese model families for regional use cases. PEKPIK LLM keeps the OpenAI-style client while expanding the model menu.

Request access Read quickstart View model catalog

Primary query

OpenAI SDK migration

Why teams search for this

Small integration diff for common chat completion workloads.

Model choice moves from code rewrites to configuration and evaluation.

Teams can run side-by-side tests before moving production traffic.

The same pattern supports future provider changes.

Where PEKPIK fits

Good fit

OKYou have OpenAI SDK code already in production.
OKYour team needs more model families but wants to avoid a full abstraction rewrite.
OKYou want staged rollout, fallback and cost comparison.

Check first

!Do not migrate all traffic at once; start with staging and representative prompts.
!Validate tool calls, JSON mode and streaming behavior if your app depends on them.
!Keep your old provider path available until evaluation passes.

OpenAI-compatible example

base_url swap

from openai import OpenAI

client = OpenAI(
    base_url="https://aiapiv2.pekpik.com/v1",
    api_key="sk-...",
)

response = client.chat.completions.create(
    model="claude-opus-4-7",
    messages=[{"role": "user", "content": "Summarize this for a product team."}],
)

Suggested rollout

01

Extract base URL, API key and model ID into environment configuration.
02

Point staging traffic at PEKPIK and keep the same SDK methods.
03

Run regression prompts for quality, formatting and latency.
04

Roll out by workload segment, then retire provider-specific paths only when stable.

FAQ

What is the smallest migration change?

For standard OpenAI SDK calls, the smallest change is usually the base URL, API key and model ID.

Should I build my own router instead?

Self-hosted routing can work for teams with infrastructure bandwidth. A managed gateway is simpler when procurement, billing and upstream operations are the bottleneck.