Integrations.
deeprelay speaks the OpenAI chat and embeddings protocol, so the SDKs and gateways you already run reach it with two settings: the base URL and your key. No deeprelay package, no adapter code. These guides show the exact settings for each tool, then the parts people get wrong.
| Base URL | https://api.deeprelay.ai/v1 |
| Auth | Authorization: Bearer deeprelay_live_… |
| Model id | deeprelay/<model> as listed by GET /v1/models |
§ 07.0Pick your tool¶
Vercel AI SDK
generateText, streamText, a Next.js route with useChat, tool calling, structured output, and embeddings through the official openai-compatible provider.
@ai-sdk/openai-compatibleRead the guide →LiteLLM
The litellm Python SDK for chat, streaming, tools, JSON output and embeddings, plus the proxy server with model aliases, fallbacks and a master key.
openai/deeprelay/<model>Read the guide →Portkey
Portkey's AI gateway in front of deeprelay as a custom host: saved providers, retries, fallback configs, caching and request logs.
custom_hostRead the guide →§ 07.4Anything else that speaks OpenAI¶
A tool that is not listed here almost certainly has an “OpenAI-compatible” or “custom base URL” option. Set it to the base URL above, paste your key, and use a deeprelay model id. The official OpenAI SDKs are the reference case:
from openai import OpenAI
client = OpenAI(
base_url="https://api.deeprelay.ai/v1",
api_key="deeprelay_live_...",
)
out = client.chat.completions.create(
model="deeprelay/deepseek-v4.1-flash",
messages=[{"role": "user", "content": "hi"}],
)Coding agents and editors (Claude Code, Codex CLI, Cursor, Cline, and the rest) have their own pages under Guides. The request and response shapes every integration is sending are in the Serverless Inference guide.