Build

SDKs and frameworks

Working setups for the OpenAI and Anthropic SDKs, LangChain, and the Vercel AI SDK.

OpenAI SDKs

Point the client at https://api.quancis.space/v1 and use your Quancis key. Both the Chat Completions and Responses APIs work.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.quancis.space/v1",
    api_key="sk-quan-...",
)

reply = client.chat.completions.create(
    model="kael-beta",
    messages=[{"role": "user", "content": "Review this function."}],
)
print(reply.choices[0].message.content)

Anthropic SDKs

Use https://api.quancis.space without /v1. Anthropic SDKs add /v1/messages themselves, so a base URL that already ends in /v1 would call /v1/v1/messages and fail.

import anthropic

client = anthropic.Anthropic(
    base_url="https://api.quancis.space",  # no /v1
    api_key="sk-quan-...",
)

message = client.messages.create(
    model="kael-beta",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Review this function."}],
)
print(message.content[0].text)

Frameworks

Any framework that lets you set an OpenAI-compatible base URL will work. These two are tested patterns:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="kael-beta",
    base_url="https://api.quancis.space/v1",
    api_key="sk-quan-...",
)

print(llm.invoke("Review this function.").content)

Environment variables

Tools that read the standard variables need no code change:

Client settings that matter

• Set request timeouts above 8 minutes, or stream. Hard requests can run for minutes.
• The SDKs retry some failures on their own. That is fine for 429 and 5xx; see Errors for what is safe to retry.