LiteLLM¶

The LiteLLM SDK takes an api_base per call. Point it at Privyx:

OpenAI models¶

privyx proxy --upstream https://api.openai.com

import litellm

reply = litellm.completion(
    model="openai/gpt-4o-mini",
    api_base="http://localhost:8000/v1",
    messages=[{"role": "user", "content": "Write a short greeting to alice@example.com"}],
)
print(reply.choices[0].message.content)

Anthropic models¶

privyx proxy --upstream https://api.anthropic.com

import litellm

reply = litellm.completion(
    model="anthropic/claude-opus-5-5",
    api_base="http://localhost:8000",
    messages=[{"role": "user", "content": "Write a short greeting to alice@example.com"}],
)
print(reply.choices[0].message.content)

LiteLLM speaks the Anthropic format to an anthropic/ model, so this request arrives at Privyx on /v1/messages.

Without changing code¶

export OPENAI_BASE_URL=http://localhost:8000/v1
export ANTHROPIC_BASE_URL=http://localhost:8000

Good to know¶

One Privyx per upstream. LiteLLM can talk to many providers; Privyx forwards to one. Run a Privyx instance per provider you want masked, each on its own port, and give each model its api_base.
Other providers. For a provider LiteLLM reaches in the OpenAI format, use the openai/ prefix with the model name and a Privyx instance whose upstream is that provider.