Use LiteLLM with an uncensored LLM API

LiteLLM routes by model prefix. For an OpenAI-compatible server like Wild West the prefix is openai/, and the server address goes in api_base.

Set it up

1. Sign in at wildwestapi.com/signup/ with the six-digit code sent to your email, open the dashboard and create an API key. It starts with sk-ww- and is shown in full once, so copy it somewhere safe.
2. Install with pip install litellm.
3. Export WILDWEST_API_KEY.
4. Call litellm.completion with model openai/outlaw-1, api_base https://wildwestapi.com/v1 and api_key from the environment.
5. For a shared gateway, add the same values to a proxy config.yaml and start it with litellm --config.

SDK call

The docs say to prefix the model name with openai/ for the chat completions route so LiteLLM sends it to an OpenAI-compatible endpoint. Without the prefix it tries to guess the provider from the name and outlaw-1 matches none.

import os, litellm

resp = litellm.completion(
    model="openai/outlaw-1",
    api_base="https://wildwestapi.com/v1",
    api_key=os.environ["WILDWEST_API_KEY"],
    messages=[{"role": "user", "content": "Write the opening of a noir story."}],
)
print(resp.choices[0].message.content)

Keep /v1 on api_base. The docs advise checking that it ends in /v1 if you see Not Found errors.

Proxy config.yaml

The proxy lets many apps share one entry point. Each model_list item has a public model_name that your apps use and litellm_params that say where it goes. The os.environ/ form reads the key from the environment so it stays out of the file.

model_list:
  - model_name: outlaw
    litellm_params:
      model: openai/outlaw-1
      api_base: https://wildwestapi.com/v1
      api_key: os.environ/WILDWEST_API_KEY

Start it with litellm --config config.yaml. It listens on port 4000 by default, and any OpenAI client can then use http://localhost:4000 with the model name outlaw.

Tool calling and streaming

Both pass through as normal OpenAI-format parameters, so stream=True and tools=[...] work the way they do against any other provider, subject to the model supporting tools.

One limit to know

The proxy is a second hop that you host. Your prompts pass through it, and anyone who can reach port 4000 without a LiteLLM key can spend your Wild West balance. Do not expose it publicly without setting up proxy keys.

FAQ

Related