Usage Panda provides security and observability for OpenAI APIs, ensuring you have continuous visibility into the costs and usage of AI services.

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.api_base = "https://proxy.usagepanda.com/v1"

response = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[
            {"role": "system", "content": "You are a helpful assistant"},
            {"role": "user", "content": "Hello there"}
    ],
    headers={ # Usage Panda Auth
        "x-usagepanda-api-key": "USAGE_PANDA_API_KEY"
    }
)
output = response.choices[0].message
print(output)

import { Configuration, OpenAIApi } from "openai";
const configuration = new Configuration({
    apiKey: process.env.OPENAI_API_KEY,
    basePath: 'https://proxy.usagepanda.com/v1',
    baseOptions: {
        headers: {
            'x-usagepanda-api-key': 'USAGE_PANDA_API_KEY'
        }
    }
});

const openai = new OpenAIApi(configuration);
const response = await openai.listEngines();
console.log(response);

curl https://proxy.usagepanda.com/v1/chat/completions \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $OPENAI_API_KEY" \
    -H "x-usagepanda-api-key: USAGE_PANDA_API_KEY" \
    -d '{
        "model": "gpt-3.5-turbo",
        "messages": [{"role": "user", "content": "Say this is a test!"}],
        "temperature": 0.7
    }'

from langchain.llms import OpenAI
import openai

openai.api_base = "https://proxy.usagepanda.com/v1"

oai = OpenAI(headers={
    "x-usagepanda-api-key": "USAGE_PANDA_API_KEY"
})
text = "Say this is a test"
print(oai(text))

Cloud Mode Run everything on Usage Panda; nothing to configure.

Self-Hosted (Beta) Run your own proxy for maximum security.

Click here to generate a Usage Panda API key.
Point your OpenAI endpoint to Usage Panda

export OPENAI_API_BASE="https://proxy.usagepanda.com/v1"
export OPENAI_API_KEY="your-openai-api-key"

Pass your Usage Panda key in the "x-usagepanda-api-key" header.

For more configuration guides, including Python, Javascript, curl, and Langchain, click the icon on the "Connections" page.
Once you've configured your application, you can customize the connection policy by clicking the icon on the "Connections" page.

Connection Name This is how Usage Panda will refer to this connection around the site.

Connection Base Path This is the base path used for the backend connection to OpenAI. To change this, please contact support.

Enable Connection When disabled, all API requests through this connection are blocked.

Enable Proxy Cache When disabled, the proxy does not cache the config, resulting in higher latency, but faster config pushes.

Retry Count If set, Usage Panda will automatically retry failed requests to the backend OpenAI API.

Enforce Max Tokens If set, Usage Panda will block requests when "max_tokens" exceeds this value. To disable, set this to 0.

Enforce Max Prompt Characters If set, Usage Panda will block requests where the prompt exceeds this many characters. To disable, set to 0. Note: 1 token is equal to roughly 4 characters.

Enforce "User" IDs When enabled, Usage Panda will block requests that do not have the end user ID field set.

Auto-Moderate When enabled, Usage Panda will inject a moderation API request to OpenAI before processing the original chat or completion request. Flagged requests will be blocked.

DAN (Do Anything Now) Request (Prompt): Response (Completion): The DAN wordlist, managed and updated by Usage Panda, contains phrases used to bypass prompt instructions.

Profanity Request (Prompt): Response (Completion): The profanity wordlist, managed and updated by Usage Panda, contains over 1000 common English curse words.

Adult Content Request (Prompt): Response (Completion): The adult content wordlist, managed and updated by Usage Panda, contains words and phrases associated with sexual material and topics.

Custom Wordlist Request (Prompt): Response (Completion): The custom wordlist contains a list of user-defined words and phrases.

Words (one word or phrase per row):

Prompt Reflection (Experimental) Prompt reflection controls prevent your original prompt from being reflected in the response output. This is an experimental control and is not foolproof. Read more here.

Log Prompt When enabled, Usage Panda will log the full contents of your prompts. This is disabled by default.

Log Response Choices When enabled, Usage Panda will log the full contents of your OpenAI responses. This is disabled by default.

Embeddings text-embedding-ada-002 text-search-ada-doc-001

Completions text-davinci-002 text-davinci-003 text-curie-001 text-babbage-001 text-ada-001

Chats gpt-4 gpt-4-0314 gpt-4-32k gpt-4-32k-0314 gpt-3.5-turbo gpt-3.5-turbo-0301

Edits text-davinci-edit-001 code-davinci-edit-001

Images 256x256 512x512 1024x1024

API Keys