Providers OpenAI

OpenAI Provider

TensorWall supports all OpenAI models through its native API integration.

Supported Models

Model	Type	Description
gpt-4o	Chat	Latest flagship model
gpt-4o-mini	Chat	Fast, cost-effective
gpt-4-turbo	Chat	Previous generation turbo
gpt-4	Chat	Base GPT-4
gpt-3.5-turbo	Chat	Fast and affordable
o1	Reasoning	Advanced reasoning
o1-mini	Reasoning	Efficient reasoning
text-embedding-3-small	Embedding	Small embedding model
text-embedding-3-large	Embedding	Large embedding model

Configuration

Environment Variables

# Optional - API URL override
OPENAI_API_URL=https://api.openai.com/v1

Request Headers

Pass your OpenAI API key in the Authorization header:

curl -X POST http://localhost:8000/v1/chat/completions \
  -H "X-API-Key: gw_your_gateway_key" \
  -H "Authorization: Bearer sk-your-openai-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Features

Streaming

Full streaming support with Server-Sent Events:

import requests

response = requests.post(
    "http://localhost:8000/v1/chat/completions",
    headers={
        "X-API-Key": "gw_xxx",
        "Authorization": "Bearer sk-xxx",
    },
    json={
        "model": "gpt-4o",
        "messages": [{"role": "user", "content": "Tell me a story"}],
        "stream": True,
    },
    stream=True,
)

for line in response.iter_lines():
    if line:
        print(line.decode())

Function Calling

Function calling is fully proxied:

response = requests.post(
    "http://localhost:8000/v1/chat/completions",
    headers={
        "X-API-Key": "gw_xxx",
        "Authorization": "Bearer sk-xxx",
    },
    json={
        "model": "gpt-4o",
        "messages": [{"role": "user", "content": "What's the weather in Paris?"}],
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "description": "Get current weather",
                    "parameters": {
                        "type": "object",
                        "properties": {
                            "location": {"type": "string"}
                        },
                        "required": ["location"]
                    }
                }
            }
        ]
    },
)

Vision

GPT-4o vision is supported:

response = requests.post(
    "http://localhost:8000/v1/chat/completions",
    headers={
        "X-API-Key": "gw_xxx",
        "Authorization": "Bearer sk-xxx",
    },
    json={
        "model": "gpt-4o",
        "messages": [
            {
                "role": "user",
                "content": [
                    {"type": "text", "text": "What's in this image?"},
                    {
                        "type": "image_url",
                        "image_url": {"url": "https://example.com/image.jpg"}
                    }
                ]
            }
        ]
    },
)

Cost Tracking

TensorWall automatically tracks costs based on OpenAI's pricing:

Model	Input (per 1M tokens)	Output (per 1M tokens)
gpt-4o	$2.50	$10.00
gpt-4o-mini	$0.15	$0.60
gpt-4-turbo	$10.00	$30.00
gpt-3.5-turbo	$0.50	$1.50

Error Handling

OpenAI errors are mapped to TensorWall error codes:

OpenAI Error	TensorWall Code	Description
401	AUTH_ERROR	Invalid API key
429	RATE_LIMITED	Rate limit exceeded
500	PROVIDER_ERROR	OpenAI service error
503	PROVIDER_UNAVAILABLE	Service temporarily unavailable