sambanova-pysdk 1.0.0

A Python SDK for the SambaNova cloud API

Sambanova Python SDK

A Python SDK for interacting with Sambanova's REST APIs, including chat, image, audio, transcription, translation, and embeddings.

Installation

pip install sambanova-sdk 

About Sambanova

Sambanova provides advanced AI infrastructure and models for generative AI, vision, audio, and language tasks. The Sambanova SDK enables seamless integration with Sambanova's REST APIs, allowing you to build, deploy, and scale AI-powered applications efficiently.

With Sambanova, you can:

• Access high-performance AI inference and training
• Use state-of-the-art models for chat, vision, audio, and more
• Scale your AI workloads with robust cloud and on-premise solutions

Documentation

The REST API documentation can be found on the Sambanova documentation portal. The full API of this library can be found in api.md.

Usage

Chat Completion

from src.sambanova.api_client import SambanovaAPIClient
from src.sambanova.chat import ChatCompletion

client = SambanovaAPIClient("your_api_key")
messages = [{"role": "user", "content": "Hello, how are you?"}]
stream = ChatCompletion.create(
    client,
    messages=messages,
    model="Llama-4-Maverick-17B-128E-Instruct"
)
response = ""
for chunk in stream:
    choices = chunk.get("choices", [])
    if not choices:
        continue
    delta = choices[0].get("delta", {})
    content = delta.get("content", "")
    response += content
print("Assistant:", response)

Image Completion

from src.sambanova.api_client import SambanovaAPIClient
from src.sambanova.image import Image
import base64

client = SambanovaAPIClient("your_api_key")
with open("path/to/image.jpg", "rb") as image_file:
    base64_image = base64.b64encode(image_file.read()).decode('utf-8')
messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Describe this image."},
            {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{base64_image}"}}
        ]
    }
]
response = Image.create(
    client,
    messages=messages,
    model="Llama-4-Maverick-17B-128E-Instruct",
    max_tokens=300,
    temperature=0.7,
    stream=False
)
print(response)

Embeddings

from src.sambanova.api_client import SambanovaAPIClient
from src.sambanova.embeddings import Embeddings

client = SambanovaAPIClient("your_api_key")
messages = [
    "Our solar system orbits the Milky Way galaxy at about 515,000 mph",
    "Jupiter's Great Red Spot is a storm that has been raging for at least 350 years."
]
response = Embeddings.create(
    client,
    messages=messages,
    model="E5-Mistral-7B-Instruct"
)
print(response)

Audio Reasoning

from src.sambanova.api_client import SambanovaAPIClient
from src.sambanova.audio import Audio
import base64

client = SambanovaAPIClient("your_api_key")
with open("path/to/audio.mp3", "rb") as audio_file:
    base64_audio = base64.b64encode(audio_file.read()).decode('utf-8')
messages = [
    {"role": "assistant", "content": "You are a helpful assistant."},
    {"role": "user", "content": [
        {"type": "audio_content", "audio_content": {"content": f"data:audio/mp3;base64,{base64_audio}"}}
    ]},
    {"role": "user", "content": "What is in this audio?"}
]
response = Audio.create(
    client,
    messages=messages,
    model="Qwen2-Audio-7B-Instruct",
    max_tokens=200,
)
print(response)

Transcription

from src.sambanova.api_client import SambanovaAPIClient
from src.sambanova.transcription import Transcription

client = SambanovaAPIClient("your_api_key")
audio_file_path = "path/to/audio.mp3"
response = Transcription.transcribe_audio_file(
    client,
    model="Whisper-Large-v3",
    audio_file_path=audio_file_path,
    language="english",
    response_format="text"
)
print(response)

API Key

Get an API Key from your Sambanova account and add it to your environment variables:

export SN_API_KEY="your-api-key-here"

You can also use python-dotenv to add SN_API_KEY="your-api-key-here" to your .env file.

Advanced Usage

Streaming Responses

The SDK supports streaming responses for chat and other endpoints. When streaming, usage and timing information will only be included in the final chunk.

stream = ChatCompletion.create(
    client,
    messages=messages,
    model="Llama-4-Maverick-17B-128E-Instruct",
    stream=True,
)
for chunk in stream:
    print(chunk.get("choices", [{}])[0].get("delta", {}).get("content", ""), end="")

Error Handling

When the SDK is unable to connect to the API (e.g., network issues or timeouts), a SambanovaAPIConnectionError is raised.

When the API returns a non-success status code (4xx or 5xx), a SambanovaAPIStatusError is raised, containing status_code and response properties.

All errors inherit from SambanovaAPIError.

from src.sambanova.api_client import SambanovaAPIClient, SambanovaAPIError

client = SambanovaAPIClient("your_api_key")
try:
    # ...your API call...
    pass
except SambanovaAPIError as e:
    print("An error occurred:", e)

Common error codes:

Status Code	Error Type
400	BadRequestError
401	AuthenticationError
403	PermissionDeniedError
404	NotFoundError
422	UnprocessableEntityError
429	RateLimitError
>=500	InternalServerError
N/A	SambanovaAPIConnectionError

Retries and Timeouts

The SDK automatically retries certain errors (connection errors, timeouts, 429, >=500) up to 2 times by default, with exponential backoff. You can configure retries and timeouts globally or per request.

client = SambanovaAPIClient("your_api_key", max_retries=3, timeout=30.0)

Logging

Enable logging by setting the environment variable:

export SAMBANOVA_LOG=info

Or use debug for more verbose output.

Custom Requests

You can make custom requests to undocumented endpoints using the request method:

response = client.request("POST", "/custom-endpoint", json={"param": "value"})
print(response)

Managing Resources

The SDK manages HTTP connections automatically. You can manually close the client if needed:

client.close()

Or use a context manager:

with SambanovaAPIClient("your_api_key") as client:
    # make requests
    pass
# client is now closed

Requirements

• Python 3.8 or higher

Contributing

Contributions are welcome! Please read our contributing guide for more information on how to get started.