OneSDK: Unified LLM API Interface

OneSDK is a Python library providing a unified interface for various Large Language Model (LLM) providers. It simplifies interactions with different LLM APIs through a consistent set of methods.

Features

• Unified API for multiple LLM providers
• Flexible usage: per-call model specification or default model setting
• Intuitive interface for common LLM operations
• Synchronous and streaming text generation support
• Token counting functionality
• Embedding creation (for supported providers)
• Image generation (for supported providers)
• File operations (for supported providers)
• Proxy setting for API calls
• Usage statistics retrieval (for supported providers)

Installation

pip install llm_onesdk

Quick Start

OneSDK supports two main usage patterns:

1. Specify model for each call

from llm_onesdk import OneSDK

sdk = OneSDK("openai", {"api_key": "your-api-key"})

response = sdk.generate(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Tell me a joke about programming."}]
)
print(response['choices'][0]['message']['content'])

2. Set a default model

from llm_onesdk import OneSDK

sdk = OneSDK("openai", {"api_key": "your-api-key"})
sdk.set_model("gpt-3.5-turbo")

response = sdk.generate(
    messages=[{"role": "user", "content": "Tell me a joke about programming."}]
)
print(response['choices'][0]['message']['content'])

Streaming Generation

for chunk in sdk.stream_generate(
    model="gpt-3.5-turbo",  # Optional if using set_model()
    messages=[{"role": "user", "content": "Write a short story about AI."}]
):
    print(chunk['choices'][0]['message']['content'], end='', flush=True)

Additional Operations

# List models (for supported providers)
models = sdk.list_models()
print(models)

# Count tokens
token_count = sdk.count_tokens(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "How many tokens is this?"}]
)
print(f"Token count: {token_count}")

# Create embeddings (for supported providers)
embeddings = sdk.create_embedding(
    model="text-embedding-ada-002",
    input="Hello, world!"
)
print(embeddings)

# Generate image (for supported providers)
image_response = sdk.create_image("A futuristic city with flying cars")
print(image_response)

Supported Providers and Core Methods

The following table shows the supported providers, their core method support, and additional features:

✓: Supported, ✗: Not supported

Notes:

1. Some providers may have additional provider-specific methods. Refer to individual provider documentation for details.
2. Providers marked with * (Ollama, Gemini, and Cohere) are currently not fully tested. The documentation for these providers is for reference only and may not be entirely accurate or up-to-date. We are working on improving these integrations and will provide more accurate information in future updates.
3. The "Additional Features" column summarizes some unique or extra functionalities of each provider. The availability and usage of specific features may change over time; please refer to the latest official documentation.

Key Methods

• set_model(model): Set default model
• list_models(): List available models (if supported)
• generate(messages, model=None, **kwargs): Generate response
• stream_generate(messages, model=None, **kwargs): Stream response
• count_tokens(model, messages): Count tokens
• create_embedding(model, input, **kwargs): Create embeddings (if supported)
• create_image(prompt, **kwargs): Create image (if supported)
• upload_file(file_path): Upload file (if supported)
• set_proxy(proxy_url): Set proxy for API calls

Error Handling

OneSDK uses custom exceptions inheriting from InvokeError. Always wrap API calls in try-except blocks:

from llm_onesdk.utils.error_handler import InvokeError

try:
    response = sdk.generate(model, messages)
except InvokeError as e:
    print(f"An error occurred: {str(e)}")

Documentation

For detailed information on each provider's capabilities and usage, please refer to the individual documentation files in the docs/ directory.

Contributing

We welcome contributions, especially new provider integrations! See our Contributing Guide for details.

License

This project is under the MIT License. See the LICENSE file for details.