smartllm 0.1.6

A unified async Python wrapper for multiple LLM providers with OpenAI Response API and reasoning support

SmartLLM

A unified async Python wrapper for multiple LLM providers with a consistent interface.

Features

• Unified Interface - Single API for multiple LLM providers (OpenAI, AWS Bedrock)
• Async/Await - Built on asyncio for high-performance concurrent requests
• Smart Caching - Two-level cache (local + DynamoDB) to reduce costs and latency
• Auto Retry - Exponential backoff retry logic for transient failures
• Structured Output - Native Pydantic model support for type-safe responses
• Streaming - Real-time streaming responses for better UX
• Rate Limiting - Built-in concurrency control per model
• Decorator Logging - Automatic function logging via Logorator
• OpenAI Response API - Full support for OpenAI's primary API including reasoning models

Installation

pip install smartllm

Optional Dependencies

Install only the providers you need:

# For OpenAI
pip install smartllm[openai]

# For AWS Bedrock
pip install smartllm[bedrock]

# For all providers
pip install smartllm[all]

DynamoDB Caching (optional)

To enable shared two-level caching across machines:

async with LLMClient(provider="openai", dynamo_table_name="my-llm-cache") as client:
    ...

Requires AWS credentials with DynamoDB access. The table is auto-created if it doesn't exist. Local file cache is always used as the first layer.

Quick Start

Basic Usage

import asyncio
from smartllm import LLMClient, TextRequest

async def main():
    # Auto-detects provider from environment variables
    async with LLMClient(provider="openai") as client:
        response = await client.generate_text(
            TextRequest(prompt="What is the capital of France?")
        )
        print(response.text)

asyncio.run(main())

Multi-turn Conversations

from smartllm import LLMClient, MessageRequest, Message

async with LLMClient(provider="openai") as client:
    messages = [
        Message(role="user", content="My name is Alice."),
        Message(role="assistant", content="Nice to meet you, Alice!"),
        Message(role="user", content="What's my name?"),
    ]
    
    response = await client.send_message(
        MessageRequest(messages=messages)
    )
    print(response.text)  # "Your name is Alice."

Streaming Responses

from smartllm import LLMClient, TextRequest

async with LLMClient(provider="openai") as client:
    request = TextRequest(
        prompt="Write a short poem about Python.",
        stream=True
    )
    
    async for chunk in client.generate_text_stream(request):
        print(chunk.text, end="", flush=True)

Structured Output with Pydantic

from pydantic import BaseModel
from smartllm import LLMClient, TextRequest

class Person(BaseModel):
    name: str
    age: int
    occupation: str

async with LLMClient(provider="openai") as client:
    response = await client.generate_text(
        TextRequest(
            prompt="Generate a person profile for a software engineer named John, age 30.",
            response_format=Person
        )
    )
    
    person = response.structured_data
    print(f"{person.name} is a {person.age} year old {person.occupation}")

Configuration

Environment Variables

OpenAI:

export OPENAI_API_KEY="your-api-key"
export OPENAI_MODEL="gpt-4o-mini"  # Optional

AWS Bedrock:

export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export AWS_REGION="us-east-1"
export BEDROCK_MODEL="anthropic.claude-3-sonnet-20240229-v1:0"  # Optional

Programmatic Configuration

from smartllm import LLMClient, LLMConfig

config = LLMConfig(
    provider="openai",
    api_key="your-api-key",
    default_model="gpt-4o",
    temperature=0.7,
    max_tokens=2048,
    max_retries=3,
)

async with LLMClient(config) as client:
    # Use client...
    pass

Customizing Defaults

from smartllm import defaults

# Modify global defaults
defaults.DEFAULT_TEMPERATURE = 0.7
defaults.DEFAULT_MAX_TOKENS = 4096
defaults.DEFAULT_MAX_RETRIES = 5

OpenAI API Types

SmartLLM supports both OpenAI APIs via the api_type parameter:

• "responses" (default) - OpenAI's primary Response API, recommended for all modern models
• "chat_completions" - Legacy Chat Completions API, supported indefinitely

# Response API (default)
response = await client.generate_text(
    TextRequest(prompt="Hello", api_type="responses")
)

# Chat Completions API (legacy)
response = await client.generate_text(
    TextRequest(prompt="Hello", api_type="chat_completions")
)

Reasoning Models

For models that support reasoning (e.g. GPT-5.x), use reasoning_effort to control how much the model reasons before responding. Reasoning tokens are returned in response.metadata:

response = await client.generate_text(
    TextRequest(
        prompt="Solve: what is the 100th Fibonacci number?",
        reasoning_effort="high",  # "low", "medium", or "high"
    )
)

print(response.text)
print(f"Reasoning tokens used: {response.metadata.get('reasoning_tokens', 0)}")

Note: reasoning models do not support temperature. Passing a value other than 1 will raise a ValueError.

Reasoning with Structured Output

from pydantic import BaseModel
from smartllm import LLMClient, TextRequest

class Solution(BaseModel):
    answer: float
    unit: str
    explanation: str

async with LLMClient(provider="openai") as client:
    response = await client.generate_text(
        TextRequest(
            prompt="A train leaves city A at 60mph toward city B (300 miles away). Another leaves B at 90mph. When do they meet?",
            response_format=Solution,
            reasoning_effort="medium",
        )
    )

    solution = response.structured_data
    print(f"{solution.answer} {solution.unit}: {solution.explanation}")
    print(f"Reasoning tokens: {response.metadata.get('reasoning_tokens', 0)}")

Advanced Features

Caching

Responses are automatically cached when temperature=0:

# First call - hits API
response1 = await client.generate_text(
    TextRequest(prompt="What is 2+2?", temperature=0)
)

# Second call - uses cache (instant, free)
response2 = await client.generate_text(
    TextRequest(prompt="What is 2+2?", temperature=0)
)

# Clear cache for specific request
response3 = await client.generate_text(
    TextRequest(prompt="What is 2+2?", temperature=0, clear_cache=True)
)

Concurrent Requests

import asyncio
from smartllm import LLMClient, TextRequest

async with LLMClient(provider="openai") as client:
    prompts = ["Question 1", "Question 2", "Question 3"]
    
    tasks = [
        client.generate_text(TextRequest(prompt=p))
        for p in prompts
    ]
    
    responses = await asyncio.gather(*tasks)

Rate Limiting

# Limit concurrent requests
client = LLMClient(provider="openai", max_concurrent=5)

Provider-Specific Clients

For advanced use cases, access provider-specific clients:

from smartllm.openai import OpenAILLMClient, OpenAIConfig
from smartllm.bedrock import BedrockLLMClient, BedrockConfig

# OpenAI-specific features
openai_config = OpenAIConfig(api_key="...", organization="...")
async with OpenAILLMClient(openai_config) as client:
    models = await client.list_available_models()

# Bedrock-specific features
bedrock_config = BedrockConfig(aws_region="us-east-1")
async with BedrockLLMClient(bedrock_config) as client:
    models = await client.list_available_model_ids()

Supported Providers

• OpenAI - GPT models via OpenAI API
• AWS Bedrock - Claude, Llama, Mistral, and Titan models

API Reference

Core Classes

• LLMClient - Unified client for all providers
• LLMConfig - Unified configuration
• TextRequest - Single prompt request
• MessageRequest - Multi-turn conversation request
• TextResponse - LLM response with metadata
• Message - Conversation message
• StreamChunk - Streaming response chunk

Request Parameters

Parameter	Type	Description	Default
prompt	str	Input text prompt	Required
model	str	Model ID to use	Config default
temperature	float	Sampling temperature (0-1)	0
max_tokens	int	Maximum output tokens	2048
top_p	float	Nucleus sampling	1.0
system_prompt	str	System context	None
stream	bool	Enable streaming	False
response_format	BaseModel	Pydantic model for structured output	None
use_cache	bool	Enable caching	True
clear_cache	bool	Clear cache before request	False
api_type	str	OpenAI API type ("responses" or "chat_completions")	"responses"
reasoning_effort	str	Reasoning effort ("low", "medium", "high")	None

Error Handling

from smartllm import LLMClient, TextRequest

async with LLMClient(provider="openai") as client:
    try:
        response = await client.generate_text(
            TextRequest(prompt="Hello")
        )
    except ValueError as e:
        print(f"Configuration error: {e}")
    except Exception as e:
        print(f"API error: {e}")

Development

Setup

git clone https://github.com/Redundando/smartllm.git
cd smartllm
pip install -e .[all,dev]

Running Tests

# Unit tests
pytest tests/unit/ -v

# Integration tests (select model interactively)
pytest tests/integration/

# Integration tests with a specific model
pytest tests/integration/ --model gpt-4o

# Integration tests with a reasoning model
pytest tests/integration/ --model gpt-5.2

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Changelog

Version 0.1.5

• Replaced custom logging with Logorator decorator-based logging
• Added two-level cache: local JSON files + optional DynamoDB via Dynamorator
• DynamoDB cache configurable via dynamo_table_name and cache_ttl_days (default: 365 days)
• Cache write-back: DynamoDB hits are written to local cache automatically
• Prompt stored in cache metadata
• Recursive Pydantic schema cleaning for OpenAI structured output compatibility
• logorator and dynamorator added as core dependencies in pyproject.toml

Version 0.1.4

• Fixed logger name from aws_llm_wrapper to smartllm
• Removed redundant response_format=json_object when using tool-based structured output
• Cache read failures now log a warning instead of silently returning None
• Added reasoning_effort warning when used with Bedrock models
• Test suite now supports model selection via --model CLI option or interactive prompt
• Integration tests support both OpenAI and AWS Bedrock models
• Bedrock streaming chunk parsing fixed for Claude models

Version 0.1.0

• Initial public release
• Unified interface for multiple providers
• OpenAI support (GPT models)
• AWS Bedrock support (Claude, Llama, Mistral, Titan)
• Async/await architecture
• Smart caching with temperature=0
• Auto retry with exponential backoff
• Structured output with Pydantic models
• Streaming responses
• Rate limiting and concurrency control
• OpenAI Response API support (primary interface)
• Reasoning model support with reasoning_effort parameter
• Comprehensive test suite

Support

• Issues: GitHub Issues
• Email: [e-mail]

Acknowledgments

Built with:

• Pydantic for data validation
• Logorator for decorator-based logging
• Dynamorator for DynamoDB caching
• aioboto3 for AWS async support
• OpenAI Python SDK for OpenAI integration