This repository demonstrates how to use Hugging Face's smolagents and LiteLLM's model abstraction with Amazon Bedrock's language models. These examples show how to leverage these libraries for both agent-based interactions (Bedrock + smolagents + LiteLLM) and direct model access (Bedrock + LiteLLM).

• Python 3.8+
• AWS account with Bedrock access
• AWS credentials configured locally

1. Clone the repository:

git clone <repository-url>
cd <project-directory>

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install required packages:

pip install -r requirements.txt

1. Ensure AWS credentials are properly configured in ~/.aws/credentials or via environment variables:

export AWS_ACCESS_KEY_ID=your_access_key
export AWS_SECRET_ACCESS_KEY=your_secret_key
export AWS_DEFAULT_REGION=us-west-2

2. The provided configuration uses:

• Model: bedrock/us.amazon.nova-lite-v1:0
• Region: us-west-2 (or whatever is in AWS_DEFAULT_REGION)

• smol-demo.py: Example using SmoL Agents with Bedrock
• litellm-demo.py: Basic example of using LiteLLM with Bedrock
• x-bonus-litellm-bedrock-openai-demo.py: Advanced LiteLLM example with OpenAI-compatible interface

The primary demo shows how to create an agent with search capabilities using smolagents. The code was taken and adapted from Hugging Face's great blog post here: https://huggingface.co/blog/smolagents

from smolagents import CodeAgent, DuckDuckGoSearchTool, LiteLLMModel

model = LiteLLMModel(model_id="bedrock/us.amazon.nova-lite-v1:0")
agent = CodeAgent(tools=[DuckDuckGoSearchTool()], model=model)

agent.run("How many seconds would it take for a leopard at full speed to run through Pont des Arts?")

This example demonstrates:

• Integration with Amazon Bedrock's Nova model
• Use of smolagents' simple CodeAgent for complex reasoning

The smolagents library introduced me to LiteLLM (https://www.litellm.ai/). LiteLLM provides an SDK and/or a proxy enabling you to access 100's of different LLMs from different providers all via an OpenAI compatible interface. Here is a simple completion example using LiteLLM with Bedrock's Nova model:

from litellm import completion
import json

response = completion(
  model="bedrock/us.amazon.nova-lite-v1:0",
  messages=[{ "content": "Hello, how are you?","role": "user"}]
)

As a bonus demo, here is sample code from OpenAI with minimal changes working with Amazon Bedrock. This shows OpenAI-compatible streaming and non-streaming completions from Amazon Bedrock's Nova models.

x-bonus-litellm-bedrock-openai-demo.py

See CONTRIBUTING for more information.

This library is licensed under the MIT-0 License. See the LICENSE file.