> ## Documentation Index
> Fetch the complete documentation index at: https://docs.freeplay.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Call Llama 3 on AWS SageMaker

> Call Llama models hosted on AWS SageMaker with Freeplay prompt management.

### 1. Configure Sagemaker Endpoint

Configure your Sagemaker endpoint in the Freeplay app and set your prompt to use it
[https://docs.freeplay.ai/account-setup/model-management#model-management](https://docs.freeplay.ai/account-setup/model-management#model-management)

### 2. Configure your Freeplay Client

### 3. Configure your Sagemaker Client

Configure your Sagemaker client using your AWS access keys

### 4. Fetch and Format your Prompt

Fetch your prompt from Freeplay, formatting it with your input variables

### 5. Call your Sagemaker Endpoint

Call your sagemaker endpoint directly using the formatted prompt object to key off needed parameters

### 6. Record to Freeplay

Record the interaction back to freeplay!

## Examples

<CodeGroup>
  ```python Python theme={null}
  from freeplay import Freeplay, RecordPayload, TestRunInfo, CallInfo
  import boto3
  import os
  from dotenv import load_dotenv
  import time
  import json

  load_dotenv("../.env")
  project_id = os.getenv("FREEPLAY_PROJECT_ID")
  freeplay_key = os.getenv("FREEPLAY_API_KEY")
  freeplay_url = os.getenv("FREEPLAY_URL")
  aws_access_key_id = os.getenv("AWS_ACCESS_KEY_ID")
  aws_secret_key = os.getenv("AWS_SECRET_KEY")

  fpClient = Freeplay(
      api_base=freeplay_url,
      freeplay_api_key=freeplay_key
  )

  sagemakerClient = boto3.client(
      'sagemaker-runtime',
      region_name='us-east-1',
      aws_access_key_id=aws_access_key_id,
      aws_secret_access_key=aws_secret_key
  )

  # fetch the prompt
  prompt_vars = {
      "pop_star": "Taylor Swift",
  }
  formatted_prompt = fpClient.prompts.get_formatted(
      project_id=project_id,
      template_name="album_bot",
      environment="latest",
      variables=prompt_vars
  )

  endpoint_name = formatted_prompt.prompt_info.provider_info['endpoint_name']
  inference_component_name = formatted_prompt.prompt_info.provider_info['inference_component_name']

  payload_body = {
      "inputs": formatted_prompt.llm_prompt_text,
      "parameters": formatted_prompt.prompt_info.model_parameters
  }

  # make the llm call
  start = time.time()
  response = sagemakerClient.invoke_endpoint(
      EndpointName=endpoint_name,
      InferenceComponentName=inference_component_name,
      ContentType='application/json',
      Accept='application/json',
      Body=json.dumps(payload_body)
  )
  end = time.time()

  response_content = json.loads(response['Body'].read().decode('utf-8'))['generated_text']
  print("response_content: ", response_content)

  all_messages = formatted_prompt.all_messages(
      {"role": "assistant",
      "content": response_content}
  )

  session = fpClient.sessions.create()

  fpClient.recordings.create(
    RecordPayload(
  				project_id=project_id,
          all_messages=all_messages,
          inputs=prompt_vars,
          session_info=session,
          prompt_version_info=formatted_prompt.prompt_info,
          call_info=CallInfo.from_prompt_info(formatted_prompt.prompt_info, start_time=start, end_time=end)
      )
  )
  ```
</CodeGroup>