Exemples d'API Converse InvokeModel Exemples d'API

Bibliothèque de codes

Cette section fournit des exemples de code pour les opérations Amazon Nova courantes à l'aide de l'API Converse ou de l' InvokeModel API.

Exemples d'API Converse

Demande de base

Envoyez une demande de texte de base aux modèles Amazon Nova à l'aide de l'API Converse.

Non-streaming


import boto3
from botocore.config import Config

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke the model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=[
        {
            "role": "user",
            "content": [{"text": "Write a short story. End the story with 'THE END'."}],
        }
    ],
    system=[{"text": "You are a children's book author."}],  # Optional
    inferenceConfig={  # These parameters are optional
        "maxTokens": 1500,
        "temperature": 0.7,
        "topP": 0.9,
        "stopSequences": ["THE END"],
    },
    additionalModelRequestFields={  # These parameters are optional
        "inferenceConfig": {
            "topK": 50,
        }
    },
)

# Extract the text response
content_list = response["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        print(content["text"])

Streaming


import boto3
from botocore.config import Config

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke the model
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=[
        {
            "role": "user",
            "content": [{"text": "Write a short story. End the story with 'THE END'."}],
        }
    ],
    system=[{"text": "You are a children's book author."}],  # Optional
    inferenceConfig={  # These parameters are optional
        "maxTokens": 1500,
        "temperature": 0.7,
        "topP": 0.9,
        "stopSequences": ["THE END"],
    },
    additionalModelRequestFields={  # These parameters are optional
        "inferenceConfig": {
            "topK": 50,
        }
    },
)

# Handle streaming events
for event in response["stream"]:
    if "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]
        if "text" in delta:
            print(delta["text"], end="", flush=True)

Entrée multimodale à l'aide d'un actif intégré

Traitez le contenu multimodal en incorporant des documents, des images, des vidéos ou des données audio directement dans la demande. Cet exemple utilise des données d'image. Pour en savoir plus sur la structure du contenu et les autres modalités, consultez la ContentBlock documentation de l'API Amazon Bedrock.

Non-streaming


import boto3
from botocore.config import Config

# Read a document, image, video, or audio file
with open("sample_image.png", "rb") as image_file:
    binary_data = image_file.read()
    data_format = "png"

# Define message with image
messages = [
    {
        "role": "user",
        "content": [
            {
                "image": {
                    "format": data_format,
                    "source": {
                        "bytes": binary_data  # For Invoke API, encode as Base64 string
                    },
                },
            },
            {"text": "Provide a brief caption for this asset."},
        ],
    }
]

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=messages,
)

# Extract the text response
content_list = response["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        print(content["text"])

Streaming


import boto3
from botocore.config import Config

# Read a document, image, video, or audio file
with open("sample_image.png", "rb") as image_file:
    binary_data = image_file.read()
    data_format = "png"

# Define message with image
messages = [
    {
        "role": "user",
        "content": [
            {
                "image": {
                    "format": data_format,
                    "source": {
                        "bytes": binary_data  # For Invoke API, encode as Base64 string
                    },
                },
            },
            {"text": "Provide a brief caption for this asset."},
        ],
    }
]

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke model with streaming
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=messages,
)

# Handle streaming events
for event in response["stream"]:
    if "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]
        if "text" in delta:
            print(delta["text"], end="", flush=True)

Entrée multimodale à l'aide de l'URI S3

Traitez le contenu multimodal en référençant des documents, des images, des vidéos ou des fichiers audio stockés dans S3. Cet exemple utilise une référence d'image. Pour en savoir plus sur la structure du contenu et les autres modalités, consultez la ContentBlock documentation de l'API Amazon Bedrock.

Non-streaming


import boto3
from botocore.config import Config

# Define message with image
messages = [
    {
        "role": "user",
        "content": [
            {
                "image": {
                    "format": "png",
                    "source": {
                        "s3Location": {
                            "uri": "s3://path/to/your/asset",
                            # "bucketOwner": "<account_id>" # Optional
                        }
                    },
                },
            },
            {"text": "Provide a brief caption for this asset."},
        ],
    }
]

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=messages,
)

# Extract the text response
content_list = response["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        print(content["text"])

Streaming


import boto3
from botocore.config import Config

# Define message with image
messages = [
    {
        "role": "user",
        "content": [
            {
                "image": {
                    "format": "png",
                    "source": {
                        "s3Location": {
                            "uri": "s3://path/to/your/asset",
                            # "bucketOwner": "<account_id>" # Optional
                        }
                    },
                },
            },
            {"text": "Provide a brief caption for this asset."},
        ],
    }
]

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke model with streaming
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=messages,
)

# Handle streaming events
for event in response["stream"]:
    if "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]
        if "text" in delta:
            print(delta["text"], end="", flush=True)

Pensée étendue (raisonnement)

Favorisez une réflexion approfondie pour les tâches complexes de résolution de problèmes.

Non-streaming


import boto3
from botocore.config import Config

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke the model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "text": 'How many capital letters appear in the following passage. Your response must include only the number: "Wilfred ordered an anvil from ACME. Shipping was expensive."'
                }
            ],
        }
    ],
    additionalModelRequestFields={
        "reasoningConfig": {
            "type": "enabled",
            "maxReasoningEffort": "low",  # "low" | "medium" | "high"
        }
    },
)

# Extract response content
content_list = response["output"]["message"]["content"]
for content in content_list:
    # Extract the reasoning response
    if "reasoningContent" in content:
        print("\n== Reasoning ==")
        print(content["reasoningContent"]["reasoningText"]["text"])
    # Extract the text response
    if "text" in content:
        print("\n== Text ==")
        print(content["text"])

Streaming


import boto3
from botocore.config import Config

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke the model
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "text": 'How many capital letters appear in the following passage. Your response must include only the number: "Wilfred ordered an anvil from ACME. Shipping was expensive."'
                }
            ],
        }
    ],
    additionalModelRequestFields={
        "reasoningConfig": {
            "type": "enabled",
            "maxReasoningEffort": "low",  # "low" | "medium" | "high"
        },
    },
)

# Process the streaming response
reasoning_output = ""
text_output = ""
for event in response["stream"]:
    if "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]

        if "reasoningContent" in delta:
            if len(reasoning_output) == 0:
                print("\n\n== Reasoning ==")
            reasoning_text_chunk = delta["reasoningContent"]["text"]
            print(reasoning_text_chunk, end="", flush=True)
            reasoning_output += reasoning_text_chunk

        elif "text" in delta:
            if len(text_output) == 0:
                print("\n\n== Text ==")
            text_chunk = delta["text"]
            print(text_chunk, end="", flush=True)
            text_output += text_chunk

Outil intégré : Nova Grounding avec citations

Utilisez Nova Grounding pour récupérer des informations en temps réel sur le Web avec des citations.

Non-streaming


import boto3
from botocore.config import Config

# Define the list of tools the model may use
tool_config = {"tools": [{"systemTool": {"name": "nova_grounding"}}]}

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

messages = [
    {
        "role": "user",
        "content": [
            {"text": "What is the latest news about renewable energy sources?"}
        ],
    }
]

# Invoke the model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

# Extract the text with interleaved citations
output_with_citations = ""
content_list = response["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        output_with_citations += content["text"]

    elif "citationsContent" in content:
        citations = content["citationsContent"]["citations"]
        for citation in citations:
            url = citation["location"]["web"]["url"]
            output_with_citations += f"[{url}]"

print(output_with_citations)

Streaming


import boto3
from botocore.config import Config

# Define the list of tools the model may use
tool_config = {"tools": [{"systemTool": {"name": "nova_grounding"}}]}

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

messages = [
    {
        "role": "user",
        "content": [
            {"text": "What is the latest news about renewable energy sources?"}
        ],
    }
]

# Invoke the model with streaming
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

# Process the streaming response with interleaved citations
for event in response["stream"]:
    if "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]

        if "text" in delta:
            print(delta["text"], end="", flush=True)

        elif "citation" in delta:
            url = delta["citation"]["location"]["web"]["url"]
            print(f"[{url}]", end="", flush=True)

Outil intégré : Interpréteur de code

Utilisez l'outil Code Interpreter pour exécuter du code Python pour les calculs et l'analyse des données.

Non-streaming


import boto3
from botocore.config import Config

# Define the list of tools the model may use
tool_config = {"tools": [{"systemTool": {"name": "nova_code_interpreter"}}]}

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

messages = [
    {
        "role": "user",
        "content": [
            {
                "text": "What is the average of 10, 24, 2, 3, 43, 52, 13, 68, 6, 7, 902, 82?"
            }
        ],
    }
]

# Invoke the model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

# Extract the text and the code the was executed
content_list = response["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        print("\n== Text ==")
        print(content["text"])

    elif "toolUse" in content and content["toolUse"]["name"] == "nova_code_interpreter":
        print("\n== Code Interpreter: input.snippet ==")
        print(content["toolUse"]["input"]["snippet"])

Streaming


import boto3
from botocore.config import Config
import json

# Define the list of tools the model may use
tool_config = {"tools": [{"systemTool": {"name": "nova_code_interpreter"}}]}

messages = [
    {
        "role": "user",
        "content": [
            {
                "text": "What is the average of 10, 24, 2, 3, 43, 52, 13, 68, 6, 7, 902, 82?"
            }
        ],
    }
]

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke the model with streaming
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

# Process the streaming response
current_block_start = None
response_text = ""
for event in response["stream"]:
    if "contentBlockStart" in event:
        current_block_start = event["contentBlockStart"]["start"]

    elif "contentBlockStop" in event:
        current_block_start = None

    elif "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]

        if (
            current_block_start
            and "toolUse" in current_block_start
            and current_block_start["toolUse"]["name"] == "nova_code_interpreter"
        ):
            # This is code interpreter content
            tool_input = json.loads(delta["toolUse"]["input"])
            print("\n== Executed Code Snippet ==")
            print(tool_input["snippet"], end="", flush=True)

        elif "text" in delta:
            # This is text response content
            if len(response_text) == 0:
                print("\n== Text ==")
            text = delta["text"]
            response_text += text
            print(text, end="", flush=True)

Utilisation d’outil

Définissez des outils personnalisés pour le modèle à utiliser pendant la conversation.

Non-streaming


import boto3
from botocore.config import Config


def get_weather(city):
    # Mock function to simulate weather API
    return {"temperatureF": 48, "conditions": "light rain"}


# Define the toolSpec for the weather tool
weather_tool = {
    "toolSpec": {
        "name": "get_weather",
        "description": "Get the current weather conditions in a given location",
        "inputSchema": {
            "json": {
                "type": "object",
                "properties": {
                    "city": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    }
                },
                "required": ["city"],
            }
        },
    }
}

# Define the list of tools the model may use
tool_config = {"tools": [weather_tool]}

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Start tracking message history
messages = []

messages.append(
    {
        "role": "user",
        "content": [
            {
                "text": "Suggest some activities to do in Seattle based on the current weather."
            }
        ],
    }
)

# Invoke the model
response = bedrock.converse(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

assistant_message = response["output"]["message"]

# Add the assistant response to the message history
messages.append(assistant_message)

content_list = assistant_message["content"]
stop_reason = response["stopReason"]

if stop_reason == "tool_use":
    # Extract the toolUse details
    tool_use = next(
        content["toolUse"] for content in content_list if "toolUse" in content
    )
    tool_name = tool_use["name"]
    tool_use_id = tool_use["toolUseId"]

    if tool_name == "get_weather":
        # Call the tool
        weather = get_weather(tool_use["input"]["city"])

        # Send the result back to the model
        messages.append(
            {
                "role": "user",
                "content": [
                    {
                        "toolResult": {
                            "toolUseId": tool_use_id,
                            "content": [{"json": weather}],
                        }
                    }
                ],
            }
        )

        # Submit the tool result back to the model
        response = bedrock.converse(
            modelId="us.amazon.nova-2-lite-v1:0",
            messages=messages,
            toolConfig=tool_config,
        )

        content_list = response["output"]["message"]["content"]
        for content in content_list:
            # Extract the text response
            if "text" in content:
                print("\n== Text ==")
                print(content["text"])
else:
    # A tool call was not needed
    for content in content_list:
        # Extract the text response
        if "text" in content:
            print("\n== Text ==")
            print(content["text"])

Streaming


import boto3
from botocore.config import Config
import json


def get_weather(city):
    # Mock function to simulate weather API
    return {"temperatureF": 48, "conditions": "light rain"}


# Define the toolSpec for the weather tool
weather_tool = {
    "toolSpec": {
        "name": "get_weather",
        "description": "Get the current weather conditions in a given location",
        "inputSchema": {
            "json": {
                "type": "object",
                "properties": {
                    "city": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    }
                },
                "required": ["city"],
            }
        },
    }
}

# Define the list of tools the model may use
tool_config = {"tools": [weather_tool]}

# Create the Bedrock Runtime client, using an extended timeout configuration
# to support long-running requests.
bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Start tracking message history
messages = []

messages.append(
    {
        "role": "user",
        "content": [
            {
                "text": "Suggest some activities to do in Seattle based on the current weather."
            }
        ],
    }
)

# Invoke the model with streaming
response = bedrock.converse_stream(
    modelId="us.amazon.nova-2-lite-v1:0", messages=messages, toolConfig=tool_config
)

# Process the streaming response
assistant_message = {"role": "assistant", "content": []}
current_tool_use = None
stop_reason = None

for event in response["stream"]:
    if "contentBlockStart" in event:
        start = event["contentBlockStart"]["start"]
        if "toolUse" in start:
            current_tool_use = start["toolUse"]
            current_tool_use["input"] = ""

    elif "contentBlockDelta" in event:
        delta = event["contentBlockDelta"]["delta"]
        if "toolUse" in delta:
            current_tool_use["input"] += delta["toolUse"]["input"]
        elif "text" in delta:
            print(delta["text"], end="", flush=True)

    elif "contentBlockStop" in event:
        if current_tool_use:
            # Parse the accumulated tool input
            current_tool_use["input"] = json.loads(current_tool_use["input"])
            assistant_message["content"].append({"toolUse": current_tool_use})
            current_tool_use = None

    elif "messageStop" in event:
        stop_reason = event["messageStop"]["stopReason"]
        if stop_reason == "end_turn":
            exit

# Add the assistant response to the message history
messages.append(assistant_message)

if stop_reason == "tool_use":
    # Extract the toolUse details
    tool_use = next(
        content["toolUse"]
        for content in assistant_message["content"]
        if "toolUse" in content
    )
    tool_name = tool_use["name"]
    tool_use_id = tool_use["toolUseId"]

    if tool_name == "get_weather":
        # Call the tool
        weather = get_weather(tool_use["input"]["city"])

        # Send the result back to the model
        messages.append(
            {
                "role": "user",
                "content": [
                    {
                        "toolResult": {
                            "toolUseId": tool_use_id,
                            "content": [{"json": weather}],
                        }
                    }
                ],
            }
        )

        # Submit the tool result back to the model with streaming
        response = bedrock.converse_stream(
            modelId="us.amazon.nova-2-lite-v1:0",
            messages=messages,
            toolConfig=tool_config,
        )

        # Handle the final streaming response
        print("\n== Text ==")
        for event in response["stream"]:
            if "contentBlockDelta" in event:
                delta = event["contentBlockDelta"]["delta"]
                if "text" in delta:
                    print(delta["text"], end="", flush=True)

InvokeModel Exemples d'API

Les exemples ci-dessous se concentrent sur les quelques domaines clés dans lesquels les structures de demande et de réponse de l'API Invoke diffèrent légèrement de celles de l'API Converse. Dans la plupart des autres cas, les deux APIs sont compatibles. Vous devriez donc être en mesure d'adapter facilement les exemples d'API Converse ci-dessus pour qu'ils fonctionnent avec l' InvokeModel API.

Demande de base

Envoyez une demande de texte de base aux modèles Amazon Nova 2 à l'aide de l' InvokeModel API.

Non-streaming


import json

import boto3
from botocore.config import Config

# Configure the request
request_body = {
    "messages": [
        {
            "role": "user",
            "content": [{"text": "Write a short story. End the story with 'THE END'."}],
        }
    ],
    "system": [{"text": "You are a children's book author."}],  # Optional
    "inferenceConfig": {  # These parameters are optional
        "maxTokens": 1500,
        "temperature": 0.7,
        "topP": 0.9,
        "topK": 50,
        "stopSequences": ["THE END"],
    },
}

bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke the model
response = bedrock.invoke_model(
    modelId="us.amazon.nova-2-lite-v1:0", body=json.dumps(request_body)
)
response_body = json.loads(response["body"].read())

# Extract the text response
content_list = response_body["output"]["message"]["content"]
for content in content_list:
    if "text" in content:
        print(content["text"])

Streaming


import json

import boto3
from botocore.config import Config

# Configure the request
request_body = {
    "messages": [
        {
            "role": "user",
            "content": [{"text": "Write a short story. End the story with 'THE END'."}],
        }
    ],
    "system": [{"text": "You are a children's book author."}],  # Optional
    "inferenceConfig": {  # These parameters are optional
        "maxTokens": 1500,
        "temperature": 0.7,
        "topP": 0.9,
        "topK": 50,
        "stopSequences": ["THE END"],
    },
}

bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke the model with streaming
response = bedrock.invoke_model_with_response_stream(
    modelId="us.amazon.nova-2-lite-v1:0", body=json.dumps(request_body)
)

# Process the streaming response
for event in response["body"]:
    chunk = json.loads(event["chunk"]["bytes"])
    if "contentBlockDelta" in chunk:
        delta = chunk["contentBlockDelta"]["delta"]
        if "text" in delta:
            print(delta["text"], end="", flush=True)

InvokeModel API avec raisonnement

Utilisez l' InvokeModel API avec le raisonnement activé pour la résolution de problèmes complexes.

Non-streaming


import json

import boto3
from botocore.config import Config

# Configure the request
request_body = {
    "messages": [
        {
            "role": "user",
            "content": [
                {
                    "text": 'How many capital letters appear in the following passage. Your response must include only the number: "Wilfred ordered an anvil from ACME. Shipping was expensive."'
                }
            ],
        }
    ],
    "reasoningConfig": {
        "type": "enabled",
        "maxReasoningEffort": "low",  # "low" | "medium" | "high"
    },
}

bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(read_timeout=3600),
)

# Invoke the model
response = bedrock.invoke_model(
    modelId="us.amazon.nova-2-lite-v1:0", body=json.dumps(request_body)
)
response_body = json.loads(response["body"].read())

# Extract response content
content_list = response_body["output"]["message"]["content"]
for content in content_list:
    # Extract the reasoning response
    if "reasoningContent" in content:
        print("\n== Reasoning ==")
        print(content["reasoningContent"]["reasoningText"]["text"])
    # Extract the text response
    if "text" in content:
        print("\n== Text ==")
        print(content["text"])

Streaming


import json

import boto3
from botocore.config import Config

# Configure the request
request_body = {
    "messages": [
        {
            "role": "user",
            "content": [
                {
                    "text": 'How many capital letters appear in the following passage. Your response must include only the number: "Wilfred ordered an anvil from ACME. Shipping was expensive."'
                }
            ],
        }
    ],
    "reasoningConfig": {
        "type": "enabled",
        "maxReasoningEffort": "low",  # "low" | "medium" | "high"
    },
}

bedrock = boto3.client(
    "bedrock-runtime",
    region_name="us-east-1",
    config=Config(connect_timeout=3600, read_timeout=3600),
)

# Invoke the model with streaming
response = bedrock.invoke_model_with_response_stream(
    modelId="us.amazon.nova-2-lite-v1:0", body=json.dumps(request_body)
)

# Process the streaming response
for event in response["body"]:
    chunk = json.loads(event["chunk"]["bytes"])

    if "contentBlockDelta" in chunk:
        delta = chunk["contentBlockDelta"]["delta"]

        # Extract the reasoning response
        if "reasoningContent" in delta:
            print("\n== Reasoning ==")
            print(delta["reasoningContent"]["reasoningText"]["text"], end="", flush=True)

        # Extract the text response
        if "text" in delta:
            print("\n== Text ==")
            print(delta["text"], end="", flush=True)

Avertissement JavaScript est désactivé ou n'est pas disponible dans votre navigateur.

Pour que vous puissiez utiliser la documentation AWS, Javascript doit être activé. Vous trouverez des instructions sur les pages d'aide de votre navigateur.

Conventions de rédaction

Code et résolution des problèmes

Résolution des problèmes