AWS nova documentation change
Summary
Restructured documentation for extended thinking feature with updated sections, configuration details, code examples, and FAQ. Added parameter constraints, response structure examples, and clarified reasoning effort levels.
Security assessment
The changes focus on feature documentation improvements, parameter usage, and response structure updates. No security vulnerabilities, mitigations, or explicit security features are mentioned. The redacted reasoning content relates to billing/usage transparency rather than security controls.
Diff
diff --git a/nova/latest/nova2-userguide/extended-thinking.md b/nova/latest/nova2-userguide/extended-thinking.md index 1aeb2bc1a..3e63565e5 100644 --- a//nova/latest/nova2-userguide/extended-thinking.md +++ b//nova/latest/nova2-userguide/extended-thinking.md @@ -5 +5 @@ -Understanding extended thinkingQuick start: Enabling extended thinkingResponse structureThree reasoning effort levelsWhen to enable extended thinkingUnderstanding reasoning tokens and pricing +How Extended Thinking WorksConfiguration OptionsSupported modelsExtended thinking with tool useUnderstanding Reasoning Tokens and PricingFrequently Asked Questions @@ -9 +9 @@ Understanding extended thinkingQuick start: Enabling extended thinkingResponse s -Amazon Nova 2 introduces extended thinking capabilities that enable the model to engage in deeper reasoning for complex problem-solving tasks. +Amazon Nova 2 Lite introduces **extended thinking** capabilities that enable the model to engage in deeper reasoning for complex problems. This optional feature gives you control over when and how the model allocates additional computational resources to think through challenging tasks. @@ -11 +11 @@ Amazon Nova 2 introduces extended thinking capabilities that enable the model to -## Understanding extended thinking +## How Extended Thinking Works @@ -13,3 +13 @@ Amazon Nova 2 introduces extended thinking capabilities that enable the model to -Extended thinking represents a fundamental enhancement to how Amazon Nova processes complex requests. When enabled, Amazon Nova creates explicit reasoning chains before generating responses, similar to how humans think through problems step by step. - -### Two operating modes +Amazon Nova 2 introduces extended thinking as a **hybrid capability**. You have complete control: @@ -24 +22,13 @@ Extended thinking represents a fundamental enhancement to how Amazon Nova proces -## Quick start: Enabling extended thinking +### Reasoning effort + +The following shows different levels of reasoning effort. + +###### Topics + + * Low Effort (maxReasoningeffort:"low") + + * Medium effort (maxReasoningEffort: "medium") + + * High Effort (maxReasoningeffort:"high") + + @@ -26 +35,0 @@ Extended thinking represents a fundamental enhancement to how Amazon Nova proces -Extended thinking is controlled through the `reasoningConfig` parameter. @@ -28 +37,15 @@ Extended thinking is controlled through the `reasoningConfig` parameter. -When using the Converse API, add this to your request: +#### Low Effort (maxReasoningeffort:"low") + +**Best for:** Tasks with added complexity requiring structured thinking. For example, you can use this for code review and improvement suggestions where the model needs to carefully consider existing code quality, analysis tasks that require thoughtful consideration of multiple factors, or problem-solving scenarios that benefit from a methodical approach. Low effort is ideal for compound tasks where basic reasoning improves accuracy without requiring deep multi-step planning. + +#### Medium effort (maxReasoningEffort: "medium") + +**Best for:** Multi-step tasks and coding workflows. For example, you can use this for software development and debugging where the model needs to understand existing code structure before implementing changes, code generation that requires coordination across multiple files or components, multi-step calculations with interdependencies, or planning tasks with multiple constraints. Medium effort is optimal for agentic workflows that coordinate multiple tools and require the model to maintain context across several sequential operations. + +#### High Effort (maxReasoningeffort:"high") + +**Best for:** STEM reasoning and advanced problem-solving. For example, you can use this for advanced mathematical problems and proofs that require rigorous step-by-step verification, scientific analysis and research tasks demanding deep investigation, complex system design with architectural considerations across multiple dimensions, or critical decision-making scenarios with significant implications. High effort delivers maximum accuracy for tasks requiring sophisticated reasoning, careful evaluation of alternatives, and thorough validation of conclusions. + +### Quick start: Enabling Extended Thinking + +Extended thinking is controlled through the `reasoningConfig` parameter. @@ -33 +56 @@ When using the Converse API, add this to your request: - bedrock_runtime = boto3.client('bedrock-runtime', region_name='us-east-1') + bedrock = boto3.client('bedrock-runtime', region_name='us-east-1') @@ -35,2 +58,3 @@ When using the Converse API, add this to your request: - response = bedrock_runtime.converse( - modelId="us.amazon.nova-2-lite-v1:0", + response = bedrock.converse( + modelId='us.amazon.nova-2-lite-v1:0', + system=[{"text": "You are a highly capable personal assistant"}], @@ -39 +63 @@ When using the Converse API, add this to your request: - "content": [{"text": "Generate a step by step guide for cooking Beef Wellington."}] + "content": [{"text": "Provide a meal plan for a gluten free family of 4."}] @@ -42,2 +66,3 @@ When using the Converse API, add this to your request: - "maxTokens": 2048, - "temperature": 0.7 + "temperature": 0.7, + "topP": 0.9, + "maxTokens": 10000 @@ -48 +73 @@ When using the Converse API, add this to your request: - "maxReasoningEffort": "medium" + "maxReasoningEffort": "low" @@ -53 +77,0 @@ When using the Converse API, add this to your request: - # Print the response @@ -55,34 +78,0 @@ When using the Converse API, add this to your request: - for item in content_list: - if "reasoningContent" in item: - print("=== REASONING ===") - print(item["reasoningContent"]["reasoningText"]["text"]) - print() - elif "text" in item: - print("=== ANSWER ===") - print(item["text"]) - -When using the InvokeModel API, add this to your request: - - - import boto3 - import json - - bedrock_runtime = boto3.client('bedrock-runtime', region_name='us-east-1') - - response = bedrock_runtime.invoke_model( - modelId="us.amazon.nova-2-lite-v1:0", - body=json.dumps({ - "messages": [{"role": "user", "content": [{"text": "Your prompt here"}]}], - "inferenceConfig": { - "maxTokens": 2048, - "temperature": 0.7 - }, - "reasoningConfig": { - "type": "enabled", - "maxReasoningEffort": "medium" - } - }) - ) - - # Parse the response - response_body = json.loads(response['body'].read()) @@ -90,2 +79,0 @@ When using the InvokeModel API, add this to your request: - # Print the response - content_list = response_body["output"]["message"]["content"] @@ -93,0 +82 @@ When using the InvokeModel API, add this to your request: + reasoning_text = item["reasoningContent"]["reasoningText"]["text"] @@ -95 +84 @@ When using the InvokeModel API, add this to your request: - print(item["reasoningContent"]["reasoningText"]["text"]) + print(reasoning_text) @@ -101,60 +90 @@ When using the InvokeModel API, add this to your request: -For full examples of code that utilizes extended thinking, see [Code library](./code-library.html). - -## Response structure - -When you enable extended thinking, responses include `reasoningContent` blocks along with text content blocks: - -###### Note - -Reasoning content is displayed as `[REDACTED]`. You are still charged for reasoning tokens as they contribute to improved response quality. - -## Three reasoning effort levels - -Amazon Nova 2 provides flexible control over reasoning depth through three effort levels: - -### Low effort (maxReasoningEffort: "low") - -**Best for:** - - * Code review and improvement suggestions - - * Basic analysis tasks requiring some additional consideration - - * Simple problem-solving that benefits from structured approach - - * Most day-to-day development tasks with moderate complexity - - - - -**Characteristics:** - - * Provides meaningful reasoning enhancement - - * Maintains reasonable cost and latency - - * Suitable for tasks that require extra thought but not deep analysis - - - - -### Medium effort (maxReasoningEffort: "medium") - -**Best for:** - - * Complex debugging scenarios requiring systematic investigation - - * Multi-step calculations with interdependencies - - * Moderate planning tasks with multiple constraints - - * Analysis requiring evaluation of several alternatives - - - - -**Characteristics:** - - * Substantial reasoning depth for problems benefiting from thorough analysis - - * Balanced approach between depth and efficiency +Reasoning Parameters: @@ -162 +92 @@ Amazon Nova 2 provides flexible control over reasoning depth through three effor - * Most common choice for complex technical tasks + * `type: enabled` or `disabled` (default: `disabled`) @@ -163,0 +94 @@ Amazon Nova 2 provides flexible control over reasoning depth through three effor + * `maxReasoningEffort`: `low`, `medium`, or `high` @@ -167 +97,0 @@ Amazon Nova 2 provides flexible control over reasoning depth through three effor -### High effort (maxReasoningEffort: "high") @@ -169,37 +99 @@ Amazon Nova 2 provides flexible control over reasoning depth through three effor -**Best for:** - - * Advanced mathematical problems and proofs - - * Complex system design with multiple architectural considerations - - * Research tasks requiring deep analysis and synthesis - - * Critical decision-making scenarios with significant implications - - * Multi-step workflows requiring careful planning and verification -