bedrock-wrapper 2.3.0 → 2.4.0

package/CHANGELOG.md

@@ -1,6 +1,24 @@
 # Changelog
 All notable changes to this project will be documented in this file.
 
+## [2.4.0] - 2025-07-24 (AWS Nova Models)
+### Added
+- Support for AWS Nova models
+  - Nova-Pro (300K context, multimodal, 5K output tokens)
+  - Nova-Lite (300K context, multimodal, optimized for speed)
+  - Nova-Micro (128K context, text-only, lowest latency)
+- Nova-specific API format handling with schemaVersion "messages-v1"
+- Proper inferenceConfig parameter structure for Nova models
+- Automatic content array formatting for Nova message compatibility
+
+## [2.3.1] - 2025-05-22 (Claude 4 Opus / Sonnet)
+### Added
+- Support for Claude 4 Opus & Claude 4 Sonnet models
+  - Claude-4-Opus
+  - Claude-4-Opus-Thinking
+  - Claude-4-Sonnet
+  - Claude-4-Sonnet-Thinking
+
 ## [2.3.0] - 2025-02-15 (Claude 3.7 & Image Support)
 ### Added
 - Support for Claude 3.7 models

package/bedrock-models.js

@@ -6,6 +6,122 @@
 //       https://us-west-2.console.aws.amazon.com/bedrock/home?region=us-west-2#/cross-region-inference
 
 export const bedrock_models = [
+    {
+        // ====================
+        // == Claude 4 Opus ==
+        // ====================
+        "modelName":                     "Claude-4-Opus",
+        // "modelId":                       "anthropic.claude-opus-4-20250514-v1:0",
+        "modelId":                       "us.anthropic.claude-opus-4-20250514-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "max_tokens",
+        "max_supported_response_tokens": 131072,
+        "response_chunk_element":        "delta.text",
+        "response_nonchunk_element":     "content[0].text",
+        "thinking_response_chunk_element": "delta.thinking",
+        "thinking_response_nonchunk_element": "content[0].thinking",
+        "special_request_schema": {
+            "anthropic_version": "bedrock-2023-05-31",
+            "anthropic_beta": ["output-128k-2025-02-19"],
+        },
+        "image_support": {
+            "max_image_size": 20971520, // 20MB
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
+    {
+        // ==============================
+        // == Claude 4 Opus Thinking ==
+        // ==============================
+        "modelName":                     "Claude-4-Opus-Thinking",
+        // "modelId":                       "anthropic.claude-opus-4-20250514-v1:0",
+        "modelId":                       "us.anthropic.claude-opus-4-20250514-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "max_tokens",
+        "max_supported_response_tokens": 131072,
+        "response_chunk_element":        "delta.text",
+        "response_nonchunk_element":     "content[0].text",
+        "thinking_response_chunk_element": "delta.thinking",
+        "thinking_response_nonchunk_element": "content[0].thinking",
+        "special_request_schema": {
+            "anthropic_version": "bedrock-2023-05-31",
+            "anthropic_beta": ["output-128k-2025-02-19"],
+            "thinking": {
+                "type": "enabled",
+                "budget_tokens": 16000
+            },
+        },
+        "image_support": {
+            "max_image_size": 20971520, // 20MB
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
+    {
+        // =====================
+        // == Claude 4 Sonnet ==
+        // =====================
+        "modelName":                     "Claude-4-Sonnet",
+        // "modelId":                       "anthropic.claude-sonnet-4-20250514-v1:0",
+        "modelId":                       "us.anthropic.claude-sonnet-4-20250514-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "max_tokens",
+        "max_supported_response_tokens": 131072,
+        "response_chunk_element":        "delta.text",
+        "response_nonchunk_element":     "content[0].text",
+        "thinking_response_chunk_element": "delta.thinking",
+        "thinking_response_nonchunk_element": "content[0].thinking",
+        "special_request_schema": {
+            "anthropic_version": "bedrock-2023-05-31",
+            "anthropic_beta": ["output-128k-2025-02-19"],
+        },
+        "image_support": {
+            "max_image_size": 20971520, // 20MB
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
+    {
+        // ==============================
+        // == Claude 4 Sonnet Thinking ==
+        // ==============================
+        "modelName":                     "Claude-4-Sonnet-Thinking",
+        // "modelId":                       "anthropic.claude-sonnet-4-20250514-v1:0",
+        "modelId":                       "us.anthropic.claude-sonnet-4-20250514-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "max_tokens",
+        "max_supported_response_tokens": 131072,
+        "response_chunk_element":        "delta.text",
+        "response_nonchunk_element":     "content[0].text",
+        "thinking_response_chunk_element": "delta.thinking",
+        "thinking_response_nonchunk_element": "content[0].thinking",
+        "special_request_schema": {
+            "anthropic_version": "bedrock-2023-05-31",
+            "anthropic_beta": ["output-128k-2025-02-19"],
+            "thinking": {
+                "type": "enabled",
+                "budget_tokens": 16000
+            },
+        },
+        "image_support": {
+            "max_image_size": 20971520, // 20MB
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
     {
         // ================================
         // == Claude 3.7 Sonnet Thinking ==
@@ -424,6 +540,73 @@ export const bedrock_models = [
         "max_supported_response_tokens": 2048,
         "response_chunk_element":        "generation"
     },
+    {
+        // ===============
+        // == Nova Pro ==
+        // ===============
+        "modelName":                     "Nova-Pro",
+        "modelId":                       "us.amazon.nova-pro-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "maxTokens",
+        "max_supported_response_tokens": 5000,
+        "response_chunk_element":        "contentBlockDelta.delta.text",
+        "response_nonchunk_element":     "output.message.content[0].text",
+        "special_request_schema": {
+            "schemaVersion": "messages-v1",
+            "inferenceConfig": {}
+        },
+        "image_support": {
+            "max_image_size": 5242880, // 5MB per image
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
+    {
+        // ================
+        // == Nova Lite ==
+        // ================
+        "modelName":                     "Nova-Lite",
+        "modelId":                       "us.amazon.nova-lite-v1:0",
+        "vision":                        true,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "maxTokens",
+        "max_supported_response_tokens": 5000,
+        "response_chunk_element":        "contentBlockDelta.delta.text",
+        "response_nonchunk_element":     "output.message.content[0].text",
+        "special_request_schema": {
+            "schemaVersion": "messages-v1",
+            "inferenceConfig": {}
+        },
+        "image_support": {
+            "max_image_size": 5242880, // 5MB per image
+            "supported_formats": ["jpeg", "png", "gif", "webp"],
+            "max_images_per_request": 10
+        }
+    },
+    {
+        // =================
+        // == Nova Micro ==
+        // =================
+        "modelName":                     "Nova-Micro",
+        "modelId":                       "us.amazon.nova-micro-v1:0",
+        "vision":                        false,
+        "messages_api":                  true,
+        "system_as_separate_field":      true,
+        "display_role_names":            true,
+        "max_tokens_param_name":         "maxTokens",
+        "max_supported_response_tokens": 5000,
+        "response_chunk_element":        "contentBlockDelta.delta.text",
+        "response_nonchunk_element":     "output.message.content[0].text",
+        "special_request_schema": {
+            "schemaVersion": "messages-v1",
+            "inferenceConfig": {}
+        }
+    },
     {
         // ================
         // == Mistral-7b ==

package/bedrock-wrapper.js

@@ -232,14 +232,71 @@ export async function* bedrockWrapper(awsCreds, openaiChatCompletionsCreateObjec
     // }
 
     // Format the request payload using the model's native structure.
-    const request = awsModel.messages_api ? {
-        messages: prompt,
-        ...(awsModel.system_as_separate_field && system_message && { system: system_message }), // Only add system field if model requires it and there's a system message
-        [awsModel.max_tokens_param_name]: max_gen_tokens,
-        temperature: temperature,
-        top_p: top_p,
-        ...awsModel.special_request_schema
-    } : {
+    const request = awsModel.messages_api ? (() => {
+        // Check if this is a Nova model (has schemaVersion in special_request_schema)
+        if (awsModel.special_request_schema?.schemaVersion === "messages-v1") {
+            // Nova model format - convert messages to Nova's expected format
+            const novaMessages = prompt.map(msg => {
+                let content;
+                
+                // Convert content to array format for Nova
+                if (typeof msg.content === 'string') {
+                    content = [{ text: msg.content }];
+                } else if (Array.isArray(msg.content)) {
+                    // Already in array format, ensure proper structure
+                    content = msg.content.map(item => {
+                        if (item.type === 'text') {
+                            return { text: item.text || item };
+                        } else if (item.type === 'image') {
+                            return {
+                                image: {
+                                    format: 'jpeg',
+                                    source: {
+                                        bytes: item.source.data
+                                    }
+                                }
+                            };
+                        }
+                        return item;
+                    });
+                } else {
+                    content = [{ text: String(msg.content) }];
+                }
+                
+                return {
+                    role: msg.role,
+                    content: content
+                };
+            });
+            
+            const novaRequest = {
+                ...awsModel.special_request_schema,
+                messages: novaMessages,
+                inferenceConfig: {
+                    [awsModel.max_tokens_param_name]: max_gen_tokens,
+                    temperature: temperature,
+                    topP: top_p
+                }
+            };
+            
+            // Add system message if present
+            if (awsModel.system_as_separate_field && system_message) {
+                novaRequest.system = [{ text: system_message }];
+            }
+            
+            return novaRequest;
+        } else {
+            // Standard messages API format (Claude, etc.)
+            return {
+                messages: prompt,
+                ...(awsModel.system_as_separate_field && system_message && { system: system_message }),
+                [awsModel.max_tokens_param_name]: max_gen_tokens,
+                temperature: temperature,
+                top_p: top_p,
+                ...awsModel.special_request_schema
+            };
+        }
+    })() : {
         prompt: typeof prompt === 'string' ? prompt : {
             messages: prompt.map(msg => ({
                 role: msg.role,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bedrock-wrapper",
-  "version": "2.3.0",
+  "version": "2.4.0",
   "description": "🪨 Bedrock Wrapper is an npm package that simplifies the integration of existing OpenAI-compatible API objects with AWS Bedrock's serverless inference LLMs.",
   "homepage": "https://www.equilllabs.com/projects/bedrock-wrapper",
   "repository": {
@@ -32,9 +32,9 @@
   "author": "",
   "license": "ISC",
   "dependencies": {
-    "@aws-sdk/client-bedrock-runtime": "^3.755.0",
-    "dotenv": "^16.4.5",
-    "sharp": "^0.33.5"
+    "@aws-sdk/client-bedrock-runtime": "^3.848.0",
+    "dotenv": "^17.2.1",
+    "sharp": "^0.34.3"
   },
   "devDependencies": {
     "chalk": "^5.4.1"

package/test-vision.js

@@ -1,6 +1,8 @@
 import { bedrockWrapper } from "./bedrock-wrapper.js";
+import { bedrock_models } from "./bedrock-models.js";
 import dotenv from 'dotenv';
 import fs from 'fs/promises';
+import chalk from 'chalk';
 
 dotenv.config();
 
@@ -10,10 +12,37 @@ const awsCreds = {
     secretAccessKey: process.env.AWS_SECRET_ACCESS_KEY,
 };
 
+async function logOutput(message, type = 'info', writeToFile = true) {
+    if (writeToFile) {
+        // Log to file
+        await fs.appendFile('test-vision-models-output.txt', message + '\n');
+    }
+    
+    // Log to console with colors
+    switch(type) {
+        case 'success':
+            console.log(chalk.green('✓ ' + message));
+            break;
+        case 'error':
+            console.log(chalk.red('✗ ' + message));
+            break;
+        case 'info':
+            console.log(chalk.blue('ℹ ' + message));
+            break;
+        case 'running':
+            console.log(chalk.yellow(message));
+            break;
+        default:
+            console.log(message);
+    }
+}
+
 async function testVisionCapabilities() {
     // Read and convert image to base64
     const imageBuffer = await fs.readFile('./test-image.jpg');
     const base64Image = imageBuffer.toString('base64');
+    
+    const testPrompt = "What's in this image? Please describe it in detail.";
 
     const messages = [
         {
@@ -21,7 +50,7 @@ async function testVisionCapabilities() {
             content: [
                 {
                     type: "text",
-                    text: "What's in this image? Please describe it in detail."
+                    text: testPrompt
                 },
                 {
                     type: "image_url",
@@ -34,11 +63,25 @@ async function testVisionCapabilities() {
         }
     ];
 
-    // Test with both Claude and Llama models that support vision
-    const visionModels = ["Claude-3-5-Sonnet-v2", "Claude-3-7-Sonnet"];
+    // Filter vision-capable models from bedrock_models
+    const visionModels = bedrock_models
+        .filter(model => model.vision === true)
+        .map(model => model.modelName);
+
+    // Clear output file and add header
+    await fs.writeFile('test-vision-models-output.txt', 
+        `Vision Test Results\n` +
+        `Test Question: "${testPrompt}"\n` +
+        `Test Date: ${new Date().toISOString()}\n` +
+        `${'='.repeat(50)}\n\n`
+    );
+
+    console.clear();
+    await logOutput(`Starting vision tests with ${visionModels.length} models...`, 'info');
+    await logOutput(`Testing image description capabilities\n`, 'info');
 
     for (const model of visionModels) {
-        console.log(`\nTesting vision capabilities with ${model}...`);
+        await logOutput(`\n${'-'.repeat(50)}\nTesting ${model} ⇢`, 'running');
         
         const openaiChatCompletionsCreateObject = {
             messages,
@@ -49,23 +92,34 @@ async function testVisionCapabilities() {
         };
 
         try {
-            console.log(`\nSending request to ${model} with format:`, 
-                JSON.stringify(openaiChatCompletionsCreateObject, null, 2));
+            console.log(`\nSending request to ${model}...`);
             
             let response = "";
-            for await (const chunk of bedrockWrapper(awsCreds, openaiChatCompletionsCreateObject, { logging: true })) {
+            for await (const chunk of bedrockWrapper(awsCreds, openaiChatCompletionsCreateObject, { logging: false })) {
                 response += chunk;
                 process.stdout.write(chunk);
             }
-            console.log("\n-------------------");
+            
+            // Write successful response to file
+            await logOutput(`\nModel: ${model}`, 'success');
+            await logOutput(`Response: ${response.trim()}\n`, 'info', true);
+            
         } catch (error) {
-            console.error(`Error with ${model}:`, error);
-            // Log the full error details
+            const errorMessage = `Error with ${model}: ${error.message}`;
+            await logOutput(errorMessage, 'error');
+            
+            // Log the full error details to file
             if (error.response) {
-                console.error('Response error:', error.response);
+                await fs.appendFile('test-vision-models-output.txt', 
+                    `Error details: ${JSON.stringify(error.response, null, 2)}\n\n`
+                );
             }
         }
+        
+        console.log("\n-------------------");
     }
+    
+    await logOutput('\nVision testing complete! Check test-vision-models-output.txt for full results.', 'info', false);
 }
 
-testVisionCapabilities().catch(console.error); 
+testVisionCapabilities().catch(console.error);

package/test-models-output.txt

@@ -1,111 +0,0 @@
-Test Question: "Respond with exactly one word: What is 1+1?"
-==================================================
-
-Starting tests with 19 models...
-Each model will be tested with streaming and non-streaming calls
-
-
---------------------------------------------------
-Testing Claude-3-7-Sonnet-Thinking ⇢
-Streaming test passed for Claude-3-7-Sonnet-Thinking: "Two"
-Non-streaming test passed for Claude-3-7-Sonnet-Thinking: "<think>The question asks for exactly one word in response to "What is 1+1?"
-
-The answer to 1+1 is 2, which is a single word (a number). 
-
-I need to respond with exactly one word, and that word should be "2".</think>
-
-2"
-
---------------------------------------------------
-Testing Claude-3-7-Sonnet ⇢
-Streaming test passed for Claude-3-7-Sonnet: "2"
-Non-streaming test passed for Claude-3-7-Sonnet: "Two"
-
---------------------------------------------------
-Testing Claude-3-5-Sonnet-v2 ⇢
-Streaming test passed for Claude-3-5-Sonnet-v2: "two"
-Non-streaming test passed for Claude-3-5-Sonnet-v2: "two"
-
---------------------------------------------------
-Testing Claude-3-5-Sonnet ⇢
-Streaming test passed for Claude-3-5-Sonnet: "Two"
-Non-streaming test passed for Claude-3-5-Sonnet: "Two"
-
---------------------------------------------------
-Testing Claude-3-5-Haiku ⇢
-Streaming test passed for Claude-3-5-Haiku: "Two"
-Non-streaming test passed for Claude-3-5-Haiku: "Two"
-
---------------------------------------------------
-Testing Claude-3-Haiku ⇢
-Streaming test passed for Claude-3-Haiku: "Two."
-Non-streaming test passed for Claude-3-Haiku: "Two."
-
---------------------------------------------------
-Testing Llama-3-3-70b ⇢
-Streaming test passed for Llama-3-3-70b: "Two"
-Non-streaming test passed for Llama-3-3-70b: "Two."
-
---------------------------------------------------
-Testing Llama-3-2-1b ⇢
-Streaming test passed for Llama-3-2-1b: "Two"
-Non-streaming test passed for Llama-3-2-1b: "Two"
-
---------------------------------------------------
-Testing Llama-3-2-3b ⇢
-Streaming test passed for Llama-3-2-3b: "2"
-Non-streaming test passed for Llama-3-2-3b: "2"
-
---------------------------------------------------
-Testing Llama-3-2-11b ⇢
-Streaming test passed for Llama-3-2-11b: "Two."
-Non-streaming test passed for Llama-3-2-11b: "Two."
-
---------------------------------------------------
-Testing Llama-3-2-90b ⇢
-Streaming test passed for Llama-3-2-90b: "Two."
-Non-streaming test passed for Llama-3-2-90b: "Two."
-
---------------------------------------------------
-Testing Llama-3-1-8b ⇢
-Streaming test passed for Llama-3-1-8b: "Two."
-Non-streaming test passed for Llama-3-1-8b: "Two."
-
---------------------------------------------------
-Testing Llama-3-1-70b ⇢
-Streaming test passed for Llama-3-1-70b: "Two."
-Non-streaming test passed for Llama-3-1-70b: "Two."
-
---------------------------------------------------
-Testing Llama-3-1-405b ⇢
-Streaming test passed for Llama-3-1-405b: "Two"
-Non-streaming test passed for Llama-3-1-405b: "Two."
-
---------------------------------------------------
-Testing Llama-3-8b ⇢
-Streaming test passed for Llama-3-8b: "Two"
-Non-streaming test passed for Llama-3-8b: "Two"
-
---------------------------------------------------
-Testing Llama-3-70b ⇢
-Streaming test passed for Llama-3-70b: "Two"
-Non-streaming test passed for Llama-3-70b: "Two"
-
---------------------------------------------------
-Testing Mistral-7b ⇢
-Streaming test passed for Mistral-7b: "Two. (I've given you two words, but the first one was "What" which was not part of the mathematical equation.)"
-Non-streaming test passed for Mistral-7b: "Two. (I've given you two words, but the first one was "What" which was not part of the mathematical equation.)"
-
---------------------------------------------------
-Testing Mixtral-8x7b ⇢
-Streaming test passed for Mixtral-8x7b: "Two.
-
-The word you are looking for is "two." The sum of 1 + 1 is equal to 2. I am programmed to provide accurate and helpful responses, so I wanted to make sure that I gave you the correct answer to your question. Is there anything else you would like to know? I'm here to help! Just let me know if you have any other questions or if there's anything else I can do for you. I'm always happy to assist with whatever you need."
-Non-streaming test passed for Mixtral-8x7b: "Two.
-
-The question you asked is a simple arithmetic addition problem, and the answer is 2. It is considered good manners to respond to a direct question with a straightforward and accurate answer. However, I noticed that you initially asked for a one-word response, so I included only the word "Two" in my answer. I hope this is what you were looking for! Is there anything else I can help you with?"
-
---------------------------------------------------
-Testing Mistral-Large ⇢
-Streaming test passed for Mistral-Large: "Two."
-Non-streaming test passed for Mistral-Large: "Two."