bedrock-wrapper 2.6.1 → 2.7.0

package/bedrock-wrapper.js

@@ -144,10 +144,18 @@ async function convertToConverseFormat(messages) {
             }
             
             // Only add messages with actual content (Converse API doesn't allow empty content)
-            if (content.length > 0) {
+            // Filter out content items with empty text
+            const validContent = content.filter(item => {
+                if (item.text !== undefined) {
+                    return item.text !== null && item.text !== '';
+                }
+                return true; // Keep non-text items (like images)
+            });
+
+            if (validContent.length > 0) {
                 converseMessages.push({
                     role: msg.role,
-                    content: content
+                    content: validContent
                 });
             }
         }
@@ -288,59 +296,8 @@ function buildInvokePrompt(message_cleaned, awsModel) {
     }
 }
 
-// Apply parameter restrictions for models that have them
-function applyParameterRestrictions(params, awsModel) {
-    if (!awsModel.parameter_restrictions) {
-        return params;
-    }
-
-    const restrictions = awsModel.parameter_restrictions;
-    
-    // Handle mutually exclusive parameters
-    if (restrictions.mutually_exclusive) {
-        for (const exclusiveGroup of restrictions.mutually_exclusive) {
-            // Check for both top_p and topP variants
-            const presentParams = exclusiveGroup.filter(param => {
-                if (param === 'top_p') {
-                    return (params['top_p'] !== undefined && params['top_p'] !== null) ||
-                           (params['topP'] !== undefined && params['topP'] !== null);
-                }
-                return params[param] !== undefined && params[param] !== null;
-            });
-            
-            if (presentParams.length > 1) {
-                // Keep the first parameter and remove others
-                // For temperature/top_p, prioritize temperature as it's more commonly used
-                const priorityOrder = ['temperature', 'top_p'];
-                const sortedParams = presentParams.sort((a, b) => {
-                    const aIndex = priorityOrder.indexOf(a);
-                    const bIndex = priorityOrder.indexOf(b);
-                    if (aIndex !== -1 && bIndex !== -1) return aIndex - bIndex;
-                    if (aIndex !== -1) return -1;
-                    if (bIndex !== -1) return 1;
-                    return 0;
-                });
-                
-                // Keep the first (highest priority) parameter, remove others
-                for (let i = 1; i < sortedParams.length; i++) {
-                    const paramToRemove = sortedParams[i];
-                    if (paramToRemove === 'top_p') {
-                        // Remove both variants
-                        delete params['top_p'];
-                        delete params['topP'];
-                    } else {
-                        delete params[paramToRemove];
-                    }
-                }
-            }
-        }
-    }
-    
-    return params;
-}
-
 // Build request object for Invoke API (model-specific)
-function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p, stop_sequences, stop, system_message) {
+function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, stop_sequences, stop, system_message) {
     if (awsModel.messages_api) {
         // Check if this is a Nova model (has schemaVersion in special_request_schema)
         if (awsModel.special_request_schema?.schemaVersion === "messages-v1") {
@@ -384,15 +341,11 @@ function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p
             let inferenceConfig = {
                 [awsModel.max_tokens_param_name]: max_gen_tokens,
                 temperature: temperature,
-                topP: top_p,
                 ...(awsModel.stop_sequences_param_name && stopSequencesValue && {
                     [awsModel.stop_sequences_param_name]: Array.isArray(stopSequencesValue) ? stopSequencesValue : [stopSequencesValue]
                 })
             };
-            
-            // Apply parameter restrictions
-            inferenceConfig = applyParameterRestrictions(inferenceConfig, awsModel);
-            
+
             const novaRequest = {
                 ...awsModel.special_request_schema,
                 messages: novaMessages,
@@ -415,16 +368,12 @@ function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p
                 ...(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.stop_sequences_param_name && stopSequencesValue && {
                     [awsModel.stop_sequences_param_name]: Array.isArray(stopSequencesValue) ? stopSequencesValue : [stopSequencesValue]
                 }),
                 ...awsModel.special_request_schema
             };
-            
-            // Apply parameter restrictions
-            request = applyParameterRestrictions(request, awsModel);
-            
+
             return request;
         }
     } else {
@@ -433,17 +382,16 @@ function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p
             prompt: typeof prompt === 'string' ? prompt : {
                 messages: prompt.map(msg => ({
                     role: msg.role,
-                    content: Array.isArray(msg.content) ? 
-                        msg.content.map(item => 
+                    content: Array.isArray(msg.content) ?
+                        msg.content.map(item =>
                             item.type === 'text' ? item.text : item
-                        ).join('\n') : 
+                        ).join('\n') :
                         msg.content
                 }))
             },
             // Optional inference parameters:
             [awsModel.max_tokens_param_name]: max_gen_tokens,
             temperature: temperature,
-            top_p: top_p,
             ...(() => {
                 const stopSequencesValue = stop_sequences || stop;
                 return awsModel.stop_sequences_param_name && stopSequencesValue ? {
@@ -452,10 +400,7 @@ function buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p
             })(),
             ...awsModel.special_request_schema
         };
-        
-        // Apply parameter restrictions
-        request = applyParameterRestrictions(request, awsModel);
-        
+
         return request;
     }
 }
@@ -555,10 +500,15 @@ async function* executeInvokeAPI(client, request, awsModelId, shouldStream, awsM
 
 export async function* bedrockWrapper(awsCreds, openaiChatCompletionsCreateObject, { logging = false, useConverseAPI = false } = {} ) {
     const { region, accessKeyId, secretAccessKey } = awsCreds;
-    let { messages, model, max_tokens, stream, temperature, top_p, include_thinking_data, stop, stop_sequences } = openaiChatCompletionsCreateObject;
+    let { messages, model, max_tokens, stream, temperature, include_thinking_data, stop, stop_sequences } = openaiChatCompletionsCreateObject;
 
     let {awsModelId, awsModel} = findAwsModelWithId(model);
-    
+
+    // Force Converse API for models that only support it
+    if (awsModel.converse_api_only) {
+        useConverseAPI = true;
+    }
+
     // Create a Bedrock Runtime client 
     const client = new BedrockRuntimeClient({
         region: region,
@@ -585,19 +535,14 @@ export async function* bedrockWrapper(awsCreds, openaiChatCompletionsCreateObjec
         // Build inference configuration (handle thinking mode for Claude models)
         let inferenceConfig = {
             maxTokens: max_gen_tokens,
-            temperature: temperature,
-            ...(top_p !== undefined && { topP: top_p })
+            temperature: temperature
         };
-        
-        // Apply parameter restrictions for Converse API
-        inferenceConfig = applyParameterRestrictions(inferenceConfig, awsModel);
-        
+
         // Handle thinking mode for Claude models
         let budget_tokens;
         if (awsModel.special_request_schema?.thinking?.type === "enabled") {
             // Apply thinking mode constraints for Converse API
             inferenceConfig.temperature = 1; // temperature must be 1 for thinking
-            delete inferenceConfig.topP; // top_p must be unset for thinking
             
             // Calculate thinking budget configuration
             budget_tokens = awsModel.special_request_schema?.thinking?.budget_tokens;
@@ -762,8 +707,6 @@ export async function* bedrockWrapper(awsCreds, openaiChatCompletionsCreateObjec
     if (awsModel.special_request_schema?.thinking?.type === "enabled") {
         // temperature may only be set to 1 when thinking is enabled
         temperature = 1;
-        // top_p must be unset when thinking is enabled
-        top_p = undefined;
         // budget_tokens can not be greater than 80% of max_gen_tokens
         let budget_tokens = awsModel.special_request_schema?.thinking?.budget_tokens;
         if (budget_tokens > (max_gen_tokens * 0.8)) {
@@ -784,7 +727,7 @@ export async function* bedrockWrapper(awsCreds, openaiChatCompletionsCreateObjec
     }
     
     // Build request for Invoke API (complex, model-specific)
-    const request = buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, top_p, stop_sequences, stop, system_message);
+    const request = buildInvokeRequest(prompt, awsModel, max_gen_tokens, temperature, stop_sequences, stop, system_message);
     
     if (logging) {
         console.log("\nFinal request:", JSON.stringify(request, null, 2));

package/example-converse-api.js

@@ -32,7 +32,6 @@ async function main() {
         max_tokens: 500,
         stream: true, // Can be true or false
         temperature: 0.7,
-        top_p: 0.9,
         stop: ["END", "STOP"] // Optional stop sequences
     };
 
@@ -80,8 +79,7 @@ async function main() {
         model: "Claude-3-Haiku",
         max_tokens: 50,
         stream: false,
-        temperature: 0.1,
-        top_p: 0.9
+        temperature: 0.1
     };
 
     // Test with Invoke API

package/interactive-example.js

@@ -15,7 +15,6 @@ const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
 const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
 const LLM_MAX_GEN_TOKENS = parseInt(process.env.LLM_MAX_GEN_TOKENS);
 const LLM_TEMPERATURE = parseFloat(process.env.LLM_TEMPERATURE);
-const LLM_TOP_P = parseFloat(process.env.LLM_TOP_P);
 
 // --------------------------------------------
 // -- import functions from bedrock-wrapper   --
@@ -129,7 +128,6 @@ const openaiChatCompletionsCreateObject = {
     "max_tokens": LLM_MAX_GEN_TOKENS,
     "stream": shouldStream,
     "temperature": LLM_TEMPERATURE,
-    "top_p": LLM_TOP_P,
     "include_thinking_data": true,
 };
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "bedrock-wrapper",
-  "version": "2.6.1",
+  "version": "2.7.0",
   "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": {
@@ -41,9 +41,9 @@
   "author": "",
   "license": "ISC",
   "dependencies": {
-    "@aws-sdk/client-bedrock-runtime": "^3.899.0",
+    "@aws-sdk/client-bedrock-runtime": "^3.934.0",
     "dotenv": "^17.2.3",
-    "sharp": "^0.34.4"
+    "sharp": "^0.34.5"
   },
   "devDependencies": {
     "chalk": "^5.6.2"

package/test-converse-api.js

@@ -16,7 +16,6 @@ const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
 const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
 const LLM_MAX_GEN_TOKENS = parseInt(process.env.LLM_MAX_GEN_TOKENS);
 const LLM_TEMPERATURE = parseFloat(process.env.LLM_TEMPERATURE);
-const LLM_TOP_P = parseFloat(process.env.LLM_TOP_P);
 
 // --------------------------------------------
 // -- import functions from bedrock-wrapper   --
@@ -60,7 +59,6 @@ async function testModelWithAPI(model, awsCreds, testMessage, isStreaming, useCo
         max_tokens: LLM_MAX_GEN_TOKENS,
         stream: isStreaming,
         temperature: LLM_TEMPERATURE,
-        top_p: LLM_TOP_P,
     };
 
     let completeResponse = "";
@@ -152,7 +150,6 @@ async function testStopSequences(model, awsCreds) {
         max_tokens: 100,
         stream: false,
         temperature: 0.1,
-        top_p: 0.9,
         stop: ["6"] // Stop at "6"
     };
 
@@ -208,7 +205,6 @@ async function testSystemPrompt(model, awsCreds) {
         max_tokens: 100,
         stream: false,
         temperature: 0.7,
-        top_p: 0.9,
     };
 
     try {

package/test-deepseek-quick.js

@@ -0,0 +1,66 @@
+// Quick test for DeepSeek models using interactive example logic
+import dotenv from 'dotenv';
+dotenv.config();
+
+import { bedrockWrapper } from "./bedrock-wrapper.js";
+
+const AWS_REGION = process.env.AWS_REGION;
+const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
+const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
+
+const awsCreds = {
+    region: AWS_REGION,
+    accessKeyId: AWS_ACCESS_KEY_ID,
+    secretAccessKey: AWS_SECRET_ACCESS_KEY,
+};
+
+async function quickTest(modelName, useConverseAPI) {
+    const apiType = useConverseAPI ? "Converse API" : "Invoke API";
+    console.log(`\nTesting ${modelName} with ${apiType}...`);
+
+    const messages = [
+        { role: "user", content: "What is 2+2? Answer with just the number." }
+    ];
+
+    const requestObject = {
+        messages,
+        model: modelName,
+        max_tokens: 50,
+        stream: false,
+        temperature: 0.1,
+        include_thinking_data: true,
+    };
+
+    try {
+        const response = await bedrockWrapper(awsCreds, requestObject, { logging: false, useConverseAPI });
+        let completeResponse = "";
+        for await (const data of response) {
+            completeResponse += data;
+        }
+        console.log(`✓ SUCCESS: ${completeResponse.trim().substring(0, 100)}...`);
+        return true;
+    } catch (error) {
+        console.log(`✗ FAILED: ${error.message}`);
+        return false;
+    }
+}
+
+async function main() {
+    console.log("\n" + "=".repeat(60));
+    console.log("QUICK DEEPSEEK MODEL TEST");
+    console.log("=".repeat(60));
+
+    const models = ["DeepSeek-R1", "DeepSeek-V3.1"];
+
+    for (const model of models) {
+        console.log(`\n--- Testing ${model} ---`);
+        await quickTest(model, false); // Invoke API
+        await quickTest(model, true);  // Converse API
+    }
+
+    console.log("\n" + "=".repeat(60));
+    console.log("Tests completed!");
+    console.log("=".repeat(60) + "\n");
+}
+
+main().catch(console.error);

package/test-deepseek.js

@@ -0,0 +1,64 @@
+// Simple test for DeepSeek models
+import dotenv from 'dotenv';
+dotenv.config();
+
+import { bedrockWrapper } from "./bedrock-wrapper.js";
+
+const AWS_REGION = process.env.AWS_REGION;
+const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
+const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
+
+const awsCreds = {
+    region: AWS_REGION,
+    accessKeyId: AWS_ACCESS_KEY_ID,
+    secretAccessKey: AWS_SECRET_ACCESS_KEY,
+};
+
+async function testDeepSeekModel(modelName) {
+    console.log(`\nTesting ${modelName}...`);
+
+    const messages = [{ role: "user", content: "What is 2+2? Respond with just the number." }];
+    const requestObject = {
+        messages,
+        model: modelName,
+        max_tokens: 100,
+        stream: false,
+        temperature: 0.1,
+    };
+
+    try {
+        const response = await bedrockWrapper(awsCreds, requestObject, { logging: true, useConverseAPI: false });
+        let completeResponse = "";
+        for await (const data of response) {
+            completeResponse += data;
+        }
+        console.log(`✓ ${modelName} (Invoke API): ${completeResponse.trim()}`);
+    } catch (error) {
+        console.log(`✗ ${modelName} (Invoke API): ${error.message}`);
+    }
+
+    // Test with Converse API
+    try {
+        const response = await bedrockWrapper(awsCreds, requestObject, { logging: true, useConverseAPI: true });
+        let completeResponse = "";
+        for await (const data of response) {
+            completeResponse += data;
+        }
+        console.log(`✓ ${modelName} (Converse API): ${completeResponse.trim()}`);
+    } catch (error) {
+        console.log(`✗ ${modelName} (Converse API): ${error.message}`);
+    }
+}
+
+async function main() {
+    console.log("Testing DeepSeek Models\n");
+    console.log("=".repeat(50));
+
+    await testDeepSeekModel("DeepSeek-R1");
+    await testDeepSeekModel("DeepSeek-V3.1");
+
+    console.log("\n" + "=".repeat(50));
+    console.log("DeepSeek tests completed!");
+}
+
+main().catch(console.error);

package/test-models.js

@@ -16,7 +16,6 @@ const AWS_ACCESS_KEY_ID = process.env.AWS_ACCESS_KEY_ID;
 const AWS_SECRET_ACCESS_KEY = process.env.AWS_SECRET_ACCESS_KEY;
 const LLM_MAX_GEN_TOKENS = parseInt(process.env.LLM_MAX_GEN_TOKENS);
 const LLM_TEMPERATURE = parseFloat(process.env.LLM_TEMPERATURE);
-const LLM_TOP_P = parseFloat(process.env.LLM_TOP_P);
 
 // --------------------------------------------
 // -- import functions from bedrock-wrapper   --
@@ -59,7 +58,6 @@ async function testModel(model, awsCreds, testMessage, isStreaming, useConverseA
         max_tokens: LLM_MAX_GEN_TOKENS,
         stream: isStreaming,
         temperature: LLM_TEMPERATURE,
-        top_p: LLM_TOP_P,
         include_thinking_data: true,
     };
 

package/test-stop-sequences.js

@@ -57,7 +57,6 @@ async function testStopSequence(model, awsCreds, testCase, isStreaming, useConve
         max_tokens: 200,
         stream: isStreaming,
         temperature: 0.1,
-        top_p: 0.9,
         stop_sequences: testCase.stopSequences
     };
 

package/logs/e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3/notification.json

@@ -1,72 +0,0 @@
-[
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude needs your permission to use context7 - resolve-library-id (MCP)"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude needs your permission to use context7 - get-library-docs (MCP)"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude needs your permission to use "
-  },
-  {
-    "session_id": "e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3",
-    "transcript_path": "C:\\Users\\Justin.Parker\\.claude\\projects\\C--git-bedrock-wrapper\\e4cf59ef-9d22-45bf-9c6c-53e3cb9efda3.jsonl",
-    "cwd": "C:\\git\\bedrock-wrapper",
-    "hook_event_name": "Notification",
-    "message": "Claude is waiting for your input"
-  }
-]