feat: add support for OpenAI Responses API

Author: feiskyer

CHANGELOG.md

@@ -1,5 +1,9 @@
 # CHANGELOG
 
+## v4.9.3
+
+* Add support for OpenAI Responses API
+
 ## v4.9.2
 
 * Add prompt-based tool calling

README.md

@@ -187,6 +187,7 @@ For Azure OpenAI Service, apiBaseUrl should be set to format `https://[YOUR-ENDP
 [Github Copilot](https://github.com/features/copilot) is supported with built-in authentication (a popup will ask your permission when using Github Copilot models).
 
 **Supported Models:**
+
 - **OpenAI Models**: `gpt-3.5-turbo`, `gpt-4`, `gpt-4-turbo`, `gpt-4o`, `gpt-4o-mini`, `gpt-4.1`, `gpt-4.5`
 - **Reasoning Models**: `o1-ga`, `o3-mini`, `o3`, `o4-mini`
 - **Claude Models**: `claude-3.5-sonnet`, `claude-3.7-sonnet`, `claude-3.7-sonnet-thought`, `claude-sonnet-4`, `claude-opus-4`
@@ -253,6 +254,7 @@ The extension provides various commands accessible through the Command Palette (
 <summary> Context Menu Commands </summary>
 
 ### **Context Menu Commands** (Right-click on selected code)
+
 | Command | Keyboard Shortcut | Description |
 | ------- | ----------------- | ----------- |
 | **Generate Code** | `Ctrl+Shift+A` / `Cmd+Shift+A` | Generate code based on comments or requirements |
@@ -268,11 +270,11 @@ The extension provides various commands accessible through the Command Palette (
 
 </details>
 
-
 <details>
 <summary> General Commands </summary>
 
 ### **General Commands**
+
 | Command | Description |
 | ------- | ----------- |
 | `ChatGPT: Ask anything` | Open input box to ask any question |
@@ -284,14 +286,14 @@ The extension provides various commands accessible through the Command Palette (
 | `Add Current File to Chat Context` | Add the currently open file to chat context |
 | `ChatGPT: Open MCP Servers` | Manage Model Context Protocol servers |
 
-
 </details>
 
 <details>
 
 <summary> Prompt Management </summary>
 
 ### **Prompt Management**
+
 - Use `#` followed by prompt name to search and apply saved prompts
 - Use `@` to add files to your conversation context
 - Access the Prompt Manager through the sidebar for full prompt management
@@ -309,12 +311,14 @@ The extension supports the **Model Context Protocol (MCP)**, allowing you to ext
 ### **What is MCP?**
 
 MCP enables AI models to securely connect to external data sources and tools, providing:
+
 - **Custom Tools**: Integrate your own tools and APIs
 - **Data Sources**: Connect to databases, file systems, APIs, and more
 - **Secure Execution**: Sandboxed tool execution environment
 - **Multi-Step Workflows**: Agent-like behavior with tool chaining
 
 ### **MCP Server Types**
+
 The extension supports three types of MCP servers:
 
 | Type | Description | Use Case |
@@ -323,14 +327,14 @@ The extension supports three types of MCP servers:
 | **sse** | Server-Sent Events over HTTP | Web-based tools and APIs |
 | **streamable-http** | HTTP streaming communication | Real-time data sources |
 
-
 </details>
 
 <details>
 
 <summary> How to configure MCP? </summary>
 
 ### **MCP Configuration**
+
 1. **Access MCP Manager**: Use `ChatGPT: Open MCP Servers` command or click the MCP icon in the sidebar
 2. **Add MCP Server**: Configure your MCP servers with:
    - **Name**: Unique identifier for the server
@@ -343,6 +347,7 @@ The extension supports three types of MCP servers:
 ### **Example MCP Configurations**
 
 **File System Access (stdio):**
+
 ```json
 {
   "name": "filesystem",
@@ -354,6 +359,7 @@ The extension supports three types of MCP servers:
 ```
 
 **Web Search (sse):**
+
 ```json
 {
   "name": "web-search",
@@ -372,6 +378,7 @@ The extension supports three types of MCP servers:
 ### **Agent Mode**
 
 When MCP servers are enabled, the extension operates in **Agent Mode**:
+
 - **Max Steps**: Configure up to 15 tool execution steps
 - **Tool Chaining**: Automatic multi-step workflows
 - **Error Handling**: Robust error recovery and retry logic
@@ -386,6 +393,7 @@ When MCP servers are enabled, the extension operates in **Agent Mode**:
 <summary> Full list of configuration options </summary>
 
 ### **Core Configuration**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.gpt3.provider` | `Auto` | AI Provider: Auto, OpenAI, Azure, AzureAI, Anthropic, GitHubCopilot, Google, Mistral, xAI, Together, DeepSeek, Groq, Perplexity, OpenRouter, Ollama |
@@ -396,6 +404,7 @@ When MCP servers are enabled, the extension operates in **Agent Mode**:
 | `chatgpt.gpt3.organization` | | Organization ID (OpenAI only) |
 
 ### **Model Parameters**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.gpt3.maxTokens` | `0` (unlimited) | Maximum tokens to generate in completion |
@@ -404,6 +413,7 @@ When MCP servers are enabled, the extension operates in **Agent Mode**:
 | `chatgpt.systemPrompt` | | System prompt for the AI assistant |
 
 ### **DeepClaude (Reasoning + Chat) Configuration**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.gpt3.reasoning.provider` | `Auto` | Provider for reasoning model (Auto, OpenAI, Azure, AzureAI, Google, DeepSeek, Groq, OpenRouter, Ollama) |
@@ -413,17 +423,21 @@ When MCP servers are enabled, the extension operates in **Agent Mode**:
 | `chatgpt.gpt3.reasoning.organization` | | Organization ID for reasoning model (OpenAI only) |
 
 ### **Agent & MCP Configuration**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.gpt3.maxSteps` | `15` | Maximum steps for agent mode when using MCP servers |
 
 ### **Feature Toggles**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.gpt3.generateCode-enabled` | `true` | Enable code generation context menu |
 | `chatgpt.gpt3.searchGrounding.enabled` | `false` | Enable search grounding (Gemini models only) |
+| `chatgpt.gpt3.responsesAPI.enabled` | `false` | Enable OpenAI Responses API. Only available for OpenAI/AzureOpenAI models |
 
 ### **Prompt Prefixes & Context Menu**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.promptPrefix.addTests` | `Implement tests for the following code` | Prompt for generating unit tests |
@@ -445,6 +459,7 @@ When MCP servers are enabled, the extension operates in **Agent Mode**:
 | `chatgpt.promptPrefix.customPrompt2-enabled` | `false` | Enable second custom prompt in context menu |
 
 ### **User Interface**
+
 | Setting | Default | Description |
 | ------- | ------- | ----------- |
 | `chatgpt.response.showNotification` | `false` | Show notification when AI responds |

package.json

@@ -4,7 +4,7 @@
   "displayName": "ChatGPT Copilot",
   "icon": "images/ai-logo.png",
   "description": "An VS Code ChatGPT Copilot Extension",
-  "version": "4.9.2",
+  "version": "4.9.3",
   "aiKey": "",
   "repository": {
     "url": "https://github.com/feiskyer/chatgpt-copilot"
@@ -361,6 +361,7 @@
             "o1-preview",
             "o3",
             "o3-mini",
+            "o3-pro",
             "o4",
             "o4-mini",
             "claude-3-5-sonnet-20240620",
@@ -372,18 +373,13 @@
             "claude-3-7-sonnet-latest",
             "claude-opus-4-20250514",
             "claude-sonnet-4-20250514",
-            "gemini-1.5-flash",
-            "gemini-1.5-pro",
-            "gemini-exp-1206",
             "gemini-2.0-flash",
-            "gemini-2.0-flash-exp",
             "gemini-2.0-flash-lite",
-            "gemini-2.0-pro-exp-02-05",
-            "gemini-2.0-flash-thinking-exp-01-21",
-            "gemini-2.5-flash-preview-05-20",
-            "gemini-2.5-pro-exp-03-25",
+            "gemini-2.0-flash-preview-image-generation",
+            "gemini-2.5-flash",
+            "gemini-2.5-flash-lite-preview-06-17",
+            "gemini-2.5-pro",
             "gemini-2.5-pro-preview-05-06",
-            "learnlm-1.5-pro-experimental",
             "grok-2-1212",
             "grok-2-vision-1212",
             "grok-vision-beta",
@@ -455,6 +451,12 @@
           "description": "Enable search grounding for Gemini model. Only available for Google Gemini models.",
           "order": 11
         },
+        "chatgpt.gpt3.responsesAPI.enabled": {
+          "type": "boolean",
+          "default": false,
+          "description": "Enable OpenAI Responses API. Only available for OpenAI/AzureOpenAI models.",
+          "order": 12
+        },
         "chatgpt.promptPrefix.addTests": {
           "type": "string",
           "default": "Implement tests for the following code",
@@ -646,7 +648,7 @@
     "build": "yarn esbuild-base --sourcemap",
     "watch": "yarn esbuild-base --sourcemap --watch",
     "fmt": "prettier --write \"src/**/*.ts\" && yarn test",
-    "update": "yarn add -D npm-check-updates && ncu -u && yarn install",
+    "update": "yarn add -D npm-check-updates && ncu -u -t latest && yarn install",
     "test": "eslint src --ext ts && tsc --noEmit",
     "package": "yarn vsce package",
     "publish": "yarn vsce publish"
@@ -655,18 +657,18 @@
     "@types/glob": "^8.1.0",
     "@types/isomorphic-fetch": "^0.0.39",
     "@types/mocha": "^10.0.10",
-    "@types/node": "^22.15.29",
+    "@types/node": "^24.0.3",
     "@types/react": "^19",
     "@types/react-dom": "^19",
     "@types/uuid": "^10.0.0",
     "@types/vscode": "1.96.0",
     "@types/vscode-webview": "^1.57.5",
     "@vscode/test-electron": "^2.5.2",
-    "@vscode/vsce": "^3.4.2",
+    "@vscode/vsce": "^3.5.1-3",
     "esbuild": "^0.25.5",
-    "eslint": "^9.28.0",
-    "glob": "^11.0.2",
-    "mocha": "^11.5.0",
+    "eslint": "^9.29.0",
+    "glob": "^11.0.3",
+    "mocha": "^11.7.0",
     "npm-check-updates": "^18.0.1",
     "prettier": "^3.5.3",
     "rimraf": "^6.0.1",
@@ -678,7 +680,7 @@
     "@ai-sdk/anthropic": "^1.2.12",
     "@ai-sdk/azure": "^1.3.23",
     "@ai-sdk/deepseek": "^0.2.14",
-    "@ai-sdk/google": "^1.2.18",
+    "@ai-sdk/google": "^1.2.19",
     "@ai-sdk/groq": "^1.2.9",
     "@ai-sdk/mistral": "^1.2.8",
     "@ai-sdk/openai": "^1.3.22",
@@ -687,25 +689,25 @@
     "@ai-sdk/replicate": "^0.2.8",
     "@ai-sdk/togetherai": "^0.2.14",
     "@ai-sdk/xai": "^1.2.16",
-    "@modelcontextprotocol/sdk": "^1.12.1",
-    "@openrouter/ai-sdk-provider": "^0.7.0",
+    "@modelcontextprotocol/sdk": "^1.13.0",
+    "@openrouter/ai-sdk-provider": "^0.7.2",
     "@quail-ai/azure-ai-provider": "1.2.0",
     "@types/minimatch": "^5.1.2",
     "ai": "^4.3.16",
     "ajv": "^8.17.1",
-    "axios": "^1.9.0",
-    "cheerio": "^1.0.0",
+    "axios": "^1.10.0",
+    "cheerio": "^1.1.0",
     "delay": "^6.0.0",
     "eventsource-parser": "^3.0.2",
     "gpt3-tokenizer": "^1.1.5",
     "isomorphic-fetch": "^3.0.0",
-    "keyv": "^5.3.3",
-    "minimatch": "^10.0.1",
+    "keyv": "^5.3.4",
+    "minimatch": "^10.0.3",
     "ollama-ai-provider": "^1.2.0",
-    "openai": "^5.0.1",
+    "openai": "^5.5.1",
     "p-timeout": "^6.1.4",
     "punycode": "^2.3.1",
-    "puppeteer": "^24.9.0",
+    "puppeteer": "^24.10.2",
     "puppeteer-extra": "^3.3.6",
     "puppeteer-extra-plugin-stealth": "^2.11.2",
     "puppeteer-extra-plugin-user-data-dir": "^2.4.1",
@@ -717,7 +719,7 @@
     "strip-markdown": "^6.0.0",
     "uri-js": "^4.4.1",
     "uuid": "^11.1.0",
-    "zod": "^3.25.46"
+    "zod": "^3.25.67"
   },
   "resolutions": {
     "punycode": "2.3.1",

src/chatgpt-view-provider.ts

@@ -599,6 +599,9 @@ export default class ChatGptViewProvider implements vscode.WebviewViewProvider {
       const searchGrounding = configuration.get(
         "gpt3.searchGrounding.enabled",
       ) as boolean;
+      const enableResponsesAPI = configuration.get(
+        "gpt3.responsesAPI.enabled",
+      ) as boolean;
 
       let systemPrompt = configuration.get("systemPrompt") as string;
       if (this.systemPromptOverride != "") {
@@ -675,6 +678,7 @@ export default class ChatGptViewProvider implements vscode.WebviewViewProvider {
         organization,
         systemPrompt,
         searchGrounding,
+        enableResponsesAPI,
         isReasoning: false,
       });
       if (this.reasoningModel != "") {
@@ -697,6 +701,7 @@ export default class ChatGptViewProvider implements vscode.WebviewViewProvider {
           organization,
           systemPrompt: "",
           searchGrounding,
+          enableResponsesAPI,
           isReasoning: true,
         });
       }

src/deepclaude.ts

@@ -197,6 +197,13 @@ export async function reasoningChat(
         temperature: provider.modelConfig.temperature,
         // topP: provider.modelConfig.topP,
       }),
+      // ...(provider.provider === "Google" && provider.modelConfig.searchGrounding && {
+      //   providerOptions: {
+      //     google: {
+      //       useSearchGrounding: true,
+      //     },
+      //   },
+      // }),
     });
     for await (const part of result.fullStream) {
       // logger.appendLine(`INFO: deepclaude.model: ${provider.model} deepclaude.question: ${question} response: ${JSON.stringify(part, null, 2)}`);

src/llms.ts

@@ -76,7 +76,7 @@ export async function initGeminiModel(
   if (config.isReasoning) {
     const model = viewProvider.reasoningModel
       ? viewProvider.reasoningModel
-      : "gemini-1.5-flash-latest";
+      : "gemini-2.5-pro";
     viewProvider.apiReasoning = wrapLanguageModel({
       model: ai(model),
       middleware: extractReasoningMiddleware({ tagName: "think" }),
@@ -93,7 +93,7 @@ export async function initGeminiModel(
   } else {
     const model = viewProvider.model
       ? viewProvider.model
-      : "gemini-1.5-flash-latest";
+      : "gemini-2.5-pro";
     viewProvider.apiChat = ai(model);
     if (config.searchGrounding) {
       viewProvider.apiChat = ai(model, {

src/mcp.ts

@@ -133,12 +133,7 @@ export async function createToolSet(config: MCPServerConfig): Promise<ToolSet> {
         if (parameters.jsonSchema.additionalProperties == null) {
           parameters.jsonSchema.additionalProperties = false;
         }
-        // const parameters = jsonSchema(
-        //   Object.fromEntries(
-        //     Object.entries(t.inputSchema as JSONSchema7)
-        //       .filter(([key]) => key !== "additionalProperties" && key !== "$schema")
-        //   )
-        // );
+
         toolset.tools[toolName] = tool({
           description: t.description || toolName,
           parameters: parameters,