Upload folder using huggingface_hub

README.md

@@ -13,19 +13,24 @@ app_port: 3000
 
 # responses.js
 
-A lightweight Express.js server that implements OpenAI's Responses API, built on top of Chat Completions and powered by Hugging Face Inference Providers.
+A lightweight Express.js server that implements a translation layer between the two main LLM APIs currently available. Works with any Chat Completion API, whether it's a local LLM or the cloud provider of your choice.
+
+## 🎮 Live Demo
+
+[**Try responses.js right now, no installation needed!**](https://huggingface.co/spaces/Wauplin/responses.js)
 
 ## ✨ Features
 
 - **ResponsesAPI**: Partial implementation of [OpenAI's Responses API](https://platform.openai.com/docs/api-reference/responses), on top of Chat Completion API
-- **Inference Providers**: Powered by Hugging Face Inference Providers
+- **Provider Agnostic**: Works with any Chat Completion API (local or remote)
 - **Streaming Support**: Support for streamed responses
 - **Structured Output**: Support for structured data responses (e.g. jsonschema)
 - **Function Calling**: Tool and function calling capabilities
 - **Multi-modal Input**: Text and image input support
+- **Remote MCP**: Execute MCP tool calls remotely
 - **Demo UI**: Interactive web interface for testing
 
-Not implemented: remote function calling, MCP server, file upload, stateful API, etc.
+Not implemented: remote function calling, file upload, stateful API, etc.
 
 ## 🚀 Quick Start
 
@@ -82,8 +87,8 @@ pnpm run example function_streaming
 ### Important Notes
 
 - Server must be running (`pnpm dev`) on `http://localhost:3000`
-- `HF_TOKEN` environment variable set with your Hugging Face token
-- Tests use real inference providers and will incur costs
+- `API_KEY` environment variable set with your LLM provider's API key
+- Tests use real inference providers and may incur costs
 - Tests are not run in CI due to billing requirements
 
 ### Running Tests
@@ -112,7 +117,7 @@ Experience the API through our interactive web interface, adapted from the [open
 ```bash
 # Create demo/.env
 cat > demo/.env << EOF
-MODEL="CohereLabs/c4ai-command-a-03-2025"
+MODEL="moonshotai/Kimi-K2-Instruct:groq"
 OPENAI_BASE_URL=http://localhost:3000/v1
 OPENAI_API_KEY=${HF_TOKEN:-<your-huggingface-token>}
 EOF
@@ -200,4 +205,3 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 - Based on OpenAI's [Responses API specification](https://platform.openai.com/docs/api-reference/responses)
 - Built on top of [OpenAI's nodejs client](https://github.com/openai/openai-node)
 - Demo UI adapted from [openai-responses-starter-app](https://github.com/openai/openai-responses-starter-app)
-- Built on top of [Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers/index)

src/routes/landingPageHtml.ts

@@ -464,7 +464,7 @@ export function getLandingPageHtml(req: Request, res: Response): void {
   <main>
     <section class="hero">
       <h2>OpenAI-compatible Responses API</h2>
-      <p><b>responses.js</b> is an open-source, lightweight server implementing OpenAI's Responses API, built on top of Chat Completions and powered by Hugging Face Inference Providers.</p>
+      <p><b>responses.js</b> is an open-source, lightweight translation layer between the two main LLM APIs currently available. Works with any Chat Completion API, whether it's a local LLM or the cloud provider of your choice.</p>
       <div class="api-endpoint-box">
         <button class="copy-endpoint-btn" onclick="copyEndpointUrl(this)">Copy</button>
         <div><b>API Endpoint:</b></div>
@@ -479,13 +479,13 @@ export function getLandingPageHtml(req: Request, res: Response): void {
           <b>OpenAI-compatible</b><br>Stateless implementation of the <a href="https://platform.openai.com/docs/api-reference/responses" target="_blank">Responses API</a>
         </div>
         <div class="feature-card">
-          <b>Inference Providers</b><br>Powered by Hugging Face Inference Providers
+          <b>Provider Agnostic</b><br>Works with any Chat Completion API (local or remote).
         </div>
         <div class="feature-card">
-          <b>Multi-modal</b><br>Text and image input support
+          <b>Multi-modal, streaming, structured output</b><br>Supports text and image inputs, streaming output, JSON schema, and function calling.
         </div>
         <div class="feature-card">
-          <b>Streaming, & Structured Output</b><br>Supports streaming, JSON schema, and function calling
+          <b>Remote MCP</b><br>Server-side MCP tool execution.
         </div>
       </div>
     </section>

src/schemas.ts

@@ -193,7 +193,7 @@ export const createResponseParamsSchema = z.object({
 		.union([
 			z.enum(["auto", "none", "required"]),
 			z.object({
-				type: z.enum(["function"]),
+				type: z.literal("function"),
 				name: z.string(),
 			}),
 			// TODO: also hosted tool and MCP tool
@@ -206,7 +206,7 @@ export const createResponseParamsSchema = z.object({
 					name: z.string(),
 					parameters: z.record(z.any()),
 					strict: z.boolean().default(true),
-					type: z.enum(["function"]),
+					type: z.literal("function"),
 					description: z.string().optional(),
 				}),
 				mcpServerParamsSchema,