@inference-gateway/sdk 0.5.1 → 0.6.1

package/CHANGELOG.md

@@ -2,6 +2,18 @@
 
 All notable changes to this project will be documented in this file.
 
+## [0.6.1](https://github.com/inference-gateway/typescript-sdk/compare/v0.6.0...v0.6.1) (2025-04-28)
+
+### ♻️ Improvements
+
+*  Remove redundant request option ([#11](https://github.com/inference-gateway/typescript-sdk/issues/11)) ([82e34e2](https://github.com/inference-gateway/typescript-sdk/commit/82e34e2ee9782fd224945bff1bd4daf2859a4f79))
+
+## [0.6.0](https://github.com/inference-gateway/typescript-sdk/compare/v0.5.1...v0.6.0) (2025-04-28)
+
+### ✨ Features
+
+* Add usage metrics handling to streaming chat completions and update tests ([#10](https://github.com/inference-gateway/typescript-sdk/issues/10)) ([576ff71](https://github.com/inference-gateway/typescript-sdk/commit/576ff711140c9e357bea4ba572e92027297c428b))
+
 ## [0.5.1](https://github.com/inference-gateway/typescript-sdk/compare/v0.5.0...v0.5.1) (2025-04-27)
 
 ### 🐛 Bug Fixes

package/README.md

@@ -127,6 +127,7 @@ try {
       onOpen: () => console.log('Stream opened'),
       onContent: (content) => process.stdout.write(content),
       onChunk: (chunk) => console.log('Received chunk:', chunk.id),
+      onUsageMetrics: (metrics) => console.log('Usage metrics:', metrics),
       onFinish: () => console.log('\nStream completed'),
       onError: (error) => console.error('Stream error:', error),
     },

package/dist/src/client.d.ts

@@ -1,10 +1,11 @@
-import type { Provider, SchemaChatCompletionMessageToolCall, SchemaCreateChatCompletionRequest, SchemaCreateChatCompletionResponse, SchemaCreateChatCompletionStreamResponse, SchemaError, SchemaListModelsResponse } from './types/generated';
+import type { Provider, SchemaChatCompletionMessageToolCall, SchemaCompletionUsage, SchemaCreateChatCompletionRequest, SchemaCreateChatCompletionResponse, SchemaCreateChatCompletionStreamResponse, SchemaError, SchemaListModelsResponse } from './types/generated';
 interface ChatCompletionStreamCallbacks {
     onOpen?: () => void;
     onChunk?: (chunk: SchemaCreateChatCompletionStreamResponse) => void;
     onReasoning?: (reasoningContent: string) => void;
     onContent?: (content: string) => void;
     onTool?: (toolCall: SchemaChatCompletionMessageToolCall) => void;
+    onUsageMetrics?: (usage: SchemaCompletionUsage) => void;
     onFinish?: (response: SchemaCreateChatCompletionStreamResponse | null) => void;
     onError?: (error: SchemaError) => void;
 }
@@ -39,11 +40,16 @@ export declare class InferenceGatewayClient {
     /**
      * Creates a chat completion.
      */
-    createChatCompletion(request: SchemaCreateChatCompletionRequest, provider?: Provider): Promise<SchemaCreateChatCompletionResponse>;
+    createChatCompletion(request: Omit<SchemaCreateChatCompletionRequest, 'stream'>, provider?: Provider): Promise<SchemaCreateChatCompletionResponse>;
     /**
      * Creates a streaming chat completion.
+     * This method always sets stream=true internally, so there's no need to specify it in the request.
+     *
+     * @param request - Chat completion request (must include at least model and messages)
+     * @param callbacks - Callbacks for handling streaming events
+     * @param provider - Optional provider to use for this request
      */
-    streamChatCompletion(request: SchemaCreateChatCompletionRequest, callbacks: ChatCompletionStreamCallbacks, provider?: Provider): Promise<void>;
+    streamChatCompletion(request: Omit<SchemaCreateChatCompletionRequest, 'stream' | 'stream_options'>, callbacks: ChatCompletionStreamCallbacks, provider?: Provider): Promise<void>;
     /**
      * Proxy a request to a specific provider.
      */

package/dist/src/client.js

@@ -87,11 +87,16 @@ class InferenceGatewayClient {
         }
         return this.request('/chat/completions', {
             method: 'POST',
-            body: JSON.stringify(request),
+            body: JSON.stringify({ ...request, stream: false }),
         }, query);
     }
     /**
      * Creates a streaming chat completion.
+     * This method always sets stream=true internally, so there's no need to specify it in the request.
+     *
+     * @param request - Chat completion request (must include at least model and messages)
+     * @param callbacks - Callbacks for handling streaming events
+     * @param provider - Optional provider to use for this request
      */
     async streamChatCompletion(request, callbacks, provider) {
         const query = {};
@@ -120,6 +125,9 @@ class InferenceGatewayClient {
                 body: JSON.stringify({
                     ...request,
                     stream: true,
+                    stream_options: {
+                        include_usage: true,
+                    },
                 }),
                 signal: controller.signal,
             });
@@ -162,6 +170,9 @@ class InferenceGatewayClient {
                         try {
                             const chunk = JSON.parse(data);
                             callbacks.onChunk?.(chunk);
+                            if (chunk.usage && callbacks.onUsageMetrics) {
+                                callbacks.onUsageMetrics(chunk.usage);
+                            }
                             const reasoning_content = chunk.choices[0]?.delta?.reasoning_content;
                             if (reasoning_content !== undefined) {
                                 callbacks.onReasoning?.(reasoning_content);

package/dist/tests/client.test.js

@@ -89,7 +89,6 @@ describe('InferenceGatewayClient', () => {
                     { role: generated_1.MessageRole.system, content: 'You are a helpful assistant' },
                     { role: generated_1.MessageRole.user, content: 'Hello' },
                 ],
-                stream: false,
             };
             const mockResponse = {
                 id: 'chatcmpl-123',
@@ -120,14 +119,13 @@ describe('InferenceGatewayClient', () => {
             expect(result).toEqual(mockResponse);
             expect(mockFetch).toHaveBeenCalledWith('http://localhost:8080/v1/chat/completions', expect.objectContaining({
                 method: 'POST',
-                body: JSON.stringify(mockRequest),
+                body: JSON.stringify({ ...mockRequest, stream: false }),
             }));
         });
         it('should create a chat completion with a specific provider', async () => {
             const mockRequest = {
                 model: 'claude-3-opus-20240229',
                 messages: [{ role: generated_1.MessageRole.user, content: 'Hello' }],
-                stream: false,
             };
             const mockResponse = {
                 id: 'chatcmpl-456',
@@ -158,7 +156,7 @@ describe('InferenceGatewayClient', () => {
             expect(result).toEqual(mockResponse);
             expect(mockFetch).toHaveBeenCalledWith('http://localhost:8080/v1/chat/completions?provider=anthropic', expect.objectContaining({
                 method: 'POST',
-                body: JSON.stringify(mockRequest),
+                body: JSON.stringify({ ...mockRequest, stream: false }),
             }));
         });
     });
@@ -167,7 +165,6 @@ describe('InferenceGatewayClient', () => {
             const mockRequest = {
                 model: 'gpt-4o',
                 messages: [{ role: generated_1.MessageRole.user, content: 'Hello' }],
-                stream: true,
             };
             const mockStream = new web_1.TransformStream();
             const writer = mockStream.writable.getWriter();
@@ -201,6 +198,9 @@ describe('InferenceGatewayClient', () => {
                 body: JSON.stringify({
                     ...mockRequest,
                     stream: true,
+                    stream_options: {
+                        include_usage: true,
+                    },
                 }),
             }));
         });
@@ -208,7 +208,6 @@ describe('InferenceGatewayClient', () => {
             const mockRequest = {
                 model: 'gpt-4o',
                 messages: [{ role: generated_1.MessageRole.user, content: 'Hello' }],
-                stream: true,
             };
             const mockStream = new web_1.TransformStream();
             const writer = mockStream.writable.getWriter();
@@ -253,6 +252,9 @@ describe('InferenceGatewayClient', () => {
                 body: JSON.stringify({
                     ...mockRequest,
                     stream: true,
+                    stream_options: {
+                        include_usage: true,
+                    },
                 }),
             }));
         });
@@ -274,7 +276,6 @@ describe('InferenceGatewayClient', () => {
                         },
                     },
                 ],
-                stream: true,
             };
             const mockStream = new web_1.TransformStream();
             const writer = mockStream.writable.getWriter();
@@ -312,12 +313,21 @@ describe('InferenceGatewayClient', () => {
                 },
             });
             expect(callbacks.onFinish).toHaveBeenCalledTimes(1);
+            expect(mockFetch).toHaveBeenCalledWith('http://localhost:8080/v1/chat/completions', expect.objectContaining({
+                method: 'POST',
+                body: JSON.stringify({
+                    ...mockRequest,
+                    stream: true,
+                    stream_options: {
+                        include_usage: true,
+                    },
+                }),
+            }));
         });
         it('should handle errors in streaming chat completions', async () => {
             const mockRequest = {
                 model: 'gpt-4o',
                 messages: [{ role: generated_1.MessageRole.user, content: 'Hello' }],
-                stream: true,
             };
             mockFetch.mockResolvedValueOnce({
                 ok: false,
@@ -330,6 +340,57 @@ describe('InferenceGatewayClient', () => {
             await expect(client.streamChatCompletion(mockRequest, callbacks)).rejects.toThrow('Bad Request');
             expect(callbacks.onError).toHaveBeenCalledTimes(1);
         });
+        it('should handle streaming chat completions with usage metrics', async () => {
+            const mockRequest = {
+                model: 'gpt-4o',
+                messages: [{ role: generated_1.MessageRole.user, content: 'Hello' }],
+            };
+            const mockStream = new web_1.TransformStream();
+            const writer = mockStream.writable.getWriter();
+            const encoder = new node_util_1.TextEncoder();
+            mockFetch.mockResolvedValueOnce({
+                ok: true,
+                body: mockStream.readable,
+            });
+            const callbacks = {
+                onOpen: jest.fn(),
+                onChunk: jest.fn(),
+                onContent: jest.fn(),
+                onUsageMetrics: jest.fn(),
+                onFinish: jest.fn(),
+                onError: jest.fn(),
+            };
+            const streamPromise = client.streamChatCompletion(mockRequest, callbacks);
+            await writer.write(encoder.encode('data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}\n\n' +
+                'data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}]}\n\n' +
+                'data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","choices":[{"index":0,"delta":{"content":"!"},"finish_reason":null}]}\n\n' +
+                'data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}\n\n' +
+                'data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","choices":[],"usage":{"prompt_tokens":10,"completion_tokens":8,"total_tokens":18}}\n\n' +
+                'data: [DONE]\n\n'));
+            await writer.close();
+            await streamPromise;
+            expect(callbacks.onOpen).toHaveBeenCalledTimes(1);
+            expect(callbacks.onChunk).toHaveBeenCalledTimes(5);
+            expect(callbacks.onContent).toHaveBeenCalledWith('Hello');
+            expect(callbacks.onContent).toHaveBeenCalledWith('!');
+            expect(callbacks.onUsageMetrics).toHaveBeenCalledTimes(1);
+            expect(callbacks.onUsageMetrics).toHaveBeenCalledWith({
+                prompt_tokens: 10,
+                completion_tokens: 8,
+                total_tokens: 18,
+            });
+            expect(callbacks.onFinish).toHaveBeenCalledTimes(1);
+            expect(mockFetch).toHaveBeenCalledWith('http://localhost:8080/v1/chat/completions', expect.objectContaining({
+                method: 'POST',
+                body: JSON.stringify({
+                    ...mockRequest,
+                    stream: true,
+                    stream_options: {
+                        include_usage: true,
+                    },
+                }),
+            }));
+        });
     });
     describe('proxy', () => {
         it('should proxy requests to a specific provider', async () => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@inference-gateway/sdk",
-  "version": "0.5.1",
+  "version": "0.6.1",
   "description": "An SDK written in Typescript for the [Inference Gateway](https://github.com/inference-gateway/inference-gateway).",
   "main": "dist/src/index.js",
   "types": "dist/src/index.d.ts",