@inference-gateway/sdk 0.7.2 → 0.7.3

package/CHANGELOG.md

@@ -2,6 +2,16 @@
 
 All notable changes to this project will be documented in this file.
 
+## [0.7.3](https://github.com/inference-gateway/typescript-sdk/compare/v0.7.2...v0.7.3) (2025-06-01)
+
+### ♻️ Improvements
+
+* Enhance stream processing with abort signal support and increase default timeout ([#18](https://github.com/inference-gateway/typescript-sdk/issues/18)) ([3778138](https://github.com/inference-gateway/typescript-sdk/commit/377813851b6635ca7aafe2a5c9888b720736c9f5))
+
+### 🔧 Miscellaneous
+
+* Update MCP example README and remove unused example file ([99b34e7](https://github.com/inference-gateway/typescript-sdk/commit/99b34e70edf0c8aada1d0e0d0874481ea8381a79))
+
 ## [0.7.2](https://github.com/inference-gateway/typescript-sdk/compare/v0.7.1...v0.7.2) (2025-05-30)
 
 ### 📚 Documentation

package/dist/src/client.d.ts

@@ -54,8 +54,9 @@ export declare class InferenceGatewayClient {
      * @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
+     * @param abortSignal - Optional AbortSignal to cancel the request
      */
-    streamChatCompletion(request: Omit<SchemaCreateChatCompletionRequest, 'stream' | 'stream_options'>, callbacks: ChatCompletionStreamCallbacks, provider?: Provider): Promise<void>;
+    streamChatCompletion(request: Omit<SchemaCreateChatCompletionRequest, 'stream' | 'stream_options'>, callbacks: ChatCompletionStreamCallbacks, provider?: Provider, abortSignal?: AbortSignal): Promise<void>;
     /**
      * Initiates a streaming request to the chat completions endpoint
      */

package/dist/src/client.js

@@ -13,12 +13,15 @@ class StreamProcessor {
         this.callbacks = callbacks;
         this.clientProvidedTools = clientProvidedTools;
     }
-    async processStream(body) {
+    async processStream(body, abortSignal) {
         const reader = body.getReader();
         const decoder = new TextDecoder();
         let buffer = '';
         try {
             while (true) {
+                if (abortSignal?.aborted) {
+                    throw new Error('Stream processing was aborted');
+                }
                 const { done, value } = await reader.read();
                 if (done)
                     break;
@@ -34,6 +37,10 @@ class StreamProcessor {
             }
         }
         catch (error) {
+            if (abortSignal?.aborted || error.name === 'AbortError') {
+                console.log('Stream processing was cancelled');
+                return;
+            }
             const apiError = {
                 error: error.message || 'Unknown error',
             };
@@ -150,10 +157,10 @@ class StreamProcessor {
         }
     }
     finalizeIncompleteToolCalls() {
-        for (const [, toolCall] of this.incompleteToolCalls.entries()) {
+        this.incompleteToolCalls.forEach((toolCall) => {
             if (!toolCall.id || !toolCall.function.name) {
                 globalThis.console.warn('Incomplete tool call detected:', toolCall);
-                continue;
+                return;
             }
             const completedToolCall = {
                 id: toolCall.id,
@@ -171,13 +178,20 @@ class StreamProcessor {
                     this.callbacks.onMCPTool?.(completedToolCall);
                 }
                 catch (argError) {
-                    globalThis.console.warn(`Invalid MCP tool arguments for ${toolCall.function.name}:`, argError);
+                    const isIncompleteJSON = toolCall.function.arguments &&
+                        !toolCall.function.arguments.trim().endsWith('}');
+                    if (isIncompleteJSON) {
+                        globalThis.console.warn(`Incomplete MCP tool arguments for ${toolCall.function.name} (stream was likely interrupted):`, toolCall.function.arguments);
+                    }
+                    else {
+                        globalThis.console.warn(`Invalid MCP tool arguments for ${toolCall.function.name}:`, argError);
+                    }
                 }
             }
             else {
                 this.callbacks.onTool?.(completedToolCall);
             }
-        }
+        });
         this.incompleteToolCalls.clear();
     }
     isMCPTool(toolName) {
@@ -199,7 +213,7 @@ class InferenceGatewayClient {
         this.apiKey = options.apiKey;
         this.defaultHeaders = options.defaultHeaders || {};
         this.defaultQuery = options.defaultQuery || {};
-        this.timeout = options.timeout || 30000;
+        this.timeout = options.timeout || 60000; // Increased default timeout to 60 seconds
         this.fetchFn = options.fetch || globalThis.fetch;
     }
     /**
@@ -291,10 +305,11 @@ class InferenceGatewayClient {
      * @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
+     * @param abortSignal - Optional AbortSignal to cancel the request
      */
-    async streamChatCompletion(request, callbacks, provider) {
+    async streamChatCompletion(request, callbacks, provider, abortSignal) {
         try {
-            const response = await this.initiateStreamingRequest(request, provider);
+            const response = await this.initiateStreamingRequest(request, provider, abortSignal);
             if (!response.body) {
                 const error = {
                     error: 'Response body is not readable',
@@ -313,7 +328,7 @@ class InferenceGatewayClient {
                 }
             }
             const streamProcessor = new StreamProcessor(callbacks, clientProvidedTools);
-            await streamProcessor.processStream(response.body);
+            await streamProcessor.processStream(response.body, abortSignal);
         }
         catch (error) {
             const apiError = {
@@ -326,7 +341,7 @@ class InferenceGatewayClient {
     /**
      * Initiates a streaming request to the chat completions endpoint
      */
-    async initiateStreamingRequest(request, provider) {
+    async initiateStreamingRequest(request, provider, abortSignal) {
         const query = {};
         if (provider) {
             query.provider = provider;
@@ -345,6 +360,9 @@ class InferenceGatewayClient {
             headers.set('Authorization', `Bearer ${this.apiKey}`);
         }
         const controller = new AbortController();
+        const combinedSignal = abortSignal
+            ? AbortSignal.any([abortSignal, controller.signal])
+            : controller.signal;
         const timeoutId = globalThis.setTimeout(() => controller.abort(), this.timeout);
         try {
             const response = await this.fetchFn(url, {
@@ -357,7 +375,7 @@ class InferenceGatewayClient {
                         include_usage: true,
                     },
                 }),
-                signal: controller.signal,
+                signal: combinedSignal,
             });
             if (!response.ok) {
                 let errorMessage = `HTTP error! status: ${response.status}`;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@inference-gateway/sdk",
-  "version": "0.7.2",
+  "version": "0.7.3",
   "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",