@layer-ai/core 2.0.14 → 2.0.16

package/dist/lib/provider-factory.d.ts

@@ -24,4 +24,11 @@ export declare function getProviderForModel(model: SupportedModel): Provider;
  * @param userId - Optional user ID for BYOK key resolution
  */
 export declare function callAdapter(request: LayerRequest, userId?: string): Promise<LayerResponse>;
+/**
+ * Calls the appropriate provider adapter for streaming responses.
+ * This is the streaming version of callAdapter.
+ * @param request - The Layer request to execute
+ * @param userId - Optional user ID for BYOK key resolution
+ */
+export declare function callAdapterStream(request: LayerRequest, userId?: string): AsyncIterable<LayerResponse>;
 //# sourceMappingURL=provider-factory.d.ts.map

package/dist/lib/provider-factory.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"provider-factory.d.ts","sourceRoot":"","sources":["../../src/lib/provider-factory.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAMH,OAAO,KAAK,EAAE,YAAY,EAAE,aAAa,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AAEjF,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAG7E,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,QAAQ,EAAE,CAAC;AAa9C;;;GAGG;AACH,wBAAgB,gBAAgB,CAAC,OAAO,EAAE,MAAM,GAAG,cAAc,CAmBhE;AAED;;GAEG;AACH,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,cAAc,GAAG,QAAQ,CAMnE;AAED;;;;;GAKG;AACH,wBAAsB,WAAW,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC,CAWhG"}
+{"version":3,"file":"provider-factory.d.ts","sourceRoot":"","sources":["../../src/lib/provider-factory.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAMH,OAAO,KAAK,EAAE,YAAY,EAAE,aAAa,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AAEjF,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAG7E,OAAO,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,QAAQ,EAAE,CAAC;AAa9C;;;GAGG;AACH,wBAAgB,gBAAgB,CAAC,OAAO,EAAE,MAAM,GAAG,cAAc,CAmBhE;AAED;;GAEG;AACH,wBAAgB,mBAAmB,CAAC,KAAK,EAAE,cAAc,GAAG,QAAQ,CAMnE;AAED;;;;;GAKG;AACH,wBAAsB,WAAW,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC,CAWhG;AAED;;;;;GAKG;AACH,wBAAuB,iBAAiB,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,aAAa,CAAC,CAgB7G"}

package/dist/lib/provider-factory.js

@@ -71,3 +71,22 @@ export async function callAdapter(request, userId) {
     const adapter = new AdapterClass();
     return await adapter.call(request, userId);
 }
+/**
+ * Calls the appropriate provider adapter for streaming responses.
+ * This is the streaming version of callAdapter.
+ * @param request - The Layer request to execute
+ * @param userId - Optional user ID for BYOK key resolution
+ */
+export async function* callAdapterStream(request, userId) {
+    const normalizedModel = normalizeModelId(request.model);
+    const provider = getProviderForModel(normalizedModel);
+    const AdapterClass = PROVIDER_ADAPTERS[provider];
+    if (!AdapterClass) {
+        throw new Error(`No adapter found for provider: ${provider}`);
+    }
+    const adapter = new AdapterClass();
+    if (!adapter.callStream) {
+        throw new Error(`Streaming not supported for provider: ${provider}`);
+    }
+    yield* adapter.callStream(request, userId);
+}

package/dist/routes/tests/test-chat-streaming.d.ts

@@ -0,0 +1,7 @@
+#!/usr/bin/env npx tsx
+/**
+ * Test streaming at the chat route level
+ * Tests routing strategies with streaming support
+ */
+export {};
+//# sourceMappingURL=test-chat-streaming.d.ts.map

package/dist/routes/tests/test-chat-streaming.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"test-chat-streaming.d.ts","sourceRoot":"","sources":["../../../src/routes/tests/test-chat-streaming.ts"],"names":[],"mappings":";AACA;;;GAGG"}

package/dist/routes/tests/test-chat-streaming.js

@@ -0,0 +1,213 @@
+#!/usr/bin/env npx tsx
+/**
+ * Test streaming at the chat route level
+ * Tests routing strategies with streaming support
+ */
+import { callAdapterStream } from '../../lib/provider-factory.js';
+console.log('='.repeat(80));
+console.log('STREAMING - CHAT ROUTE LEVEL TESTS');
+console.log('='.repeat(80));
+console.log('');
+// Simplified versions of the routing functions from chat.ts
+async function* executeWithFallbackStream(request, modelsToTry) {
+    let lastError = null;
+    for (const modelToTry of modelsToTry) {
+        try {
+            const modelRequest = { ...request, model: modelToTry };
+            yield* callAdapterStream(modelRequest);
+            return; // Success!
+        }
+        catch (error) {
+            lastError = error;
+            console.log(`  Model ${modelToTry} failed, trying next fallback...`);
+            continue;
+        }
+    }
+    throw lastError || new Error('All models failed during streaming');
+}
+async function* executeWithRoundRobinStream(gateConfig, request) {
+    let selectedModel = request.model;
+    if (gateConfig.fallbackModels?.length) {
+        const allModels = [gateConfig.model, ...gateConfig.fallbackModels];
+        const modelIndex = Math.floor(Math.random() * allModels.length);
+        selectedModel = allModels[modelIndex];
+        console.log(`  Selected model via round-robin: ${selectedModel}`);
+    }
+    const modelRequest = { ...request, model: selectedModel };
+    yield* callAdapterStream(modelRequest);
+}
+// Test 1: Single routing with streaming
+async function testSingleRouting() {
+    console.log('Test 1: Single Routing with Streaming');
+    console.log('-'.repeat(80));
+    const request = {
+        gateId: 'test-gate',
+        model: 'gpt-4o',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'Say "test passed" and nothing else.' }
+            ],
+            maxTokens: 10,
+            stream: true,
+        }
+    };
+    let chunkCount = 0;
+    let fullContent = '';
+    for await (const chunk of callAdapterStream(request)) {
+        chunkCount++;
+        if (chunk.content) {
+            fullContent += chunk.content;
+        }
+    }
+    console.log(`  Chunks received: ${chunkCount}`);
+    console.log(`  Content: ${fullContent.trim()}`);
+    console.log('  ✅ Single routing test passed\n');
+}
+// Test 2: Fallback routing with streaming
+async function testFallbackRouting() {
+    console.log('Test 2: Fallback Routing with Streaming');
+    console.log('-'.repeat(80));
+    const request = {
+        gateId: 'test-gate',
+        model: 'invalid-model-1', // This will fail
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'Say "fallback worked" and nothing else.' }
+            ],
+            maxTokens: 10,
+            stream: true,
+        }
+    };
+    const modelsToTry = [
+        'invalid-model-1', // Should fail
+        'invalid-model-2', // Should fail
+        'gpt-4o', // Should succeed
+    ];
+    let chunkCount = 0;
+    let fullContent = '';
+    let succeeded = false;
+    try {
+        for await (const chunk of executeWithFallbackStream(request, modelsToTry)) {
+            chunkCount++;
+            if (chunk.content) {
+                fullContent += chunk.content;
+            }
+        }
+        succeeded = true;
+    }
+    catch (error) {
+        console.error('  ❌ Fallback failed:', error instanceof Error ? error.message : error);
+    }
+    if (succeeded) {
+        console.log(`  Chunks received: ${chunkCount}`);
+        console.log(`  Content: ${fullContent.trim()}`);
+        console.log('  ✅ Fallback routing test passed\n');
+    }
+}
+// Test 3: Round-robin routing with streaming
+async function testRoundRobinRouting() {
+    console.log('Test 3: Round-Robin Routing with Streaming');
+    console.log('-'.repeat(80));
+    const gateConfig = {
+        id: 'test-gate-id',
+        name: 'Test Gate',
+        model: 'gpt-4o',
+        taskType: 'chat',
+        routingStrategy: 'round-robin',
+        fallbackModels: ['gpt-4o', 'gpt-3.5-turbo'],
+        userId: 'test-user',
+        createdAt: new Date(),
+        updatedAt: new Date(),
+    };
+    const request = {
+        gateId: 'test-gate',
+        model: 'gpt-4o',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'Say "round-robin worked" and nothing else.' }
+            ],
+            maxTokens: 10,
+            stream: true,
+        }
+    };
+    let chunkCount = 0;
+    let fullContent = '';
+    for await (const chunk of executeWithRoundRobinStream(gateConfig, request)) {
+        chunkCount++;
+        if (chunk.content) {
+            fullContent += chunk.content;
+        }
+    }
+    console.log(`  Chunks received: ${chunkCount}`);
+    console.log(`  Content: ${fullContent.trim()}`);
+    console.log('  ✅ Round-robin routing test passed\n');
+}
+// Test 4: Streaming with tool calls
+async function testStreamingWithTools() {
+    console.log('Test 4: Streaming with Tool Calls');
+    console.log('-'.repeat(80));
+    const request = {
+        gateId: 'test-gate',
+        model: 'gpt-4o',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'What is the weather in Paris?' }
+            ],
+            tools: [
+                {
+                    type: 'function',
+                    function: {
+                        name: 'get_weather',
+                        description: 'Get weather for a location',
+                        parameters: {
+                            type: 'object',
+                            properties: {
+                                location: { type: 'string' },
+                            },
+                            required: ['location'],
+                        },
+                    },
+                },
+            ],
+            stream: true,
+        }
+    };
+    let toolCallsFound = false;
+    let finishReason = null;
+    for await (const chunk of callAdapterStream(request)) {
+        if (chunk.toolCalls && chunk.toolCalls.length > 0) {
+            toolCallsFound = true;
+        }
+        if (chunk.finishReason) {
+            finishReason = chunk.finishReason;
+        }
+    }
+    console.log(`  Tool calls found: ${toolCallsFound}`);
+    console.log(`  Finish reason: ${finishReason}`);
+    if (toolCallsFound && finishReason === 'tool_call') {
+        console.log('  ✅ Tool calls streaming test passed\n');
+    }
+    else {
+        console.log('  ⚠️  Tool calls may not have been invoked (model chose not to use tools)\n');
+    }
+}
+// Run all tests
+(async () => {
+    try {
+        await testSingleRouting();
+        await testFallbackRouting();
+        await testRoundRobinRouting();
+        await testStreamingWithTools();
+        console.log('='.repeat(80));
+        console.log('✅ ALL STREAMING ROUTE TESTS PASSED');
+        console.log('='.repeat(80));
+    }
+    catch (error) {
+        console.error('❌ Test failed:', error);
+        process.exit(1);
+    }
+})();

package/dist/routes/v3/chat.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"chat.d.ts","sourceRoot":"","sources":["../../../src/routes/v3/chat.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,MAAM,IAAI,UAAU,EAAE,MAAM,SAAS,CAAC;AAIpD,OAAO,KAAK,EAAE,YAAY,EAAiB,IAAI,EAA+C,MAAM,eAAe,CAAC;AAGpH,QAAA,MAAM,MAAM,EAAE,UAAqB,CAAC;AAiBpC,wBAAgB,mBAAmB,CACjC,UAAU,EAAE,IAAI,EAChB,OAAO,EAAE,YAAY,GACpB,YAAY,CAkFd;AA6MD,eAAe,MAAM,CAAC"}
+{"version":3,"file":"chat.d.ts","sourceRoot":"","sources":["../../../src/routes/v3/chat.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,MAAM,IAAI,UAAU,EAAE,MAAM,SAAS,CAAC;AAIpD,OAAO,KAAK,EAAE,YAAY,EAAiB,IAAI,EAA+C,MAAM,eAAe,CAAC;AAGpH,QAAA,MAAM,MAAM,EAAE,UAAqB,CAAC;AAiBpC,wBAAgB,mBAAmB,CACjC,UAAU,EAAE,IAAI,EAChB,OAAO,EAAE,YAAY,GACpB,YAAY,CAiFd;AAgWD,eAAe,MAAM,CAAC"}

package/dist/routes/v3/chat.js

@@ -1,7 +1,7 @@
 import { Router } from 'express';
 import { db } from '../../lib/db/postgres.js';
 import { authenticate } from '../../middleware/auth.js';
-import { callAdapter, normalizeModelId, getProviderForModel, PROVIDER } from '../../lib/provider-factory.js';
+import { callAdapter, callAdapterStream, normalizeModelId, getProviderForModel, PROVIDER } from '../../lib/provider-factory.js';
 import { OverrideField } from '@layer-ai/sdk';
 const router = Router();
 // MARK:- Helper Functions
@@ -22,7 +22,6 @@ export function resolveFinalRequest(gateConfig, request) {
             finalModel = gateConfig.model;
         }
     }
-    // Since this is v3/chat endpoint, we know the data is ChatRequest
     const chatData = { ...request.data };
     if (!chatData.systemPrompt && gateConfig.systemPrompt) {
         chatData.systemPrompt = gateConfig.systemPrompt;
@@ -141,6 +140,48 @@ async function executeWithRouting(gateConfig, request, userId) {
             return { result, modelUsed: request.model };
     }
 }
+async function* executeWithFallbackStream(request, modelsToTry, userId) {
+    let lastError = null;
+    for (const modelToTry of modelsToTry) {
+        try {
+            const modelRequest = { ...request, model: modelToTry };
+            yield* callAdapterStream(modelRequest, userId);
+            return;
+        }
+        catch (error) {
+            lastError = error;
+            console.log(`Model ${modelToTry} failed during streaming, trying next fallback...`, error instanceof Error ? error.message : error);
+            continue;
+        }
+    }
+    throw lastError || new Error('All models failed during streaming');
+}
+async function* executeWithRoundRobinStream(gateConfig, request, userId) {
+    if (!gateConfig.fallbackModels?.length) {
+        yield* callAdapterStream(request, userId);
+        return;
+    }
+    const allModels = [gateConfig.model, ...gateConfig.fallbackModels];
+    const modelIndex = Math.floor(Math.random() * allModels.length);
+    const selectedModel = allModels[modelIndex];
+    const modelRequest = { ...request, model: selectedModel };
+    yield* callAdapterStream(modelRequest, userId);
+}
+async function* executeWithRoutingStream(gateConfig, request, userId) {
+    const modelsToTry = getModelsToTry(gateConfig, request.model);
+    switch (gateConfig.routingStrategy) {
+        case 'fallback':
+            yield* executeWithFallbackStream(request, modelsToTry, userId);
+            break;
+        case 'round-robin':
+            yield* executeWithRoundRobinStream(gateConfig, request, userId);
+            break;
+        case 'single':
+        default:
+            yield* callAdapterStream(request, userId);
+            break;
+    }
+}
 // MARK:- Route Handler
 router.post('/', authenticate, async (req, res) => {
     const startTime = Date.now();
@@ -185,9 +226,96 @@ router.post('/', authenticate, async (req, res) => {
             metadata: rawRequest.metadata
         };
         const finalRequest = resolveFinalRequest(gateConfig, request);
+        const isStreaming = finalRequest.data && 'stream' in finalRequest.data && finalRequest.data.stream === true;
+        if (isStreaming) {
+            res.setHeader('Content-Type', 'text/event-stream');
+            res.setHeader('Cache-Control', 'no-cache');
+            res.setHeader('Connection', 'keep-alive');
+            res.setHeader('X-Accel-Buffering', 'no');
+            let promptTokens = 0;
+            let completionTokens = 0;
+            let totalCost = 0;
+            let modelUsed = finalRequest.model;
+            try {
+                for await (const chunk of executeWithRoutingStream(gateConfig, finalRequest, userId)) {
+                    if (chunk.usage) {
+                        promptTokens = chunk.usage.promptTokens || 0;
+                        completionTokens = chunk.usage.completionTokens || 0;
+                    }
+                    if (chunk.cost) {
+                        totalCost = chunk.cost;
+                    }
+                    if (chunk.model) {
+                        modelUsed = chunk.model;
+                    }
+                    res.write(`data: ${JSON.stringify(chunk)}\n\n`);
+                }
+                res.write(`data: [DONE]\n\n`);
+                res.end();
+                const latencyMs = Date.now() - startTime;
+                db.logRequest({
+                    userId,
+                    gateId: gateConfig.id,
+                    gateName: gateConfig.name,
+                    modelRequested: request.model || gateConfig.model,
+                    modelUsed: modelUsed,
+                    promptTokens,
+                    completionTokens,
+                    totalTokens: promptTokens + completionTokens,
+                    costUsd: totalCost,
+                    latencyMs,
+                    success: true,
+                    errorMessage: null,
+                    userAgent: req.headers['user-agent'] || null,
+                    ipAddress: req.ip || null,
+                    requestPayload: {
+                        gateId: request.gateId,
+                        type: request.type,
+                        model: request.model,
+                        data: request.data,
+                        metadata: request.metadata,
+                    },
+                    responsePayload: {
+                        streamed: true,
+                        model: modelUsed,
+                        usage: { promptTokens, completionTokens, totalTokens: promptTokens + completionTokens },
+                        cost: totalCost,
+                    },
+                }).catch(err => console.error('Failed to log request:', err));
+            }
+            catch (streamError) {
+                const errorMessage = streamError instanceof Error ? streamError.message : 'Unknown streaming error';
+                res.write(`data: ${JSON.stringify({ error: 'stream_error', message: errorMessage })}\n\n`);
+                res.end();
+                db.logRequest({
+                    userId,
+                    gateId: gateConfig.id,
+                    gateName: gateConfig.name,
+                    modelRequested: request.model || gateConfig.model,
+                    modelUsed: null,
+                    promptTokens: 0,
+                    completionTokens: 0,
+                    totalTokens: 0,
+                    costUsd: 0,
+                    latencyMs: Date.now() - startTime,
+                    success: false,
+                    errorMessage,
+                    userAgent: req.headers['user-agent'] || null,
+                    ipAddress: req.ip || null,
+                    requestPayload: {
+                        gateId: request.gateId,
+                        type: request.type,
+                        model: request.model,
+                        data: request.data,
+                        metadata: request.metadata,
+                    },
+                    responsePayload: null,
+                }).catch(err => console.error('Failed to log request:', err));
+            }
+            return;
+        }
         const { result, modelUsed } = await executeWithRouting(gateConfig, finalRequest, userId);
         const latencyMs = Date.now() - startTime;
-        // Log request to database
         db.logRequest({
             userId,
             gateId: gateConfig.id,
@@ -218,7 +346,6 @@ router.post('/', authenticate, async (req, res) => {
                 finishReason: result.finishReason,
             },
         }).catch(err => console.error('Failed to log request:', err));
-        // Return LayerResponse with additional metadata
         const response = {
             ...result,
             model: modelUsed,

package/dist/services/providers/base-adapter.d.ts

@@ -17,6 +17,7 @@ export declare abstract class BaseProviderAdapter {
     protected imageMimeTypeMappings?: Record<ImageMimeType, string>;
     protected encodingFormatMappings?: Record<EncodingFormat, string>;
     abstract call(request: LayerRequest, userId?: string): Promise<LayerResponse>;
+    callStream?(request: LayerRequest, userId?: string): AsyncIterable<LayerResponse>;
     protected mapRole(role: Role): string;
     protected mapImageDetail(detail: ImageDetail): string | undefined;
     protected mapImageSize(size: ImageSize): string | undefined;

package/dist/services/providers/base-adapter.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"base-adapter.d.ts","sourceRoot":"","sources":["../../../src/services/providers/base-adapter.ts"],"names":[],"mappings":"AAAA,OAAO,EACL,YAAY,EACZ,aAAa,EACb,IAAI,EACJ,WAAW,EACX,SAAS,EACT,YAAY,EACZ,UAAU,EACV,SAAS,EACT,WAAW,EACX,aAAa,EACb,aAAa,EACb,YAAY,EACZ,UAAU,EACV,cAAc,EACd,eAAe,EAGhB,MAAM,eAAe,CAAC;AACvB,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,iCAAiC,CAAC;AAEhE,OAAO,EAAE,eAAe,EAAE,CAAC;AAE3B,8BAAsB,mBAAmB;IACvC,SAAS,CAAC,QAAQ,CAAC,QAAQ,EAAE,QAAQ,CAAC;IACtC,SAAS,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IAE1B,SAAS,CAAC,YAAY,CAAC,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAC9C,SAAS,CAAC,mBAAmB,CAAC,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IAC5D,SAAS,CAAC,kBAAkB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAAC,CAAC;IAC/D,SAAS,CAAC,oBAAoB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC;IAC9D,SAAS,CAAC,iBAAiB,CAAC,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACxD,SAAS,CAAC,oBAAoB,CAAC,EAAE,MAAM,CAAC,YAAY,EAAE,MAAM,CAAC,CAAC;IAC9D,SAAS,CAAC,kBAAkB,CAAC,EAAE,MAAM,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC;IAC1D,SAAS,CAAC,iBAAiB,CAAC,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACxD,SAAS,CAAC,mBAAmB,CAAC,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IAC5D,SAAS,CAAC,qBAAqB,CAAC,EAAE,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC,CAAC;IAChE,SAAS,CAAC,qBAAqB,CAAC,EAAE,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC,CAAC;IAChE,SAAS,CAAC,sBAAsB,CAAC,EAAE,MAAM,CAAC,cAAc,EAAE,MAAM,CAAC,CAAC;IAElE,QAAQ,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC;IAE7E,SAAS,CAAC,OAAO,CAAC,IAAI,EAAE,IAAI,GAAG,MAAM;IAcrC,SAAS,CAAC,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,MAAM,GAAG,SAAS;IAQjE,SAAS,CAAC,YAAY,CAAC,IAAI,EAAE,SAAS,GAAG,MAAM,GAAG,SAAS;IAQ3D,SAAS,CAAC,eAAe,CAAC,OAAO,EAAE,YAAY,GAAG,MAAM,GAAG,SAAS;IAQpE,SAAS,CAAC,aAAa,CAAC,KAAK,EAAE,UAAU,GAAG,MAAM,GAAG,SAAS;IAQ9D,SAAS,CAAC,YAAY,CAAC,IAAI,EAAE,SAAS,GAAG,MAAM,GAAG,SAAS;IAQ3D,SAAS,CAAC,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,MAAM,GAAG,SAAS;IAQjE,SAAS,CAAC,gBAAgB,CAAC,QAAQ,EAAE,aAAa,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,gBAAgB,CAAC,QAAQ,EAAE,aAAa,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,iBAAiB,CAAC,MAAM,EAAE,cAAc,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,eAAe,CAAC,oBAAoB,EAAE,MAAM,GAAG,YAAY;IAQrE,SAAS,CAAC,aAAa,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS;IAYxE,SAAS,CAAC,aAAa,CACrB,KAAK,EAAE,MAAM,EACb,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,MAAM,GACvB,MAAM;IAWT,SAAS,CAAC,kBAAkB,CAC1B,KAAK,EAAE,MAAM,EACb,OAAO,CAAC,EAAE,MAAM,EAChB,IAAI,CAAC,EAAE,MAAM,EACb,KAAK,GAAE,MAAU,GAChB,MAAM;IAqBT,SAAS,CAAC,kBAAkB,CAC1B,KAAK,EAAE,MAAM,EACb,QAAQ,CAAC,EAAE,MAAM,EACjB,KAAK,GAAE,MAAU,GAChB,MAAM;CAiBV"}
+{"version":3,"file":"base-adapter.d.ts","sourceRoot":"","sources":["../../../src/services/providers/base-adapter.ts"],"names":[],"mappings":"AAAA,OAAO,EACL,YAAY,EACZ,aAAa,EACb,IAAI,EACJ,WAAW,EACX,SAAS,EACT,YAAY,EACZ,UAAU,EACV,SAAS,EACT,WAAW,EACX,aAAa,EACb,aAAa,EACb,YAAY,EACZ,UAAU,EACV,cAAc,EACd,eAAe,EAGhB,MAAM,eAAe,CAAC;AACvB,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,iCAAiC,CAAC;AAEhE,OAAO,EAAE,eAAe,EAAE,CAAC;AAE3B,8BAAsB,mBAAmB;IACvC,SAAS,CAAC,QAAQ,CAAC,QAAQ,EAAE,QAAQ,CAAC;IACtC,SAAS,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC;IAE1B,SAAS,CAAC,YAAY,CAAC,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,CAAC;IAC9C,SAAS,CAAC,mBAAmB,CAAC,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IAC5D,SAAS,CAAC,kBAAkB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,GAAG,MAAM,CAAC,CAAC;IAC/D,SAAS,CAAC,oBAAoB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC;IAC9D,SAAS,CAAC,iBAAiB,CAAC,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACxD,SAAS,CAAC,oBAAoB,CAAC,EAAE,MAAM,CAAC,YAAY,EAAE,MAAM,CAAC,CAAC;IAC9D,SAAS,CAAC,kBAAkB,CAAC,EAAE,MAAM,CAAC,UAAU,EAAE,MAAM,CAAC,CAAC;IAC1D,SAAS,CAAC,iBAAiB,CAAC,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAAC;IACxD,SAAS,CAAC,mBAAmB,CAAC,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAAC;IAC5D,SAAS,CAAC,qBAAqB,CAAC,EAAE,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC,CAAC;IAChE,SAAS,CAAC,qBAAqB,CAAC,EAAE,MAAM,CAAC,aAAa,EAAE,MAAM,CAAC,CAAC;IAChE,SAAS,CAAC,sBAAsB,CAAC,EAAE,MAAM,CAAC,cAAc,EAAE,MAAM,CAAC,CAAC;IAElE,QAAQ,CAAC,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC;IAG7E,UAAU,CAAC,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,aAAa,CAAC;IAEjF,SAAS,CAAC,OAAO,CAAC,IAAI,EAAE,IAAI,GAAG,MAAM;IAcrC,SAAS,CAAC,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,MAAM,GAAG,SAAS;IAQjE,SAAS,CAAC,YAAY,CAAC,IAAI,EAAE,SAAS,GAAG,MAAM,GAAG,SAAS;IAQ3D,SAAS,CAAC,eAAe,CAAC,OAAO,EAAE,YAAY,GAAG,MAAM,GAAG,SAAS;IAQpE,SAAS,CAAC,aAAa,CAAC,KAAK,EAAE,UAAU,GAAG,MAAM,GAAG,SAAS;IAQ9D,SAAS,CAAC,YAAY,CAAC,IAAI,EAAE,SAAS,GAAG,MAAM,GAAG,SAAS;IAQ3D,SAAS,CAAC,cAAc,CAAC,MAAM,EAAE,WAAW,GAAG,MAAM,GAAG,SAAS;IAQjE,SAAS,CAAC,gBAAgB,CAAC,QAAQ,EAAE,aAAa,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,gBAAgB,CAAC,QAAQ,EAAE,aAAa,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,iBAAiB,CAAC,MAAM,EAAE,cAAc,GAAG,MAAM,GAAG,SAAS;IAQvE,SAAS,CAAC,eAAe,CAAC,oBAAoB,EAAE,MAAM,GAAG,YAAY;IAQrE,SAAS,CAAC,aAAa,CAAC,MAAM,EAAE,UAAU,GAAG,MAAM,GAAG,MAAM,GAAG,SAAS;IAYxE,SAAS,CAAC,aAAa,CACrB,KAAK,EAAE,MAAM,EACb,YAAY,EAAE,MAAM,EACpB,gBAAgB,EAAE,MAAM,GACvB,MAAM;IAWT,SAAS,CAAC,kBAAkB,CAC1B,KAAK,EAAE,MAAM,EACb,OAAO,CAAC,EAAE,MAAM,EAChB,IAAI,CAAC,EAAE,MAAM,EACb,KAAK,GAAE,MAAU,GAChB,MAAM;IAqBT,SAAS,CAAC,kBAAkB,CAC1B,KAAK,EAAE,MAAM,EACb,QAAQ,CAAC,EAAE,MAAM,EACjB,KAAK,GAAE,MAAU,GAChB,MAAM;CAiBV"}

package/dist/services/providers/openai-adapter.d.ts

@@ -12,7 +12,9 @@ export declare class OpenAIAdapter extends BaseProviderAdapter {
     protected videoSizeMappings: Record<VideoSize, string>;
     protected audioFormatMappings: Record<AudioFormat, string>;
     call(request: LayerRequest, userId?: string): Promise<LayerResponse>;
+    callStream(request: LayerRequest, userId?: string): AsyncIterable<LayerResponse>;
     private handleChat;
+    private handleChatStream;
     private handleImageGeneration;
     private handleEmbeddings;
     private handleTextToSpeech;

package/dist/services/providers/openai-adapter.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"openai-adapter.d.ts","sourceRoot":"","sources":["../../../src/services/providers/openai-adapter.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EACL,YAAY,EACZ,aAAa,EACb,IAAI,EACJ,WAAW,EACX,SAAS,EACT,YAAY,EACZ,UAAU,EACV,SAAS,EACT,WAAW,EACX,YAAY,EACb,MAAM,eAAe,CAAC;AACvB,OAAO,EAAY,KAAK,QAAQ,EAAE,MAAM,iCAAiC,CAAC;AAoB1E,qBAAa,aAAc,SAAQ,mBAAmB;IACpD,SAAS,CAAC,QAAQ,EAAE,QAAQ,CAAmB;IAE/C,SAAS,CAAC,YAAY,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,CAQ1C;IAEF,SAAS,CAAC,mBAAmB,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAIxD;IAEF,SAAS,CAAC,oBAAoB,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAK1D;IAEF,SAAS,CAAC,iBAAiB,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAQpD;IAEF,SAAS,CAAC,oBAAoB,EAAE,MAAM,CAAC,YAAY,EAAE,MAAM,CAAC,CAG1D;IAEF,SAAS,CAAC,kBAAkB,EAAE,MAAM,CAAC,UAAU,EAAE,MAAM,CAAC,CAGtD;IAEF,SAAS,CAAC,iBAAiB,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAKpD;IAEF,SAAS,CAAC,mBAAmB,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAOxD;IAEI,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC;YAoB5D,UAAU;YAoHV,qBAAqB;YAuCrB,gBAAgB;YAkChB,kBAAkB;CA+BjC"}
+{"version":3,"file":"openai-adapter.d.ts","sourceRoot":"","sources":["../../../src/services/providers/openai-adapter.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EACL,YAAY,EACZ,aAAa,EACb,IAAI,EACJ,WAAW,EACX,SAAS,EACT,YAAY,EACZ,UAAU,EACV,SAAS,EACT,WAAW,EACX,YAAY,EACb,MAAM,eAAe,CAAC;AACvB,OAAO,EAAY,KAAK,QAAQ,EAAE,MAAM,iCAAiC,CAAC;AAkB1E,qBAAa,aAAc,SAAQ,mBAAmB;IACpD,SAAS,CAAC,QAAQ,EAAE,QAAQ,CAAmB;IAE/C,SAAS,CAAC,YAAY,EAAE,MAAM,CAAC,IAAI,EAAE,MAAM,CAAC,CAQ1C;IAEF,SAAS,CAAC,mBAAmB,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAIxD;IAEF,SAAS,CAAC,oBAAoB,EAAE,MAAM,CAAC,MAAM,EAAE,YAAY,CAAC,CAK1D;IAEF,SAAS,CAAC,iBAAiB,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAQpD;IAEF,SAAS,CAAC,oBAAoB,EAAE,MAAM,CAAC,YAAY,EAAE,MAAM,CAAC,CAG1D;IAEF,SAAS,CAAC,kBAAkB,EAAE,MAAM,CAAC,UAAU,EAAE,MAAM,CAAC,CAGtD;IAEF,SAAS,CAAC,iBAAiB,EAAE,MAAM,CAAC,SAAS,EAAE,MAAM,CAAC,CAKpD;IAEF,SAAS,CAAC,mBAAmB,EAAE,MAAM,CAAC,WAAW,EAAE,MAAM,CAAC,CAOxD;IAEI,IAAI,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,aAAa,CAAC;IAmBnE,UAAU,CAAC,OAAO,EAAE,YAAY,EAAE,MAAM,CAAC,EAAE,MAAM,GAAG,aAAa,CAAC,aAAa,CAAC;YAYzE,UAAU;YA+GT,gBAAgB;YAyIjB,qBAAqB;YAsCrB,gBAAgB;YAkChB,kBAAkB;CA+BjC"}

package/dist/services/providers/openai-adapter.js

@@ -4,11 +4,9 @@ import { PROVIDER } from "../../lib/provider-constants.js";
 import { resolveApiKey } from '../../lib/key-resolver.js';
 let openai = null;
 function getOpenAIClient(apiKey) {
-    // If custom API key provided, create new client
     if (apiKey) {
         return new OpenAI({ apiKey });
     }
-    // Otherwise use singleton with platform key
     if (!openai) {
         openai = new OpenAI({
             apiKey: process.env.OPENAI_API_KEY,
@@ -73,7 +71,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
         };
     }
     async call(request, userId) {
-        // Resolve API key (BYOK → Platform key)
         const resolved = await resolveApiKey(this.provider, userId, process.env.OPENAI_API_KEY);
         switch (request.type) {
             case 'chat':
@@ -90,6 +87,16 @@ export class OpenAIAdapter extends BaseProviderAdapter {
                 throw new Error(`Unknown modality: ${request.type}`);
         }
     }
+    async *callStream(request, userId) {
+        const resolved = await resolveApiKey(this.provider, userId, process.env.OPENAI_API_KEY);
+        switch (request.type) {
+            case 'chat':
+                yield* this.handleChatStream(request, resolved.key, resolved.usedPlatformKey);
+                break;
+            default:
+                throw new Error(`Streaming not supported for type: ${request.type}`);
+        }
+    }
     async handleChat(request, apiKey, usedPlatformKey) {
         const startTime = Date.now();
         const client = getOpenAIClient(apiKey);
@@ -103,7 +110,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
         }
         for (const msg of chat.messages) {
             const role = this.mapRole(msg.role);
-            // Handle vision messages (content + images)
             if (msg.images && msg.images.length > 0) {
                 const content = [];
                 if (msg.content) {
@@ -121,7 +127,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
                 }
                 messages.push({ role: role, content });
             }
-            // Handle tool responses (mutually exclusive)
             else if (msg.toolCallId) {
                 messages.push({
                     role: 'tool',
@@ -129,7 +134,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
                     tool_call_id: msg.toolCallId,
                 });
             }
-            // Handle assistant messages with tool calls (can have content + tool_calls)
             else if (msg.toolCalls) {
                 messages.push({
                     role: 'assistant',
@@ -137,7 +141,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
                     tool_calls: msg.toolCalls,
                 });
             }
-            // Handle regular text messages
             else {
                 messages.push({
                     role,
@@ -161,7 +164,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
                 tools: chat.tools,
                 ...(chat.toolChoice && { tool_choice: chat.toolChoice }),
             }),
-            // Structured output support: convert camelCase to snake_case
             ...(chat.responseFormat && {
                 response_format: (typeof chat.responseFormat === 'string'
                     ? { type: chat.responseFormat }
@@ -191,6 +193,119 @@ export class OpenAIAdapter extends BaseProviderAdapter {
             raw: response,
         };
     }
+    async *handleChatStream(request, apiKey, usedPlatformKey) {
+        const startTime = Date.now();
+        const client = getOpenAIClient(apiKey);
+        const { data: chat, model } = request;
+        if (!model) {
+            throw new Error('Model is required for chat completion');
+        }
+        const messages = [];
+        if (chat.systemPrompt) {
+            messages.push({ role: 'system', content: chat.systemPrompt });
+        }
+        for (const msg of chat.messages) {
+            const role = this.mapRole(msg.role);
+            if (msg.images && msg.images.length > 0) {
+                const content = [];
+                if (msg.content) {
+                    content.push({ type: 'text', text: msg.content });
+                }
+                for (const image of msg.images) {
+                    const imageUrl = image.url || `data:${image.mimeType || 'image/jpeg'};base64,${image.base64}`;
+                    content.push({
+                        type: 'image_url',
+                        image_url: {
+                            url: imageUrl,
+                            ...(image.detail && { detail: this.mapImageDetail(image.detail) }),
+                        },
+                    });
+                }
+                messages.push({ role: role, content });
+            }
+            else if (msg.toolCallId) {
+                messages.push({
+                    role: 'tool',
+                    content: msg.content || '',
+                    tool_call_id: msg.toolCallId,
+                });
+            }
+            else if (msg.toolCalls) {
+                messages.push({
+                    role: 'assistant',
+                    content: msg.content || null,
+                    tool_calls: msg.toolCalls,
+                });
+            }
+            else {
+                messages.push({
+                    role,
+                    content: msg.content || '',
+                    ...(role === 'function' && msg.name && { name: msg.name }),
+                });
+            }
+        }
+        const openaiRequest = {
+            model: model,
+            messages,
+            stream: true,
+            ...(chat.temperature !== undefined && { temperature: chat.temperature }),
+            ...(chat.maxTokens !== undefined && { max_completion_tokens: chat.maxTokens }),
+            ...(chat.topP !== undefined && { top_p: chat.topP }),
+            ...(chat.stopSequences !== undefined && { stop: chat.stopSequences }),
+            ...(chat.frequencyPenalty !== undefined && { frequency_penalty: chat.frequencyPenalty }),
+            ...(chat.presencePenalty !== undefined && { presence_penalty: chat.presencePenalty }),
+            ...(chat.seed !== undefined && { seed: chat.seed }),
+            ...(chat.tools && {
+                tools: chat.tools,
+                ...(chat.toolChoice && { tool_choice: chat.toolChoice }),
+            }),
+            ...(chat.responseFormat && {
+                response_format: (typeof chat.responseFormat === 'string'
+                    ? { type: chat.responseFormat }
+                    : chat.responseFormat),
+            }),
+        };
+        const stream = await client.chat.completions.create(openaiRequest);
+        let promptTokens = 0;
+        let completionTokens = 0;
+        let fullContent = '';
+        for await (const chunk of stream) {
+            const delta = chunk.choices[0]?.delta;
+            const finishReason = chunk.choices[0]?.finish_reason;
+            if (delta?.content) {
+                fullContent += delta.content;
+            }
+            if (chunk.usage) {
+                promptTokens = chunk.usage.prompt_tokens || 0;
+                completionTokens = chunk.usage.completion_tokens || 0;
+            }
+            yield {
+                content: delta?.content || undefined,
+                toolCalls: delta?.tool_calls,
+                model: chunk.model || model,
+                finishReason: finishReason ? this.mapFinishReason(finishReason) : undefined,
+                rawFinishReason: finishReason || undefined,
+                stream: true,
+            };
+        }
+        const cost = this.calculateCost(model, promptTokens, completionTokens);
+        const latencyMs = Date.now() - startTime;
+        yield {
+            content: '',
+            model: model,
+            usage: {
+                promptTokens,
+                completionTokens,
+                totalTokens: promptTokens + completionTokens,
+            },
+            cost,
+            latencyMs,
+            usedPlatformKey,
+            stream: true,
+            finishReason: 'completed',
+        };
+    }
     async handleImageGeneration(request, apiKey, usedPlatformKey) {
         const startTime = Date.now();
         const client = getOpenAIClient(apiKey);
@@ -206,7 +321,6 @@ export class OpenAIAdapter extends BaseProviderAdapter {
             ...(image.count && { n: image.count }),
             ...(image.style && { style: this.mapImageStyle(image.style) }),
         });
-        // Calculate cost based on quality, size, and count
         const cost = this.calculateImageCost(model, image.quality || 'standard', image.size || '1024x1024', image.count || 1);
         return {
             images: (response.data || []).map(img => ({

package/dist/services/providers/tests/test-openai-streaming.d.ts

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=test-openai-streaming.d.ts.map

package/dist/services/providers/tests/test-openai-streaming.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"test-openai-streaming.d.ts","sourceRoot":"","sources":["../../../../src/services/providers/tests/test-openai-streaming.ts"],"names":[],"mappings":""}

package/dist/services/providers/tests/test-openai-streaming.js

@@ -0,0 +1,139 @@
+import { OpenAIAdapter } from '../openai-adapter.js';
+const adapter = new OpenAIAdapter();
+async function testChatStreamingBasic() {
+    console.log('Testing basic chat streaming...');
+    const request = {
+        gateId: 'test-gate',
+        model: 'gpt-4o-mini',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'Count from 1 to 5 slowly, one number per line.' }
+            ],
+            temperature: 0.7,
+            maxTokens: 50,
+            stream: true,
+        }
+    };
+    if (!adapter.callStream) {
+        throw new Error('callStream method not available');
+    }
+    let chunkCount = 0;
+    let fullContent = '';
+    let finalUsage = null;
+    let finalCost = null;
+    console.log('\nStreaming chunks:');
+    console.log('---');
+    for await (const chunk of adapter.callStream(request)) {
+        chunkCount++;
+        if (chunk.content) {
+            process.stdout.write(chunk.content);
+            fullContent += chunk.content;
+        }
+        if (chunk.usage) {
+            finalUsage = chunk.usage;
+        }
+        if (chunk.cost) {
+            finalCost = chunk.cost;
+        }
+    }
+    console.log('\n---\n');
+    console.log('Total chunks received:', chunkCount);
+    console.log('Full content:', fullContent);
+    console.log('Final usage:', finalUsage);
+    console.log('Final cost:', finalCost);
+    console.log('✅ Basic streaming test passed\n');
+}
+async function testChatStreamingWithToolCalls() {
+    console.log('Testing chat streaming with tool calls...');
+    const request = {
+        gateId: 'test-gate',
+        model: 'gpt-4o-mini',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'What is the weather in San Francisco?' }
+            ],
+            tools: [
+                {
+                    type: 'function',
+                    function: {
+                        name: 'get_weather',
+                        description: 'Get the current weather for a location',
+                        parameters: {
+                            type: 'object',
+                            properties: {
+                                location: {
+                                    type: 'string',
+                                    description: 'The city and state, e.g. San Francisco, CA',
+                                },
+                            },
+                            required: ['location'],
+                        },
+                    },
+                },
+            ],
+            toolChoice: 'auto',
+            stream: true,
+        }
+    };
+    if (!adapter.callStream) {
+        throw new Error('callStream method not available');
+    }
+    let toolCallsReceived = false;
+    for await (const chunk of adapter.callStream(request)) {
+        if (chunk.toolCalls && chunk.toolCalls.length > 0) {
+            console.log('Tool calls received:', JSON.stringify(chunk.toolCalls, null, 2));
+            toolCallsReceived = true;
+        }
+        if (chunk.finishReason === 'tool_call') {
+            console.log('Finish reason: tool_call');
+        }
+    }
+    if (!toolCallsReceived) {
+        console.warn('⚠️  No tool calls received (model may have chosen not to use tools)');
+    }
+    else {
+        console.log('✅ Tool calls streaming test passed\n');
+    }
+}
+async function testChatStreamingError() {
+    console.log('Testing streaming with invalid model (error handling)...');
+    const request = {
+        gateId: 'test-gate',
+        model: 'invalid-model-name-that-does-not-exist',
+        type: 'chat',
+        data: {
+            messages: [
+                { role: 'user', content: 'Hello' }
+            ],
+            stream: true,
+        }
+    };
+    if (!adapter.callStream) {
+        throw new Error('callStream method not available');
+    }
+    try {
+        for await (const chunk of adapter.callStream(request)) {
+            console.log('Received chunk:', chunk);
+        }
+        console.error('❌ Should have thrown an error for invalid model');
+    }
+    catch (error) {
+        console.log('✅ Correctly threw error:', error instanceof Error ? error.message : error);
+        console.log('✅ Error handling test passed\n');
+    }
+}
+// Run all tests
+(async () => {
+    try {
+        await testChatStreamingBasic();
+        await testChatStreamingWithToolCalls();
+        await testChatStreamingError();
+        console.log('✅ All streaming tests completed successfully!');
+    }
+    catch (error) {
+        console.error('❌ Test failed:', error);
+        process.exit(1);
+    }
+})();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@layer-ai/core",
-  "version": "2.0.14",
+  "version": "2.0.16",
   "description": "Core API routes and services for Layer AI",
   "type": "module",
   "main": "./dist/index.js",
@@ -36,7 +36,7 @@
     "nanoid": "^5.0.4",
     "openai": "^4.24.0",
     "pg": "^8.11.3",
-    "@layer-ai/sdk": "^2.5.5"
+    "@layer-ai/sdk": "^2.5.6"
   },
   "devDependencies": {
     "@types/bcryptjs": "^2.4.6",