feat(cloudflare,vercel-edge): Add support for OpenAI instrumentation

CHANGELOG.md

@@ -25,6 +25,22 @@ Sentry.init({
 
 Spans matching the filter criteria will not be recorded. Potential child spans of filtered spans will be re-parented, if possible.
 
+- **feat(cloudflare,vercel-edge): Add support for OpenAI instrumentation ([#17338](https://github.com/getsentry/sentry-javascript/pull/17338))**
+
+Adds support for OpenAI manual instrumentation in `@sentry/cloudflare` and `@sentry/vercel-edge`.
+
+To instrument the OpenAI client, wrap it with `Sentry.instrumentOpenAiClient` and set recording settings.
+
+```js
+import * as Sentry from '@sentry/cloudflare';
+import OpenAI from 'openai';
+
+const openai = new OpenAI();
+const client = Sentry.instrumentOpenAiClient(openai, { recordInputs: true, recordOutputs: true });
+
+// use the wrapped client
+```
+
 ## 10.1.0
 
 - feat(nuxt): Align build-time options to follow bundler plugins structure ([#17255](https://github.com/getsentry/sentry-javascript/pull/17255))

dev-packages/cloudflare-integration-tests/package.json

@@ -16,6 +16,7 @@
     "@sentry/cloudflare": "10.1.0"
   },
   "devDependencies": {
+    "@sentry/core": "10.0.0",
     "@cloudflare/workers-types": "^4.20250708.0",
     "@sentry-internal/test-utils": "link:../test-utils",
     "vitest": "^3.2.4",

dev-packages/cloudflare-integration-tests/suites/tracing/openai/index.ts

@@ -0,0 +1,34 @@
+import * as Sentry from '@sentry/cloudflare';
+import { MockOpenAi } from './mocks';
+
+interface Env {
+  SENTRY_DSN: string;
+}
+
+const mockClient = new MockOpenAi({
+  apiKey: 'mock-api-key',
+});
+
+const client = Sentry.instrumentOpenAiClient(mockClient);
+
+export default Sentry.withSentry(
+  (env: Env) => ({
+    dsn: env.SENTRY_DSN,
+    tracesSampleRate: 1.0,
+  }),
+  {
+    async fetch(_request, _env, _ctx) {
+      const response = await client.chat?.completions?.create({
+        model: 'gpt-3.5-turbo',
+        messages: [
+          { role: 'system', content: 'You are a helpful assistant.' },
+          { role: 'user', content: 'What is the capital of France?' },
+        ],
+        temperature: 0.7,
+        max_tokens: 100,
+      });
+
+      return new Response(JSON.stringify(response));
+    },
+  },
+);

dev-packages/cloudflare-integration-tests/suites/tracing/openai/mocks.ts

@@ -0,0 +1,245 @@
+import type { OpenAiClient } from '@sentry/core';
+
+export class MockOpenAi implements OpenAiClient {
+  public chat: Record<string, unknown>;
+  public responses: {
+    create: (...args: unknown[]) => Promise<unknown>;
+  };
+
+  public apiKey: string;
+
+  public constructor(config: { apiKey: string }) {
+    this.apiKey = config.apiKey;
+
+    this.chat = {
+      completions: {
+        create: async (...args: unknown[]) => {
+          const params = args[0] as { model: string; stream?: boolean };
+          // Simulate processing time
+          await new Promise(resolve => setTimeout(resolve, 10));
+
+          if (params.model === 'error-model') {
+            const error = new Error('Model not found');
+            (error as unknown as { status: number }).status = 404;
+            (error as unknown as { headers: Record<string, string> }).headers = { 'x-request-id': 'mock-request-123' };
+            throw error;
+          }
+
+          // If stream is requested, return an async generator
+          if (params.stream) {
+            return this.createChatCompletionStream(params);
+          }
+
+          return {
+            id: 'chatcmpl-mock123',
+            object: 'chat.completion',
+            created: 1677652288,
+            model: params.model,
+            system_fingerprint: 'fp_44709d6fcb',
+            choices: [
+              {
+                index: 0,
+                message: {
+                  role: 'assistant',
+                  content: 'Hello from OpenAI mock!',
+                },
+                finish_reason: 'stop',
+              },
+            ],
+            usage: {
+              prompt_tokens: 10,
+              completion_tokens: 15,
+              total_tokens: 25,
+            },
+          };
+        },
+      },
+    };
+
+    this.responses = {
+      create: async (...args: unknown[]) => {
+        const params = args[0] as { model: string; input: string; instructions: string; stream?: boolean };
+        await new Promise(resolve => setTimeout(resolve, 10));
+
+        // If stream is requested, return an async generator
+        if (params.stream) {
+          return this.createResponsesApiStream(params);
+        }
+
+        return {
+          id: 'resp_mock456',
+          object: 'response',
+          created_at: 1677652290,
+          model: params.model,
+          input_text: params.input,
+          output_text: `Response to: ${params.input}`,
+          status: 'completed',
+          usage: {
+            input_tokens: 5,
+            output_tokens: 8,
+            total_tokens: 13,
+          },
+        };
+      },
+    };
+  }
+
+  // Create a mock streaming response for chat completions
+  public async *createChatCompletionStream(params: { model: string }): AsyncGenerator<unknown> {
+    // First chunk with basic info
+    yield {
+      id: 'chatcmpl-stream-123',
+      object: 'chat.completion.chunk',
+      created: 1677652300,
+      model: params.model,
+      system_fingerprint: 'fp_stream_123',
+      choices: [
+        {
+          index: 0,
+          delta: {
+            role: 'assistant',
+            content: 'Hello',
+          },
+          finish_reason: null,
+        },
+      ],
+    };
+
+    // Second chunk with more content
+    yield {
+      id: 'chatcmpl-stream-123',
+      object: 'chat.completion.chunk',
+      created: 1677652300,
+      model: params.model,
+      system_fingerprint: 'fp_stream_123',
+      choices: [
+        {
+          index: 0,
+          delta: {
+            content: ' from OpenAI streaming!',
+          },
+          finish_reason: 'stop',
+        },
+      ],
+      usage: {
+        prompt_tokens: 12,
+        completion_tokens: 18,
+        total_tokens: 30,
+        completion_tokens_details: {
+          accepted_prediction_tokens: 0,
+          audio_tokens: 0,
+          reasoning_tokens: 0,
+          rejected_prediction_tokens: 0,
+        },
+        prompt_tokens_details: {
+          audio_tokens: 0,
+          cached_tokens: 0,
+        },
+      },
+    };
+  }
+
+  // Create a mock streaming response for responses API
+  public async *createResponsesApiStream(params: {
+    model: string;
+    input: string;
+    instructions: string;
+  }): AsyncGenerator<unknown> {
+    // Response created event
+    yield {
+      type: 'response.created',
+      response: {
+        id: 'resp_stream_456',
+        object: 'response',
+        created_at: 1677652310,
+        model: params.model,
+        status: 'in_progress',
+        error: null,
+        incomplete_details: null,
+        instructions: params.instructions,
+        max_output_tokens: 1000,
+        parallel_tool_calls: false,
+        previous_response_id: null,
+        reasoning: {
+          effort: null,
+          summary: null,
+        },
+        store: false,
+        temperature: 0.7,
+        text: {
+          format: {
+            type: 'text',
+          },
+        },
+        tool_choice: 'auto',
+        top_p: 1.0,
+        truncation: 'disabled',
+        user: null,
+        metadata: {},
+        output: [],
+        output_text: '',
+        usage: {
+          input_tokens: 0,
+          output_tokens: 0,
+          total_tokens: 0,
+        },
+      },
+      sequence_number: 1,
+    };
+
+    // Response in progress with output text delta
+    yield {
+      type: 'response.output_text.delta',
+      delta: 'Streaming response to: ',
+      sequence_number: 2,
+    };
+
+    yield {
+      type: 'response.output_text.delta',
+      delta: params.input,
+      sequence_number: 3,
+    };
+
+    // Response completed event
+    yield {
+      type: 'response.completed',
+      response: {
+        id: 'resp_stream_456',
+        object: 'response',
+        created_at: 1677652310,
+        model: params.model,
+        status: 'completed',
+        error: null,
+        incomplete_details: null,
+        instructions: params.instructions,
+        max_output_tokens: 1000,
+        parallel_tool_calls: false,
+        previous_response_id: null,
+        reasoning: {
+          effort: null,
+          summary: null,
+        },
+        store: false,
+        temperature: 0.7,
+        text: {
+          format: {
+            type: 'text',
+          },
+        },
+        tool_choice: 'auto',
+        top_p: 1.0,
+        truncation: 'disabled',
+        user: null,
+        metadata: {},
+        output: [],
+        output_text: params.input,
+        usage: {
+          input_tokens: 6,
+          output_tokens: 10,
+          total_tokens: 16,
+        },
+      },
+      sequence_number: 4,
+    };
+  }
+}

dev-packages/cloudflare-integration-tests/suites/tracing/openai/test.ts

@@ -0,0 +1,42 @@
+import { expect, it } from 'vitest';
+import { createRunner } from '../../../runner';
+
+// These tests are not exhaustive because the instrumentation is
+// already tested in the node integration tests and we merely
+// want to test that the instrumentation does not break in our
+// cloudflare SDK.
+
+it('traces a basic chat completion request', async () => {
+  const runner = createRunner(__dirname)
+    .ignore('event')
+    .expect(envelope => {
+      const transactionEvent = envelope[1]?.[0]?.[1];
+
+      expect(transactionEvent.transaction).toBe('GET /');
+      expect(transactionEvent.spans).toEqual(
+        expect.arrayContaining([
+          expect.objectContaining({
+            data: expect.objectContaining({
+              'gen_ai.operation.name': 'chat',
+              'sentry.op': 'gen_ai.chat',
+              'gen_ai.system': 'openai',
+              'gen_ai.request.model': 'gpt-3.5-turbo',
+              'gen_ai.request.temperature': 0.7,
+              'gen_ai.response.model': 'gpt-3.5-turbo',
+              'gen_ai.response.id': 'chatcmpl-mock123',
+              'gen_ai.usage.input_tokens': 10,
+              'gen_ai.usage.output_tokens': 15,
+              'gen_ai.usage.total_tokens': 25,
+              'gen_ai.response.finish_reasons': '["stop"]',
+            }),
+            description: 'chat gpt-3.5-turbo',
+            op: 'gen_ai.chat',
+            origin: 'manual',
+          }),
+        ]),
+      );
+    })
+    .start();
+  await runner.makeRequest('get', '/');
+  await runner.completed();
+});

dev-packages/cloudflare-integration-tests/suites/tracing/openai/wrangler.jsonc

@@ -0,0 +1,6 @@
+{
+  "name": "worker-name",
+  "compatibility_date": "2025-06-17",
+  "main": "index.ts",
+  "compatibility_flags": ["nodejs_compat"]
+}

dev-packages/cloudflare-integration-tests/vite.config.mts

@@ -22,6 +22,12 @@ export default defineConfig({
     // already run in their own processes. We use threads instead because the
     // overhead is significantly less.
     pool: 'threads',
+    // Run tests sequentially to avoid port conflicts with wrangler dev processes
+    poolOptions: {
+      threads: {
+        singleThread: true,
+      },
+    },
     reporters: process.env.DEBUG
       ? ['default', { summary: false }]
       : process.env.GITHUB_ACTIONS

packages/cloudflare/src/index.ts

@@ -69,6 +69,7 @@ export {
   functionToStringIntegration,
   // eslint-disable-next-line deprecation/deprecation
   inboundFiltersIntegration,
+  instrumentOpenAiClient,
   eventFiltersIntegration,
   linkedErrorsIntegration,
   requestDataIntegration,