fixup! test(e2e): Add a cloudflare-pi-durable end-to-end application

Author: JPeer264

dev-packages/e2e-tests/test-applications/cloudflare-pi-durable/package.json

@@ -22,6 +22,7 @@
     "@cloudflare/workers-types": "^4.20260426.0",
     "@playwright/test": "~1.63.0",
     "@sentry-internal/test-utils": "link:../../../test-utils",
+    "@sentry/conventions": "0.26.0",
     "typescript": "^5.5.2",
     "vite": "7.3.5",
     "wrangler": "^4.136.2"

dev-packages/e2e-tests/test-applications/cloudflare-pi-durable/tests/pi-durable.test.ts

@@ -1,4 +1,18 @@
 import { expect, test } from '@playwright/test';
+import {
+  GEN_AI_CONVERSATION_ID,
+  GEN_AI_PROVIDER_NAME,
+  GEN_AI_REQUEST_MODEL,
+  GEN_AI_RESPONSE_FINISH_REASONS,
+  GEN_AI_TOOL_CALL_ARGUMENTS,
+  GEN_AI_TOOL_CALL_RESULT,
+  GEN_AI_TOOL_NAME,
+  GEN_AI_USAGE_INPUT_TOKENS,
+  GEN_AI_USAGE_OUTPUT_TOKENS,
+  SENTRY_ORIGIN,
+  SERVER_ADDRESS,
+} from '@sentry/conventions/attributes';
+import { DB_QUERY, GEN_AI_CHAT, GEN_AI_INVOKE_AGENT, HTTP_CLIENT, HTTP_SERVER } from '@sentry/conventions/op';
 import { collectStreamedSpans, getSpanOp, waitForError } from '@sentry-internal/test-utils';
 import { newAgentId, prompt, submitPrompt, waitForOperation } from './utils';
 
@@ -9,8 +23,8 @@ test('traces a PiHarness prompt as invoke_agent with chat, execute_tool and prov
   const spansPromise = collectStreamedSpans(
     APP,
     spansOfTrace =>
-      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === 'gen_ai.invoke_agent') &&
-      spansOfTrace.some(span => span.attributes['gen_ai.tool.name']?.value === 'get_weather'),
+      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === GEN_AI_INVOKE_AGENT) &&
+      spansOfTrace.some(span => span.attributes[GEN_AI_TOOL_NAME]?.value === 'get_weather'),
   );
 
   const answer = await prompt(
@@ -22,52 +36,52 @@ test('traces a PiHarness prompt as invoke_agent with chat, execute_tool and prov
 
   const spans = await spansPromise;
   const agent = spans.find(span => span.is_segment)!;
-  const chats = spans.filter(span => getSpanOp(span) === 'gen_ai.chat');
+  const chats = spans.filter(span => getSpanOp(span) === GEN_AI_CHAT);
   // The manual span of the tool is the one way to pick the call that ran, should the model call twice.
   const manualSpan = spans.find(span => span.name === 'resolve-weather')!;
   const tool = spans.find(span => span.span_id === manualSpan.parent_span_id)!;
-  const providerCalls = spans.filter(span => getSpanOp(span) === 'http.client');
+  const providerCalls = spans.filter(span => getSpanOp(span) === HTTP_CLIENT);
 
-  expect(agent.attributes['sentry.origin']?.value).toBe('auto.ai.pi_durable');
+  expect(agent.attributes[SENTRY_ORIGIN]?.value).toBe('auto.ai.pi_durable');
   // The PiHarness root session is pi conversation 1, in a Harness of its own per Durable Object.
-  expect(agent.attributes['gen_ai.conversation.id']?.value).toMatch(/^[0-9a-f]{32}:1$/);
+  expect(agent.attributes[GEN_AI_CONVERSATION_ID]?.value).toMatch(/^[0-9a-f]{32}:1$/);
   // The request that submitted the prompt is a trace of its own: the scheduler runs the work.
-  expect(spans.some(span => getSpanOp(span) === 'http.server')).toBe(false);
+  expect(spans.some(span => getSpanOp(span) === HTTP_SERVER)).toBe(false);
   // pi-durable keeps its state in the `pi_` tables of `PiHarness`, whose statements start no span.
-  expect(spans.filter(span => getSpanOp(span) === 'db.query' && /\bpi_/.test(String(span.name)))).toEqual([]);
+  expect(spans.filter(span => getSpanOp(span) === DB_QUERY && /\bpi_/.test(String(span.name)))).toEqual([]);
   // pi-ai sends the requests through `@anthropic-ai/sdk`, whose own integration must stay out so
   // each request is reported once.
-  expect(spans.filter(span => String(span.attributes['sentry.origin']?.value).startsWith('auto.ai.'))).toEqual(
-    spans.filter(span => span.attributes['sentry.origin']?.value === 'auto.ai.pi_durable'),
+  expect(spans.filter(span => String(span.attributes[SENTRY_ORIGIN]?.value).startsWith('auto.ai.'))).toEqual(
+    spans.filter(span => span.attributes[SENTRY_ORIGIN]?.value === 'auto.ai.pi_durable'),
   );
 
   // One tool-calling response, then the answer.
   expect(chats.length).toBeGreaterThanOrEqual(2);
   for (const chat of chats) {
     expect(chat.parent_span_id).toBe(agent.span_id);
-    expect(chat.attributes['gen_ai.provider.name']?.value).toBe('openrouter');
-    expect(chat.attributes['gen_ai.request.model']?.value).toBe('anthropic/claude-haiku-4.5');
-    expect(chat.attributes['gen_ai.conversation.id']?.value).toBe(agent.attributes['gen_ai.conversation.id']?.value);
+    expect(chat.attributes[GEN_AI_PROVIDER_NAME]?.value).toBe('openrouter');
+    expect(chat.attributes[GEN_AI_REQUEST_MODEL]?.value).toBe('anthropic/claude-haiku-4.5');
+    expect(chat.attributes[GEN_AI_CONVERSATION_ID]?.value).toBe(agent.attributes[GEN_AI_CONVERSATION_ID]?.value);
     // pi-durable retries a provider error inside the run; such a request has no usage and no HTTP span
     // of its own to assert on.
     if (chat.status === 'ok') {
-      expect(typeof chat.attributes['gen_ai.usage.input_tokens']?.value).toBe('number');
-      expect(typeof chat.attributes['gen_ai.usage.output_tokens']?.value).toBe('number');
+      expect(typeof chat.attributes[GEN_AI_USAGE_INPUT_TOKENS]?.value).toBe('number');
+      expect(typeof chat.attributes[GEN_AI_USAGE_OUTPUT_TOKENS]?.value).toBe('number');
       expect(providerCalls.some(providerCall => providerCall.parent_span_id === chat.span_id)).toBe(true);
     }
   }
 
-  expect(tool.attributes['gen_ai.tool.name']?.value).toBe('get_weather');
+  expect(tool.attributes[GEN_AI_TOOL_NAME]?.value).toBe('get_weather');
   expect(tool.parent_span_id).toBe(agent.span_id);
   expect(tool.status).toBe('ok');
-  expect(tool.attributes['gen_ai.tool.call.arguments']?.value).toContain('Vienna');
-  expect(tool.attributes['gen_ai.tool.call.result']?.value).toContain('21 degrees and sunny in');
+  expect(tool.attributes[GEN_AI_TOOL_CALL_ARGUMENTS]?.value).toContain('Vienna');
+  expect(tool.attributes[GEN_AI_TOOL_CALL_RESULT]?.value).toContain('21 degrees and sunny in');
 
   // The provider's HTTP calls nest inside the `chat` span that sent them.
   expect(providerCalls.length).toBeGreaterThan(0);
   for (const providerCall of providerCalls) {
     expect(chats.map(chat => chat.span_id)).toContain(providerCall.parent_span_id);
-    expect(providerCall.attributes['server.address']?.value).toBe('openrouter.ai');
+    expect(providerCall.attributes[SERVER_ADDRESS]?.value).toBe('openrouter.ai');
   }
 });
 
@@ -79,8 +93,8 @@ test('reports a throwing tool as an error on its execute_tool span', async ({ ba
   const spansPromise = collectStreamedSpans(
     APP,
     spansOfTrace =>
-      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === 'gen_ai.invoke_agent') &&
-      spansOfTrace.some(span => span.attributes['gen_ai.tool.name']?.value === 'fail_now'),
+      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === GEN_AI_INVOKE_AGENT) &&
+      spansOfTrace.some(span => span.attributes[GEN_AI_TOOL_NAME]?.value === 'fail_now'),
   );
 
   const answer = await prompt(baseURL!, newAgentId('fail'), 'Call the fail_now tool, then tell me what happened.');
@@ -90,7 +104,7 @@ test('reports a throwing tool as an error on its execute_tool span', async ({ ba
   // The error is reported on the span of the call that threw.
   const tool = spans.find(span => span.span_id === error.contexts?.trace?.span_id)!;
 
-  expect(tool.attributes['gen_ai.tool.name']?.value).toBe('fail_now');
+  expect(tool.attributes[GEN_AI_TOOL_NAME]?.value).toBe('fail_now');
   expect(tool.status).toBe('error');
   expect(tool.trace_id).toBe(error.contexts?.trace?.trace_id);
   expect(error.exception?.values?.[0]?.mechanism).toEqual({ type: 'auto.ai.pi_durable', handled: false });
@@ -104,8 +118,8 @@ test('resumes a run in a new trace after the Durable Object resets during a tool
   const spansPromise = collectStreamedSpans(
     APP,
     spansOfTrace =>
-      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === 'gen_ai.invoke_agent') &&
-      spansOfTrace.some(span => span.attributes['gen_ai.tool.name']?.value === 'crash_once' && span.status === 'ok'),
+      spansOfTrace.some(span => span.is_segment && getSpanOp(span) === GEN_AI_INVOKE_AGENT) &&
+      spansOfTrace.some(span => span.attributes[GEN_AI_TOOL_NAME]?.value === 'crash_once' && span.status === 'ok'),
   );
 
   // The reset fails the request that submitted the prompt, but pi has stored the input already.
@@ -120,11 +134,11 @@ test('resumes a run in a new trace after the Durable Object resets during a tool
 
   const spans = await spansPromise;
   const agent = spans.find(span => span.is_segment)!;
-  const tool = spans.find(span => span.attributes['gen_ai.tool.name']?.value === 'crash_once')!;
-  const chats = spans.filter(span => getSpanOp(span) === 'gen_ai.chat');
-  const finalAnswer = chats.find(span => span.attributes['gen_ai.response.finish_reasons']?.value === '["stop"]');
+  const tool = spans.find(span => span.attributes[GEN_AI_TOOL_NAME]?.value === 'crash_once')!;
+  const chats = spans.filter(span => getSpanOp(span) === GEN_AI_CHAT);
+  const finalAnswer = chats.find(span => span.attributes[GEN_AI_RESPONSE_FINISH_REASONS]?.value === '["stop"]');
 
-  expect(getSpanOp(agent)).toBe('gen_ai.invoke_agent');
+  expect(getSpanOp(agent)).toBe(GEN_AI_INVOKE_AGENT);
   expect(tool.parent_span_id).toBe(agent.span_id);
   expect(finalAnswer?.parent_span_id).toBe(agent.span_id);
   // The request that called the tool ran before the reset, so this trace starts with the rerun of the