Re recorderd ai sdk tests + added the ai sdk example

Author: dvirski

packages/sample-app/src/sample_vercel_ai_otel140.ts

@@ -0,0 +1,295 @@
+/**
+ * Vercel AI SDK - OTel 1.40 Migration Validation Sample
+ *
+ * Covers all major use cases to validate span shapes after the OTel 1.40 migration:
+ *   1. generateText  - OpenAI + Anthropic (basic chat)
+ *   2. streamText    - OpenAI streaming
+ *   3. generateObject - structured output
+ *   4. generateText with tools + multi-step (agentic loop)
+ *   5. Agent span naming via ai.telemetry.metadata.agent
+ *   6. Conversation ID propagation
+ *
+ * Run via run.sh (sets OPENAI_API_KEY, ANTHROPIC_API_KEY, TRACELOOP_API_KEY).
+ *
+ * Key attributes to verify in Traceloop dashboard:
+ *   - gen_ai.provider.name  → "openai" / "anthropic"  (OTel 1.40 lowercase)
+ *   - gen_ai.request.model  → model name
+ *   - gen_ai.operation.name → "chat" / "execute_tool"
+ *   - gen_ai.input.messages → [{role, parts:[{type:"text",content:...}]}]
+ *   - gen_ai.output.messages → [{role,finish_reason,parts:[...]}]
+ *   - gen_ai.usage.input_tokens / output_tokens
+ *   - gen_ai.response.finish_reasons → ["stop"] / ["tool_call"]
+ *   - gen_ai.tool.definitions → JSON array of tool objects (source format)
+ *   - Span name format: "chat {model}" / "execute_tool {toolName}"
+ */
+
+import * as traceloop from "@traceloop/node-server-sdk";
+import { openai } from "@ai-sdk/openai";
+import { anthropic } from "@ai-sdk/anthropic";
+import {
+  generateText,
+  streamText,
+  generateObject,
+  tool,
+  stepCountIs,
+} from "ai";
+import { z } from "zod";
+
+traceloop.initialize({
+  appName: "sample_vercel_ai_otel140",
+  disableBatch: true,
+});
+
+// ─── Shared tools ──────────────────────────────────────────────────────────
+
+const getWeather = tool({
+  description: "Get current weather for a location",
+  inputSchema: z.object({
+    location: z.string().describe("City name"),
+    units: z.enum(["celsius", "fahrenheit"]).optional().default("celsius"),
+  }),
+  execute: async ({ location, units }) => {
+    console.log(`  [tool] getWeather(${location}, ${units})`);
+    return {
+      location,
+      temperature: units === "celsius" ? 22 : 72,
+      condition: "Sunny",
+      humidity: 55,
+    };
+  },
+});
+
+const searchWeb = tool({
+  description: "Search the web for information",
+  inputSchema: z.object({
+    query: z.string().describe("Search query"),
+  }),
+  execute: async ({ query }) => {
+    console.log(`  [tool] searchWeb("${query}")`);
+    return {
+      results: [
+        { title: `Result 1 for: ${query}`, url: "https://example.com/1" },
+        { title: `Result 2 for: ${query}`, url: "https://example.com/2" },
+      ],
+    };
+  },
+});
+
+// ─── 1. Basic generateText — OpenAI ────────────────────────────────────────
+
+async function testGenerateTextOpenAI() {
+  console.log("\n▶ [1] generateText — OpenAI");
+
+  const result = await traceloop.withWorkflow(
+    { name: "openai_basic_chat" },
+    async () =>
+      generateText({
+        model: openai("gpt-4o-mini"),
+        messages: [
+          { role: "system", content: "You are a concise assistant." },
+          { role: "user", content: "What is the capital of France?" },
+        ],
+        experimental_telemetry: {
+          isEnabled: true,
+          metadata: { scenario: "basic_chat", provider: "openai" },
+        },
+      }),
+  );
+
+  console.log(`  Response: ${result.text}`);
+  console.log(
+    `  Tokens: ${result.usage.promptTokens} in / ${result.usage.completionTokens} out`,
+  );
+}
+
+// ─── 2. Basic generateText — Anthropic ─────────────────────────────────────
+
+async function testGenerateTextAnthropic() {
+  console.log("\n▶ [2] generateText — Anthropic");
+
+  const result = await traceloop.withWorkflow(
+    { name: "anthropic_basic_chat" },
+    async () =>
+      generateText({
+        model: anthropic("claude-haiku-4-5"),
+        messages: [
+          { role: "user", content: "What is the capital of Germany?" },
+        ],
+        experimental_telemetry: {
+          isEnabled: true,
+          metadata: { scenario: "basic_chat", provider: "anthropic" },
+        },
+      }),
+  );
+
+  console.log(`  Response: ${result.text}`);
+  console.log(
+    `  Tokens: ${result.usage.promptTokens} in / ${result.usage.completionTokens} out`,
+  );
+}
+
+// ─── 3. streamText — OpenAI ─────────────────────────────────────────────────
+
+async function testStreamText() {
+  console.log("\n▶ [3] streamText — OpenAI");
+
+  const result = await traceloop.withWorkflow(
+    { name: "openai_stream" },
+    async () => {
+      const stream = streamText({
+        model: openai("gpt-4o-mini"),
+        prompt: "Count from 1 to 5, one number per line.",
+        experimental_telemetry: {
+          isEnabled: true,
+          metadata: { scenario: "streaming" },
+        },
+      });
+
+      let fullText = "";
+      process.stdout.write("  Stream: ");
+      for await (const chunk of stream.textStream) {
+        process.stdout.write(chunk);
+        fullText += chunk;
+      }
+      console.log();
+      return fullText;
+    },
+  );
+
+  console.log(`  Streamed ${result.length} chars`);
+}
+
+// ─── 4. generateObject — structured output ──────────────────────────────────
+
+async function testGenerateObject() {
+  console.log("\n▶ [4] generateObject — OpenAI structured output");
+
+  const result = await traceloop.withWorkflow(
+    { name: "openai_structured_output" },
+    async () =>
+      generateObject({
+        model: openai("gpt-4o-mini"),
+        schema: z.object({
+          city: z.string(),
+          country: z.string(),
+          population: z.number(),
+          famousFor: z.array(z.string()).max(3),
+        }),
+        prompt: "Give me facts about Paris, France.",
+        experimental_telemetry: {
+          isEnabled: true,
+          metadata: { scenario: "structured_output" },
+        },
+      }),
+  );
+
+  console.log(`  Object:`, result.object);
+}
+
+// ─── 5. generateText with tools — multi-step agent loop ─────────────────────
+
+async function testToolsOpenAI() {
+  console.log("\n▶ [5] generateText + tools (multi-step) — OpenAI");
+
+  await traceloop.withAgent({ name: "travel_researcher" }, async () =>
+    generateText({
+      model: openai("gpt-4o-mini"),
+      prompt:
+        "What's the weather in Tokyo right now? Also search for 'best things to do in Tokyo'.",
+      tools: { getWeather, searchWeb },
+      stopWhen: stepCountIs(4),
+      experimental_telemetry: {
+        isEnabled: true,
+        metadata: {
+          agent: "travel_researcher",
+          scenario: "multi_step_tools",
+        },
+      },
+    }),
+  );
+
+  console.log("  Agent completed tool-use loop");
+}
+
+// ─── 6. generateText with tools — Anthropic ─────────────────────────────────
+
+async function testToolsAnthropic() {
+  console.log("\n▶ [6] generateText + tools — Anthropic");
+
+  await traceloop.withAgent({ name: "weather_agent_anthropic" }, async () =>
+    generateText({
+      model: anthropic("claude-haiku-4-5"),
+      prompt: "What is the weather like in London?",
+      tools: { getWeather },
+      stopWhen: stepCountIs(3),
+      experimental_telemetry: {
+        isEnabled: true,
+        metadata: {
+          agent: "weather_agent_anthropic",
+          scenario: "tool_call_anthropic",
+        },
+      },
+    }),
+  );
+
+  console.log("  Anthropic agent completed");
+}
+
+// ─── 7. Conversation ID propagation ─────────────────────────────────────────
+
+async function testConversationId() {
+  console.log("\n▶ [7] Conversation ID — OpenAI multi-turn");
+
+  const conversationId = `conv-${Date.now()}`;
+
+  await traceloop.withWorkflow({ name: "multi_turn_chat" }, async () => {
+    const turn1 = await generateText({
+      model: openai("gpt-4o-mini"),
+      messages: [{ role: "user", content: "My name is Alice." }],
+      experimental_telemetry: {
+        isEnabled: true,
+        metadata: { conversationId, turn: "1" },
+      },
+    });
+    console.log(`  Turn 1: ${turn1.text}`);
+
+    const turn2 = await generateText({
+      model: openai("gpt-4o-mini"),
+      messages: [
+        { role: "user", content: "My name is Alice." },
+        { role: "assistant", content: turn1.text },
+        { role: "user", content: "What is my name?" },
+      ],
+      experimental_telemetry: {
+        isEnabled: true,
+        metadata: { conversationId, turn: "2" },
+      },
+    });
+    console.log(`  Turn 2: ${turn2.text}`);
+  });
+}
+
+// ─── main ────────────────────────────────────────────────────────────────────
+
+async function main() {
+  console.log("=".repeat(60));
+  console.log("  Vercel AI SDK — OTel 1.40 Migration Validation");
+  console.log("=".repeat(60));
+
+  await testGenerateTextOpenAI();
+  await testGenerateTextAnthropic();
+  await testStreamText();
+  await testGenerateObject();
+  await testToolsOpenAI();
+  await testToolsAnthropic();
+  await testConversationId();
+
+  console.log("\n" + "=".repeat(60));
+  console.log("  All scenarios complete — check Traceloop dashboard");
+  console.log("=".repeat(60));
+}
+
+main().catch((err) => {
+  console.error("Fatal error:", err);
+  process.exit(1);
+});