feat(cases): added 2-multi-step for js/anthropic, js/google-genai, js/langgraph, js/vercel, py/anthropic, py/langchain, py/langgraph, py/openai-agents, py/pydantic-ai

sdks/js/anthropic/cases/2-multi-step.js

@@ -0,0 +1,75 @@
+/**
+ * 2-multi-step: Multi-step Conversation
+ *
+ * Tests a multi-step conversation with conversation history using Anthropic SDK
+ * and verifies that Sentry captures all spans for both API calls.
+ */
+
+const { Sentry } = require("../setup");
+const Anthropic = require("@anthropic-ai/sdk");
+const { runTestCase } = require("../../_test-utils/test-runner.cjs");
+
+async function testLogic(inputs) {
+  const { model, system, first_prompt, second_prompt } = inputs;
+
+  // Initialize Anthropic client
+  const anthropic = new Anthropic({
+    apiKey: process.env.ANTHROPIC_API_KEY,
+  });
+
+  // First call
+  const firstMessage = await anthropic.messages.create({
+    model: model,
+    max_tokens: 1024,
+    system: system,
+    messages: [
+      {
+        role: "user",
+        content: first_prompt,
+      },
+    ],
+  });
+
+  if (!firstMessage.content || firstMessage.content.length === 0) {
+    throw new Error("No completion returned from Anthropic (first call)");
+  }
+
+  const firstText = firstMessage.content[0].text;
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    First response: ${firstText}`);
+  }
+
+  // Second call with conversation history
+  const secondMessage = await anthropic.messages.create({
+    model: model,
+    max_tokens: 1024,
+    system: system,
+    messages: [
+      {
+        role: "user",
+        content: first_prompt,
+      },
+      {
+        role: "assistant",
+        content: firstText,
+      },
+      {
+        role: "user",
+        content: second_prompt,
+      },
+    ],
+  });
+
+  if (!secondMessage.content || secondMessage.content.length === 0) {
+    throw new Error("No completion returned from Anthropic (second call)");
+  }
+
+  const secondText = secondMessage.content[0].text;
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    Second response: ${secondText}`);
+  }
+}
+
+module.exports = runTestCase("2-multi-step", testLogic, Sentry);

sdks/js/anthropic/config.json

@@ -10,6 +10,11 @@
       "model": "claude-haiku-4-5",
       "gen_ai.request.model": "claude-haiku-4-5",
       "gen_ai.response.model": "claude-haiku-4-5*"
+    },
+    "2-multi-step": {
+      "model": "claude-haiku-4-5",
+      "gen_ai.request.model": "claude-haiku-4-5",
+      "gen_ai.response.model": "claude-haiku-4-5*"
     }
   }
 }

sdks/js/google-genai/cases/2-multi-step.js

@@ -0,0 +1,62 @@
+/**
+ * 2-multi-step: Multi-step Conversation
+ *
+ * Tests a multi-step conversation with conversation history using Google GenAI SDK
+ * and verifies that Sentry captures all spans for both API calls.
+ */
+
+const { Sentry } = require("../setup");
+const { GoogleGenAI } = require("@google/genai");
+const { runTestCase } = require("../../_test-utils/test-runner.cjs");
+
+async function testLogic(inputs) {
+  const { model, system, first_prompt, second_prompt } = inputs;
+
+  const client = new GoogleGenAI({
+    apiKey: process.env.GOOGLE_GENAI_API_KEY,
+  });
+
+  // First call
+  const firstResponse = await client.models.generateContent({
+    model,
+    contents: first_prompt,
+    config: {
+      systemInstruction: [system],
+    },
+  });
+
+  const firstText = firstResponse.text;
+
+  if (!firstText) {
+    throw new Error("No completion returned from Google GenAI (first call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    First response: ${firstText}`);
+  }
+
+  // Second call with conversation history
+  const secondResponse = await client.models.generateContent({
+    model,
+    contents: [
+      { role: "user", parts: [{ text: first_prompt }] },
+      { role: "model", parts: [{ text: firstText }] },
+      { role: "user", parts: [{ text: second_prompt }] },
+    ],
+    config: {
+      systemInstruction: [system],
+    },
+  });
+
+  const secondText = secondResponse.text;
+
+  if (!secondText) {
+    throw new Error("No completion returned from Google GenAI (second call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    Second response: ${secondText}`);
+  }
+}
+
+module.exports = runTestCase("2-multi-step", testLogic, Sentry);

sdks/js/google-genai/config.json

@@ -6,6 +6,11 @@
       "model": "gemini-2.5-flash-lite",
       "gen_ai.request.model": "gemini-2.5-flash-lite",
       "gen_ai.response.model": "gemini-2.5-flash-lite"
+    },
+    "2-multi-step": {
+      "model": "gemini-2.5-flash-lite",
+      "gen_ai.request.model": "gemini-2.5-flash-lite",
+      "gen_ai.response.model": "gemini-2.5-flash-lite"
     }
   }
 }

sdks/js/langgraph/cases/2-multi-step.js

@@ -0,0 +1,70 @@
+/**
+ * 2-multi-step: Multi-step Conversation
+ *
+ * Tests a multi-step conversation with conversation history using LangGraph SDK
+ * and verifies that Sentry captures all spans for both API calls.
+ */
+
+const { Sentry } = require("../setup");
+const { ChatOpenAI } = require("@langchain/openai");
+const { createReactAgent } = require("@langchain/langgraph/prebuilt");
+const {
+  HumanMessage,
+  SystemMessage,
+  AIMessage,
+} = require("@langchain/core/messages");
+const { runTestCase } = require("../../_test-utils/test-runner.cjs");
+
+async function testLogic(inputs) {
+  const { model, system, first_prompt, second_prompt } = inputs;
+
+  // Create LLM instance
+  const llm = new ChatOpenAI({
+    modelName: model,
+    apiKey: process.env.OPENAI_API_KEY,
+  });
+
+  // Create a simple react agent with no tools
+  const agent = createReactAgent({ llm, tools: [] });
+
+  // First call
+  const firstResult = await agent.invoke({
+    messages: [new SystemMessage(system), new HumanMessage(first_prompt)],
+  });
+
+  const firstMessages = firstResult.messages;
+  const firstLastMessage = firstMessages[firstMessages.length - 1];
+  const firstText = firstLastMessage.content;
+
+  if (!firstText) {
+    throw new Error("No completion returned from LangGraph (first call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    First response: ${firstText}`);
+  }
+
+  // Second call with conversation history
+  const secondResult = await agent.invoke({
+    messages: [
+      new SystemMessage(system),
+      new HumanMessage(first_prompt),
+      new AIMessage(firstText),
+      new HumanMessage(second_prompt),
+    ],
+  });
+
+  const secondMessages = secondResult.messages;
+  const secondLastMessage = secondMessages[secondMessages.length - 1];
+  const secondText = secondLastMessage.content;
+
+  if (!secondText) {
+    throw new Error("No completion returned from LangGraph (second call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    Second response: ${secondText}`);
+  }
+}
+
+module.exports = runTestCase("2-multi-step", testLogic, Sentry);

sdks/js/vercel/cases/2-multi-step.js

@@ -0,0 +1,55 @@
+/**
+ * 2-multi-step: Multi-step Conversation
+ *
+ * Tests a multi-step conversation with conversation history using Vercel AI SDK
+ * and verifies that Sentry captures all spans for both API calls.
+ */
+
+const { Sentry } = require("../setup");
+const { generateText } = require("ai");
+const { openai } = require("@ai-sdk/openai");
+const { runTestCase } = require("../../_test-utils/test-runner.cjs");
+
+async function testLogic(inputs) {
+  const { model, system, first_prompt, second_prompt } = inputs;
+
+  // First call
+  const firstResult = await generateText({
+    model: openai(model),
+    system,
+    prompt: first_prompt,
+  });
+
+  const firstText = firstResult.text;
+
+  if (!firstText) {
+    throw new Error("No completion returned from Vercel AI (first call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    First response: ${firstText}`);
+  }
+
+  // Second call with conversation history
+  const secondResult = await generateText({
+    model: openai(model),
+    system,
+    messages: [
+      { role: "user", content: first_prompt },
+      { role: "assistant", content: firstText },
+      { role: "user", content: second_prompt },
+    ],
+  });
+
+  const secondText = secondResult.text;
+
+  if (!secondText) {
+    throw new Error("No completion returned from Vercel AI (second call)");
+  }
+
+  if (process.env.SENTRY_AI_TEST_VERBOSE === "true") {
+    console.log(`    Second response: ${secondText}`);
+  }
+}
+
+module.exports = runTestCase("2-multi-step", testLogic, Sentry);

sdks/py/anthropic/cases/2-multi-step.py

@@ -0,0 +1,66 @@
+"""
+2-multi-step: Multi-step Conversation
+
+Tests a multi-step conversation with conversation history using Anthropic SDK
+and verifies that Sentry captures all spans for both API calls.
+"""
+
+import os
+from anthropic import Anthropic
+from test_runner import run_test_case
+
+
+async def test_logic(inputs):
+    """The actual test logic"""
+    model = inputs["model"]
+    system = inputs["system"]
+    first_prompt = inputs["first_prompt"]
+    second_prompt = inputs["second_prompt"]
+
+    client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))
+
+    # First call
+    first_response = client.messages.create(
+        model=model,
+        system=system,
+        messages=[
+            {"role": "user", "content": first_prompt}
+        ],
+        max_tokens=1024,
+    )
+
+    first_text = first_response.content[0].text
+
+    if not first_text:
+        raise Exception("No output returned from Anthropic (first call)")
+
+    if os.getenv("SENTRY_AI_TEST_VERBOSE") == "true":
+        print(f"    First response: {first_text}")
+
+    # Second call with conversation history
+    second_response = client.messages.create(
+        model=model,
+        system=system,
+        messages=[
+            {"role": "user", "content": first_prompt},
+            {"role": "assistant", "content": first_text},
+            {"role": "user", "content": second_prompt}
+        ],
+        max_tokens=1024,
+    )
+
+    second_text = second_response.content[0].text
+
+    if not second_text:
+        raise Exception("No output returned from Anthropic (second call)")
+
+    if os.getenv("SENTRY_AI_TEST_VERBOSE") == "true":
+        print(f"    Second response: {second_text}")
+
+    return second_text
+
+
+# Export test case functions
+test_case = run_test_case("2-multi-step", test_logic)
+main = test_case["main"]
+assert_sentry = test_case["assert_sentry"]

sdks/py/anthropic/config.json

@@ -6,6 +6,11 @@
       "model": "claude-haiku-4-5",
       "gen_ai.request.model": "claude-haiku-4-5",
       "gen_ai.response.model": "claude-haiku-4-5*"
+    },
+    "2-multi-step": {
+      "model": "claude-haiku-4-5",
+      "gen_ai.request.model": "claude-haiku-4-5",
+      "gen_ai.response.model": "claude-haiku-4-5*"
     }
   }
 }