feat(core): simplify truncation logic to only keep the newest message

dev-packages/node-integration-tests/suites/tracing/anthropic/scenario-media-truncation.mjs

@@ -49,10 +49,15 @@ async function run() {
     const client = instrumentAnthropicAiClient(mockClient);
 
     // Send the image showing the number 3
+    // Put the image in the last message so it doesn't get dropped
     await client.messages.create({
       model: 'claude-3-haiku-20240307',
       max_tokens: 1024,
       messages: [
+        {
+          role: 'user',
+          content: 'what number is this?',
+        },
         {
           role: 'user',
           content: [
@@ -66,10 +71,6 @@ async function run() {
             },
           ],
         },
-        {
-          role: 'user',
-          content: 'what number is this?',
-        },
       ],
       temperature: 0.7,
     });

dev-packages/node-integration-tests/suites/tracing/anthropic/scenario-message-truncation.mjs

@@ -48,12 +48,11 @@ async function run() {
 
     const client = instrumentAnthropicAiClient(mockClient);
 
-    // Create 3 large messages where:
-    // - First 2 messages are very large (will be dropped)
-    // - Last message is large but will be truncated to fit within the 20KB limit
+    // Test 1: Given an array of messages only the last message should be kept
+    // The last message should be truncated to fit within the 20KB limit
     const largeContent1 = 'A'.repeat(15000); // ~15KB
     const largeContent2 = 'B'.repeat(15000); // ~15KB
-    const largeContent3 = 'C'.repeat(25000); // ~25KB (will be truncated)
+    const largeContent3 = 'C'.repeat(25000) + 'D'.repeat(25000); // ~50KB (will be truncated, only C's remain)
 
     await client.messages.create({
       model: 'claude-3-haiku-20240307',
@@ -65,6 +64,20 @@ async function run() {
       ],
       temperature: 0.7,
     });
+
+    // Test 2: Given an array of messages only the last message should be kept
+    // The last message is small, so it should be kept intact
+    const smallContent = 'This is a small message that fits within the limit';
+    await client.messages.create({
+      model: 'claude-3-haiku-20240307',
+      max_tokens: 100,
+      messages: [
+        { role: 'user', content: largeContent1 },
+        { role: 'assistant', content: largeContent2 },
+        { role: 'user', content: smallContent },
+      ],
+      temperature: 0.7,
+    });
   });
 }
 

dev-packages/node-integration-tests/suites/tracing/anthropic/test.ts

@@ -86,8 +86,7 @@ describe('Anthropic integration', () => {
         data: expect.objectContaining({
           'gen_ai.operation.name': 'messages',
           'gen_ai.request.max_tokens': 100,
-          'gen_ai.request.messages':
-            '[{"role":"system","content":"You are a helpful assistant."},{"role":"user","content":"What is the capital of France?"}]',
+          'gen_ai.request.messages': '[{"role":"user","content":"What is the capital of France?"}]',
           'gen_ai.request.model': 'claude-3-haiku-20240307',
           'gen_ai.request.temperature': 0.7,
           'gen_ai.response.id': 'msg_mock123',
@@ -638,6 +637,7 @@ describe('Anthropic integration', () => {
             transaction: {
               transaction: 'main',
               spans: expect.arrayContaining([
+                // First call: Last message is large and gets truncated (only C's remain, D's are cropped)
                 expect.objectContaining({
                   data: expect.objectContaining({
                     'gen_ai.operation.name': 'messages',
@@ -653,6 +653,24 @@ describe('Anthropic integration', () => {
                   origin: 'auto.ai.anthropic',
                   status: 'ok',
                 }),
+                // Second call: Last message is small and kept without truncation
+                expect.objectContaining({
+                  data: expect.objectContaining({
+                    'gen_ai.operation.name': 'messages',
+                    'sentry.op': 'gen_ai.messages',
+                    'sentry.origin': 'auto.ai.anthropic',
+                    'gen_ai.system': 'anthropic',
+                    'gen_ai.request.model': 'claude-3-haiku-20240307',
+                    // Small message should be kept intact
+                    'gen_ai.request.messages': JSON.stringify([
+                      { role: 'user', content: 'This is a small message that fits within the limit' },
+                    ]),
+                  }),
+                  description: 'messages claude-3-haiku-20240307',
+                  op: 'gen_ai.messages',
+                  origin: 'auto.ai.anthropic',
+                  status: 'ok',
+                }),
               ]),
             },
           })
@@ -677,6 +695,7 @@ describe('Anthropic integration', () => {
                   'sentry.origin': 'auto.ai.anthropic',
                   'gen_ai.system': 'anthropic',
                   'gen_ai.request.model': 'claude-3-haiku-20240307',
+                  // Only the last message (with filtered media) should be kept
                   'gen_ai.request.messages': JSON.stringify([
                     {
                       role: 'user',
@@ -691,10 +710,6 @@ describe('Anthropic integration', () => {
                         },
                       ],
                     },
-                    {
-                      role: 'user',
-                      content: 'what number is this?',
-                    },
                   ]),
                 }),
                 description: 'messages claude-3-haiku-20240307',

dev-packages/node-integration-tests/suites/tracing/google-genai/scenario-message-truncation.mjs

@@ -43,12 +43,11 @@ async function run() {
 
     const client = instrumentGoogleGenAIClient(mockClient);
 
-    // Create 3 large messages where:
-    // - First 2 messages are very large (will be dropped)
-    // - Last message is large but will be truncated to fit within the 20KB limit
+    // Test 1: Given an array of messages only the last message should be kept
+    // The last message should be truncated to fit within the 20KB limit
     const largeContent1 = 'A'.repeat(15000); // ~15KB
     const largeContent2 = 'B'.repeat(15000); // ~15KB
-    const largeContent3 = 'C'.repeat(25000); // ~25KB (will be truncated)
+    const largeContent3 = 'C'.repeat(25000) + 'D'.repeat(25000); // ~50KB (will be truncated, only C's remain)
 
     await client.models.generateContent({
       model: 'gemini-1.5-flash',
@@ -63,6 +62,23 @@ async function run() {
         { role: 'user', parts: [{ text: largeContent3 }] },
       ],
     });
+
+    // Test 2: Given an array of messages only the last message should be kept
+    // The last message is small, so it should be kept intact
+    const smallContent = 'This is a small message that fits within the limit';
+    await client.models.generateContent({
+      model: 'gemini-1.5-flash',
+      config: {
+        temperature: 0.7,
+        topP: 0.9,
+        maxOutputTokens: 100,
+      },
+      contents: [
+        { role: 'user', parts: [{ text: largeContent1 }] },
+        { role: 'model', parts: [{ text: largeContent2 }] },
+        { role: 'user', parts: [{ text: smallContent }] },
+      ],
+    });
   });
 }
 

dev-packages/node-integration-tests/suites/tracing/google-genai/test.ts

@@ -94,9 +94,7 @@ describe('Google GenAI integration', () => {
           'gen_ai.request.temperature': 0.8,
           'gen_ai.request.top_p': 0.9,
           'gen_ai.request.max_tokens': 150,
-          'gen_ai.request.messages': expect.stringMatching(
-            /\[\{"role":"system","content":"You are a friendly robot who likes to be funny."\},/,
-          ), // Should include history when recordInputs: true
+          'gen_ai.request.messages': '[{"role":"user","parts":[{"text":"Hello, how are you?"}]}]',
         }),
         description: 'chat gemini-1.5-pro create',
         op: 'gen_ai.chat',
@@ -504,6 +502,7 @@ describe('Google GenAI integration', () => {
             transaction: {
               transaction: 'main',
               spans: expect.arrayContaining([
+                // First call: Last message is large and gets truncated (only C's remain, D's are cropped)
                 expect.objectContaining({
                   data: expect.objectContaining({
                     'gen_ai.operation.name': 'models',
@@ -521,6 +520,27 @@ describe('Google GenAI integration', () => {
                   origin: 'auto.ai.google_genai',
                   status: 'ok',
                 }),
+                // Second call: Last message is small and kept without truncation
+                expect.objectContaining({
+                  data: expect.objectContaining({
+                    'gen_ai.operation.name': 'models',
+                    'sentry.op': 'gen_ai.models',
+                    'sentry.origin': 'auto.ai.google_genai',
+                    'gen_ai.system': 'google_genai',
+                    'gen_ai.request.model': 'gemini-1.5-flash',
+                    // Small message should be kept intact
+                    'gen_ai.request.messages': JSON.stringify([
+                      {
+                        role: 'user',
+                        parts: [{ text: 'This is a small message that fits within the limit' }],
+                      },
+                    ]),
+                  }),
+                  description: 'models gemini-1.5-flash',
+                  op: 'gen_ai.models',
+                  origin: 'auto.ai.google_genai',
+                  status: 'ok',
+                }),
               ]),
             },
           })

dev-packages/node-integration-tests/suites/tracing/langchain/scenario-message-truncation.mjs

@@ -51,17 +51,27 @@ async function run() {
 
     const largeContent1 = 'A'.repeat(15000); // ~15KB
     const largeContent2 = 'B'.repeat(15000); // ~15KB
-    const largeContent3 = 'C'.repeat(25000); // ~25KB (will be truncated)
+    const largeContent3 = 'C'.repeat(25000) + 'D'.repeat(25000); // ~50KB (will be truncated, only C's remain)
 
-    // Create one very large string that gets truncated to only include Cs
-    await model.invoke(largeContent3 + largeContent2);
+    // Test 1: Create one very large string that gets truncated to only include Cs
+    await model.invoke(largeContent3);
 
-    // Create an array of messages that gets truncated to only include the last message (result should again contain only Cs)
+    // Test 2: Create an array of messages that gets truncated to only include the last message
+    // The last message should be truncated to fit within the 20KB limit (result should again contain only Cs)
     await model.invoke([
       { role: 'system', content: largeContent1 },
       { role: 'user', content: largeContent2 },
       { role: 'user', content: largeContent3 },
     ]);
+
+    // Test 3: Given an array of messages only the last message should be kept
+    // The last message is small, so it should be kept intact
+    const smallContent = 'This is a small message that fits within the limit';
+    await model.invoke([
+      { role: 'system', content: largeContent1 },
+      { role: 'user', content: largeContent2 },
+      { role: 'user', content: smallContent },
+    ]);
   });
 
   await Sentry.flush(2000);

dev-packages/node-integration-tests/suites/tracing/langchain/test.ts

@@ -198,6 +198,7 @@ describe('LangChain integration', () => {
   const EXPECTED_TRANSACTION_MESSAGE_TRUNCATION = {
     transaction: 'main',
     spans: expect.arrayContaining([
+      // First call: String input truncated (only C's remain, D's are cropped)
       expect.objectContaining({
         data: expect.objectContaining({
           'gen_ai.operation.name': 'chat',
@@ -213,6 +214,7 @@ describe('LangChain integration', () => {
         origin: 'auto.ai.langchain',
         status: 'ok',
       }),
+      // Second call: Array input, last message truncated (only C's remain, D's are cropped)
       expect.objectContaining({
         data: expect.objectContaining({
           'gen_ai.operation.name': 'chat',
@@ -228,6 +230,24 @@ describe('LangChain integration', () => {
         origin: 'auto.ai.langchain',
         status: 'ok',
       }),
+      // Third call: Last message is small and kept without truncation
+      expect.objectContaining({
+        data: expect.objectContaining({
+          'gen_ai.operation.name': 'chat',
+          'sentry.op': 'gen_ai.chat',
+          'sentry.origin': 'auto.ai.langchain',
+          'gen_ai.system': 'anthropic',
+          'gen_ai.request.model': 'claude-3-5-sonnet-20241022',
+          // Small message should be kept intact
+          'gen_ai.request.messages': JSON.stringify([
+            { role: 'user', content: 'This is a small message that fits within the limit' },
+          ]),
+        }),
+        description: 'chat claude-3-5-sonnet-20241022',
+        op: 'gen_ai.chat',
+        origin: 'auto.ai.langchain',
+        status: 'ok',
+      }),
     ]),
   };
 

dev-packages/node-integration-tests/suites/tracing/langchain/v1/scenario-message-truncation.mjs

@@ -51,17 +51,27 @@ async function run() {
 
     const largeContent1 = 'A'.repeat(15000); // ~15KB
     const largeContent2 = 'B'.repeat(15000); // ~15KB
-    const largeContent3 = 'C'.repeat(25000); // ~25KB (will be truncated)
+    const largeContent3 = 'C'.repeat(25000) + 'D'.repeat(25000); // ~50KB (will be truncated, only C's remain)
 
-    // Create one very large string that gets truncated to only include Cs
-    await model.invoke(largeContent3 + largeContent2);
+    // Test 1: Create one very large string that gets truncated to only include Cs
+    await model.invoke(largeContent3);
 
-    // Create an array of messages that gets truncated to only include the last message (result should again contain only Cs)
+    // Test 2: Create an array of messages that gets truncated to only include the last message
+    // The last message should be truncated to fit within the 20KB limit (result should again contain only Cs)
     await model.invoke([
       { role: 'system', content: largeContent1 },
       { role: 'user', content: largeContent2 },
       { role: 'user', content: largeContent3 },
     ]);
+
+    // Test 3: Given an array of messages only the last message should be kept
+    // The last message is small, so it should be kept intact
+    const smallContent = 'This is a small message that fits within the limit';
+    await model.invoke([
+      { role: 'system', content: largeContent1 },
+      { role: 'user', content: largeContent2 },
+      { role: 'user', content: smallContent },
+    ]);
   });
 
   await Sentry.flush(2000);

dev-packages/node-integration-tests/suites/tracing/langchain/v1/test.ts

@@ -241,6 +241,7 @@ conditionalTest({ min: 20 })('LangChain integration (v1)', () => {
   const EXPECTED_TRANSACTION_MESSAGE_TRUNCATION = {
     transaction: 'main',
     spans: expect.arrayContaining([
+      // First call: String input truncated (only C's remain, D's are cropped)
       expect.objectContaining({
         data: expect.objectContaining({
           'gen_ai.operation.name': 'chat',
@@ -256,6 +257,7 @@ conditionalTest({ min: 20 })('LangChain integration (v1)', () => {
         origin: 'auto.ai.langchain',
         status: 'ok',
       }),
+      // Second call: Array input, last message truncated (only C's remain, D's are cropped)
       expect.objectContaining({
         data: expect.objectContaining({
           'gen_ai.operation.name': 'chat',
@@ -271,6 +273,24 @@ conditionalTest({ min: 20 })('LangChain integration (v1)', () => {
         origin: 'auto.ai.langchain',
         status: 'ok',
       }),
+      // Third call: Last message is small and kept without truncation
+      expect.objectContaining({
+        data: expect.objectContaining({
+          'gen_ai.operation.name': 'chat',
+          'sentry.op': 'gen_ai.chat',
+          'sentry.origin': 'auto.ai.langchain',
+          'gen_ai.system': 'anthropic',
+          'gen_ai.request.model': 'claude-3-5-sonnet-20241022',
+          // Small message should be kept intact
+          'gen_ai.request.messages': JSON.stringify([
+            { role: 'user', content: 'This is a small message that fits within the limit' },
+          ]),
+        }),
+        description: 'chat claude-3-5-sonnet-20241022',
+        op: 'gen_ai.chat',
+        origin: 'auto.ai.langchain',
+        status: 'ok',
+      }),
     ]),
   };
 

dev-packages/node-integration-tests/suites/tracing/openai/scenario-embeddings.mjs

@@ -67,6 +67,12 @@ async function run() {
     } catch {
       // Error is expected and handled
     }
+
+    // Third test: embeddings API with multiple inputs
+    await client.embeddings.create({
+      input: ['First input text', 'Second input text', 'Third input text'],
+      model: 'text-embedding-3-small',
+    });
   });
 
   server.close();