Skip to content

fix(core): Fix truncation to only keep last message in vercel #19080

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
nicohrubec merged 16 commits into from
Feb 9, 2026
Merged

fix(core): Fix truncation to only keep last message in vercel #19080

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
a0fabf0
fix(core): Fix truncation to only keep last message in vercel
nicohrubec Jan 28, 2026
a14a741
no longer set gen_ai.prompt in vercel
nicohrubec Jan 29, 2026
0a8a976
clean
nicohrubec Jan 29, 2026
c1d8eea
handle messages array input
nicohrubec Jan 29, 2026
85f963b
truncate ai.prompt as well
nicohrubec Jan 29, 2026
d746121
fix tests
nicohrubec Jan 29, 2026
ed69ebd
let's try again
nicohrubec Jan 29, 2026
93e43f3
remove cast and handle system instructions along with messages array
nicohrubec Feb 2, 2026
94e70da
adjust size limit
nicohrubec Feb 2, 2026
be3b7cd
be safe
nicohrubec Feb 2, 2026
50fcded
rename prompt variable + conversion method and add more comments
nicohrubec Feb 5, 2026
df96710
Prompt can be an array
nicohrubec Feb 5, 2026
d81a9a5
clean
nicohrubec Feb 5, 2026
a049b0f
add unit tests
nicohrubec Feb 5, 2026
ffb7817
yarn fix
nicohrubec Feb 5, 2026
af9bd7e
Merge branch 'develop' into nh/fix-vercel-message-truncation
nicohrubec Feb 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 51 additions & 0 deletions dev-packages/node-integration-tests/suites/tracing/vercelai/scenario-message-truncation.mjs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,51 @@
import * as Sentry from '@sentry/node';
import { generateText } from 'ai';
import { MockLanguageModelV1 } from 'ai/test';

async function run() {
await Sentry.startSpan({ op: 'function', name: 'main' }, async () => {
const largeContent1 = 'A'.repeat(15000); // ~15KB
const largeContent2 = 'B'.repeat(15000); // ~15KB
const largeContent3 = 'C'.repeat(25000) + 'D'.repeat(25000); // ~50KB (will be truncated)

// Test 1: Messages array with large last message that gets truncated
// Only the last message should be kept, and it should be truncated to only Cs
await generateText({
experimental_telemetry: { isEnabled: true },
model: new MockLanguageModelV1({
doGenerate: async () => ({
rawCall: { rawPrompt: null, rawSettings: {} },
finishReason: 'stop',
usage: { promptTokens: 10, completionTokens: 5 },
text: 'Response to truncated messages',
}),
}),
messages: [
{ role: 'user', content: largeContent1 },
{ role: 'assistant', content: largeContent2 },
{ role: 'user', content: largeContent3 },
],
});

// Test 2: Messages array where last message is small and kept intact
const smallContent = 'This is a small message that fits within the limit';
await generateText({
experimental_telemetry: { isEnabled: true },
model: new MockLanguageModelV1({
doGenerate: async () => ({
rawCall: { rawPrompt: null, rawSettings: {} },
finishReason: 'stop',
usage: { promptTokens: 10, completionTokens: 5 },
text: 'Response to small message',
}),
}),
messages: [
{ role: 'user', content: largeContent1 },
{ role: 'assistant', content: largeContent2 },
{ role: 'user', content: smallContent },
],
});
});
}

run();
52 changes: 43 additions & 9 deletions dev-packages/node-integration-tests/suites/tracing/vercelai/test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@ import {
GEN_AI_INPUT_MESSAGES_ATTRIBUTE,
GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE,
GEN_AI_OPERATION_NAME_ATTRIBUTE,
GEN_AI_PROMPT_ATTRIBUTE,
GEN_AI_REQUEST_AVAILABLE_TOOLS_ATTRIBUTE,
GEN_AI_REQUEST_MODEL_ATTRIBUTE,
GEN_AI_RESPONSE_FINISH_REASONS_ATTRIBUTE,
Expand Down Expand Up @@ -90,7 +89,6 @@ describe('Vercel AI integration', () => {
// Third span - explicit telemetry enabled, should record inputs/outputs regardless of sendDefaultPii
expect.objectContaining({
data: {
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
[GEN_AI_REQUEST_MODEL_ATTRIBUTE]: 'mock-model-id',
Expand All @@ -105,7 +103,7 @@ describe('Vercel AI integration', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.response.finishReason': 'stop',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.settings.maxSteps': 1,
Expand Down Expand Up @@ -230,7 +228,6 @@ describe('Vercel AI integration', () => {
// First span - no telemetry config, should enable telemetry AND record inputs/outputs when sendDefaultPii: true
expect.objectContaining({
data: {
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the first span?"}',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the first span?"}]',
[GEN_AI_REQUEST_MODEL_ATTRIBUTE]: 'mock-model-id',
Expand All @@ -245,7 +242,7 @@ describe('Vercel AI integration', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the first span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the first span?"}]',
'vercel.ai.response.finishReason': 'stop',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.settings.maxSteps': 1,
Expand Down Expand Up @@ -303,7 +300,6 @@ describe('Vercel AI integration', () => {
// Third span - explicitly enabled telemetry, should record inputs/outputs regardless of sendDefaultPii
expect.objectContaining({
data: {
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
[GEN_AI_REQUEST_MODEL_ATTRIBUTE]: 'mock-model-id',
Expand All @@ -318,7 +314,7 @@ describe('Vercel AI integration', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.response.finishReason': 'stop',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.settings.maxSteps': 1,
Expand Down Expand Up @@ -375,7 +371,6 @@ describe('Vercel AI integration', () => {
// Fifth span - tool call generateText span (should include prompts when sendDefaultPii: true)
expect.objectContaining({
data: {
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"What is the weather in San Francisco?"}',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"What is the weather in San Francisco?"}]',
[GEN_AI_REQUEST_MODEL_ATTRIBUTE]: 'mock-model-id',
Expand All @@ -391,7 +386,7 @@ describe('Vercel AI integration', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"What is the weather in San Francisco?"}',
'vercel.ai.prompt': '[{"role":"user","content":"What is the weather in San Francisco?"}]',
'vercel.ai.response.finishReason': 'tool-calls',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.settings.maxSteps': 1,
Expand Down Expand Up @@ -796,4 +791,43 @@ describe('Vercel AI integration', () => {
});
},
);

createEsmAndCjsTests(
__dirname,
'scenario-message-truncation.mjs',
'instrument-with-pii.mjs',
(createRunner, test) => {
test('truncates messages when they exceed byte limit', async () => {
await createRunner()
.ignore('event')
.expect({
transaction: {
transaction: 'main',
spans: expect.arrayContaining([
// First call: Last message truncated (only C's remain, D's are cropped)
expect.objectContaining({
data: expect.objectContaining({
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 3,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: expect.stringMatching(/^\[.*"(?:text|content)":"C+".*\]$/),
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: 'Response to truncated messages',
}),
}),
// Second call: Last message is small and kept intact
expect.objectContaining({
data: expect.objectContaining({
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 3,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: expect.stringContaining(
'This is a small message that fits within the limit',
),
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: 'Response to small message',
}),
}),
]),
},
})
.start()
.completed();
});
},
);
});
13 changes: 4 additions & 9 deletions dev-packages/node-integration-tests/suites/tracing/vercelai/v5/test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,6 @@ import {
GEN_AI_INPUT_MESSAGES_ATTRIBUTE,
GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE,
GEN_AI_OPERATION_NAME_ATTRIBUTE,
GEN_AI_PROMPT_ATTRIBUTE,
GEN_AI_REQUEST_AVAILABLE_TOOLS_ATTRIBUTE,
GEN_AI_REQUEST_MODEL_ATTRIBUTE,
GEN_AI_RESPONSE_FINISH_REASONS_ATTRIBUTE,
Expand Down Expand Up @@ -92,12 +91,11 @@ describe('Vercel AI integration (V5)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
Comment thread
nicohrubec marked this conversation as resolved.
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
Expand Down Expand Up @@ -229,14 +227,13 @@ describe('Vercel AI integration (V5)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the first span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the first span?"}]',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the first span?"}]',
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: 'First span here!',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the first span?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 10,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 20,
Expand Down Expand Up @@ -290,14 +287,13 @@ describe('Vercel AI integration (V5)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 10,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 20,
Expand Down Expand Up @@ -350,14 +346,13 @@ describe('Vercel AI integration (V5)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"What is the weather in San Francisco?"}',
'vercel.ai.prompt': '[{"role":"user","content":"What is the weather in San Francisco?"}]',
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: 1,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"What is the weather in San Francisco?"}]',
'vercel.ai.response.finishReason': 'tool-calls',
[GEN_AI_RESPONSE_TOOL_CALLS_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"What is the weather in San Francisco?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 15,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 25,
Expand Down
13 changes: 4 additions & 9 deletions dev-packages/node-integration-tests/suites/tracing/vercelai/v6/test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,6 @@ import { afterAll, describe, expect } from 'vitest';
import {
GEN_AI_INPUT_MESSAGES_ATTRIBUTE,
GEN_AI_OPERATION_NAME_ATTRIBUTE,
GEN_AI_PROMPT_ATTRIBUTE,
GEN_AI_REQUEST_AVAILABLE_TOOLS_ATTRIBUTE,
GEN_AI_REQUEST_MODEL_ATTRIBUTE,
GEN_AI_RESPONSE_FINISH_REASONS_ATTRIBUTE,
Expand Down Expand Up @@ -93,13 +92,12 @@ describe('Vercel AI integration (V6)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.request.headers.user-agent': expect.any(String),
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 10,
Expand Down Expand Up @@ -232,14 +230,13 @@ describe('Vercel AI integration (V6)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the first span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the first span?"}]',
'vercel.ai.request.headers.user-agent': expect.any(String),
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the first span?"}]',
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: 'First span here!',
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the first span?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 10,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 20,
Expand Down Expand Up @@ -293,14 +290,13 @@ describe('Vercel AI integration (V6)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"Where is the second span?"}',
'vercel.ai.prompt': '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.request.headers.user-agent': expect.any(String),
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"Where is the second span?"}]',
'vercel.ai.response.finishReason': 'stop',
[GEN_AI_RESPONSE_TEXT_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"Where is the second span?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 10,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 20,
Expand Down Expand Up @@ -353,14 +349,13 @@ describe('Vercel AI integration (V6)', () => {
'vercel.ai.model.provider': 'mock-provider',
'vercel.ai.operationId': 'ai.generateText',
'vercel.ai.pipeline.name': 'generateText',
'vercel.ai.prompt': '{"prompt":"What is the weather in San Francisco?"}',
'vercel.ai.prompt': '[{"role":"user","content":"What is the weather in San Francisco?"}]',
'vercel.ai.request.headers.user-agent': expect.any(String),
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: '[{"role":"user","content":"What is the weather in San Francisco?"}]',
'vercel.ai.response.finishReason': 'tool-calls',
[GEN_AI_RESPONSE_TOOL_CALLS_ATTRIBUTE]: expect.any(String),
'vercel.ai.settings.maxRetries': 2,
'vercel.ai.streaming': false,
[GEN_AI_PROMPT_ATTRIBUTE]: '{"prompt":"What is the weather in San Francisco?"}',
[GEN_AI_RESPONSE_MODEL_ATTRIBUTE]: 'mock-model-id',
[GEN_AI_USAGE_INPUT_TOKENS_ATTRIBUTE]: 15,
[GEN_AI_USAGE_OUTPUT_TOKENS_ATTRIBUTE]: 25,
Expand Down
31 changes: 22 additions & 9 deletions packages/core/src/tracing/vercel-ai/utils.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -111,6 +111,16 @@ export function convertPromptToMessages(prompt: string): { role: string; content
try {
const p = JSON.parse(prompt);
if (!!p && typeof p === 'object') {
// Handle messages array format: { messages: [...] }
const { messages } = p as { messages?: unknown };
Copy link
Copy Markdown
Contributor

@RulaKhaled RulaKhaled Feb 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

m-h: Function expects prompt to be of type string, but we cast it here to unknown, which could cause some unexpected behavior.

q: is it ever a stringified messages array?

Copy link
Copy Markdown
Member Author

@nicohrubec nicohrubec Feb 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I removed the cast and reorganized this method a bit because I noticed that in the previous version I would not add the system instruction in case we get a messages array

Copy link
Copy Markdown
Member Author

@nicohrubec nicohrubec Feb 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think it should be a stringified messages array at that point but I added an additional explicit check to be safe

if (Array.isArray(messages)) {
return messages.filter(
(m: unknown): m is { role: string; content: string } =>
!!m && typeof m === 'object' && 'role' in m && 'content' in m,
);
}

// Handle prompt/system string format: { prompt: "...", system: "..." }
const { prompt, system } = p;
if (typeof prompt === 'string' || typeof system === 'string') {
const messages: { role: string; content: string }[] = [];
Expand All @@ -133,16 +143,13 @@ export function convertPromptToMessages(prompt: string): { role: string; content
* invoke_agent op
*/
export function requestMessagesFromPrompt(span: Span, attributes: SpanAttributes): void {
if (attributes[AI_PROMPT_ATTRIBUTE]) {
const truncatedPrompt = getTruncatedJsonString(attributes[AI_PROMPT_ATTRIBUTE] as string | string[]);
span.setAttribute('gen_ai.prompt', truncatedPrompt);
}
const prompt = attributes[AI_PROMPT_ATTRIBUTE];
if (
typeof prompt === 'string' &&
typeof attributes[AI_PROMPT_ATTRIBUTE] === 'string' &&
!attributes[GEN_AI_INPUT_MESSAGES_ATTRIBUTE] &&
!attributes[AI_PROMPT_MESSAGES_ATTRIBUTE]
) {
// No messages array is present, so we need to convert the prompt to the proper messages format
const prompt = attributes[AI_PROMPT_ATTRIBUTE];
const messages = convertPromptToMessages(prompt);
if (messages.length) {
const { systemInstructions, filteredMessages } = extractSystemInstructions(messages);
Expand All @@ -152,12 +159,16 @@ export function requestMessagesFromPrompt(span: Span, attributes: SpanAttributes
}

const filteredLength = Array.isArray(filteredMessages) ? filteredMessages.length : 0;
const truncatedMessages = getTruncatedJsonString(filteredMessages);

span.setAttributes({
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: getTruncatedJsonString(filteredMessages),
[AI_PROMPT_ATTRIBUTE]: truncatedMessages,
Copy link
Copy Markdown
Contributor

@RulaKhaled RulaKhaled Feb 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i thought we're going to git rid of AI_PROMPT_ATTRIBUTE as it's deprecated, right?

Copy link
Copy Markdown
Member Author

@nicohrubec nicohrubec Feb 2, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we are getting rid of GEN_AI_PROMPT_ATTRIBUTE in our namespace, this is the original Vercel attribute. If we don't overwrite it here we send the full non-truncated message

[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: truncatedMessages,
Comment thread
nicohrubec marked this conversation as resolved.
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: filteredLength,
});
}
} else if (typeof attributes[AI_PROMPT_MESSAGES_ATTRIBUTE] === 'string') {
// In this case we already get a properly formatted messages array, this is the preferred way to get the messages
try {
const messages = JSON.parse(attributes[AI_PROMPT_MESSAGES_ATTRIBUTE]);
if (Array.isArray(messages)) {
Expand All @@ -168,9 +179,11 @@ export function requestMessagesFromPrompt(span: Span, attributes: SpanAttributes
}

const filteredLength = Array.isArray(filteredMessages) ? filteredMessages.length : 0;
const truncatedMessages = getTruncatedJsonString(filteredMessages);

span.setAttributes({
[AI_PROMPT_MESSAGES_ATTRIBUTE]: undefined,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: getTruncatedJsonString(filteredMessages),
[AI_PROMPT_MESSAGES_ATTRIBUTE]: truncatedMessages,
[GEN_AI_INPUT_MESSAGES_ATTRIBUTE]: truncatedMessages,
[GEN_AI_INPUT_MESSAGES_ORIGINAL_LENGTH_ATTRIBUTE]: filteredLength,
});
Comment thread
sentry[bot] marked this conversation as resolved.
}
Expand Down
Loading
Loading