feat(core): Support truncation for LangChain integration request messages

[nicohrubec]

This PR adds truncation support for LangChain integration request messages. All messages already get normalized to arrays of messages, so here we need no case distinction for strings.

Adds tests to verify behavior for 1. simple string inputs and 2. conversations in the form of arrays of strings.

Closes #18018

[github-actions]

node-overhead report 🧳

Note: This is a synthetic benchmark with a minimal express app and does not necessarily reflect the real-world performance impact in an application.

Scenario	Requests/s	% of Baseline	Prev. Requests/s	Change %
GET Baseline	8,884	-	8,762	+1%
GET With Sentry	1,412	16%	1,394	+1%
GET With Sentry (error only)	6,052	68%	6,136	-1%
POST Baseline	1,181	-	1,200	-2%
POST With Sentry	551	47%	542	+2%
POST With Sentry (error only)	1,067	90%	1,062	+0%
MYSQL Baseline	3,320	-	3,361	-1%
MYSQL With Sentry	513	15%	476	+8%
MYSQL With Sentry (error only)	2,715	82%	2,740	-1%

View base workflow run

[cursor]

Bug: LLM Prompts: Inconsistent Truncation, Exceeding Limits

Plain LLM string prompts aren't truncated when converted to messages, creating an inconsistency with chat model messages. The extractLLMRequestAttributes function wraps prompts into the same {role, content} message format as chat models but doesn't apply truncateGenAiMessages before stringifying, potentially causing large prompts to exceed byte limits while chat messages get properly truncated.

packages/core/src/utils/langchain/utils.ts#L253-L257

sentry-javascript/packages/core/src/utils/langchain/utils.ts

Lines 253 to 257 in 02d2b96

	const attrs = baseRequestAttributes(system, modelName, 'pipeline', llm, invocationParams, langSmithMetadata);
	if (recordInputs && Array.isArray(prompts) && prompts.length > 0) {
	const messages = prompts.map(p => ({ role: 'user', content: p }));
	setIfDefined(attrs, GEN_AI_REQUEST_MESSAGES_ATTRIBUTE, asString(messages));

 

---

[cursor]

Bug: Truncation Inconsistency Causes Oversized LLM Data

extractLLMRequestAttributes converts prompts to messages but doesn't apply truncation like extractChatModelRequestAttributes does. This creates inconsistent behavior where chat model messages get truncated to the byte limit but LLM string prompts don't, potentially sending oversized data that exceeds the intended 20KB limit.

packages/core/src/utils/langchain/utils.ts#L254-L259

sentry-javascript/packages/core/src/utils/langchain/utils.ts

Lines 254 to 259 in a42c548

	if (recordInputs && Array.isArray(prompts) && prompts.length > 0) {
	const messages = prompts.map(p => ({ role: 'user', content: p }));
	setIfDefined(attrs, GEN_AI_REQUEST_MESSAGES_ATTRIBUTE, asString(messages));
	}

 

---

[andreiborza]

LGTM