fix(llmobs): set default model provider to unknown in openai and anthropic

packages/datadog-instrumentations/src/anthropic.js

@@ -41,7 +41,7 @@ function wrapCreate (create) {
     const options = arguments[0]
     const stream = options.stream
 
-    const ctx = { options, resource: 'create' }
+    const ctx = { options, resource: 'create', baseUrl: this._client?.baseURL }
 
     return anthropicTracingChannel.start.runStores(ctx, () => {
       let apiPromise

packages/dd-trace/src/llmobs/constants/tags.js

@@ -23,6 +23,7 @@ module.exports = {
 
   MODEL_NAME: '_ml_obs.meta.model_name',
   MODEL_PROVIDER: '_ml_obs.meta.model_provider',
+  UNKNOWN_MODEL_PROVIDER: 'unknown',
 
   INPUT_DOCUMENTS: '_ml_obs.meta.input.documents',
   INPUT_MESSAGES: '_ml_obs.meta.input.messages',

packages/dd-trace/src/llmobs/plugins/anthropic.js

@@ -1,5 +1,6 @@
 'use strict'
 
+const { UNKNOWN_MODEL_PROVIDER } = require('../constants/tags')
 const LLMObsPlugin = require('./base')
 
 const ALLOWED_METADATA_KEYS = new Set([
@@ -108,16 +109,24 @@ class AnthropicLLMObsPlugin extends LLMObsPlugin {
   }
 
   getLLMObsSpanRegisterOptions (ctx) {
-    const { options } = ctx
+    const { options, baseUrl } = ctx
     const { model } = options
+    const modelProvider = this._getModelProvider(baseUrl)
 
     return {
       kind: 'llm',
       modelName: model,
-      modelProvider: 'anthropic',
+      modelProvider,
     }
   }
 
+  _getModelProvider (baseUrl = '') {
+    if (baseUrl.includes('anthropic')) {
+      return 'anthropic'
+    }
+    return UNKNOWN_MODEL_PROVIDER
+  }
+
   setLLMObsTags (ctx) {
     const span = ctx.currentStore?.span
     if (!span) return

packages/dd-trace/src/llmobs/plugins/openai/index.js

@@ -5,6 +5,7 @@ const {
   PROMPT_TRACKING_INSTRUMENTATION_METHOD,
   PROMPT_MULTIMODAL,
   INSTRUMENTATION_METHOD_AUTO,
+  UNKNOWN_MODEL_PROVIDER,
 } = require('../../constants/tags')
 const {
   extractChatTemplateFromInstructions,
@@ -90,10 +91,12 @@ class OpenAiLLMObsPlugin extends LLMObsPlugin {
   _getModelProviderAndClient (baseUrl = '') {
     if (baseUrl.includes('azure')) {
       return { modelProvider: 'azure_openai', client: 'AzureOpenAI' }
+    } else if (baseUrl.includes('openai')) {
+      return { modelProvider: 'openai', client: 'OpenAI' }
     } else if (baseUrl.includes('deepseek')) {
       return { modelProvider: 'deepseek', client: 'DeepSeek' }
     }
-    return { modelProvider: 'openai', client: 'OpenAI' }
+    return { modelProvider: UNKNOWN_MODEL_PROVIDER, client: 'OpenAI' }
   }
 
   _extractMetrics (response) {

packages/dd-trace/test/llmobs/plugins/anthropic/index.spec.js

@@ -69,6 +69,26 @@ describe('Plugin', () => {
         assertLLMObsSpan(apmSpans, llmobsSpans)
       })
 
+      it('sets model_provider to unknown for unrecognized base URLs', async () => {
+        const { Anthropic } = require(`../../../../../../versions/@anthropic-ai/sdk@${version}`).get()
+        const customClient = new Anthropic({ baseURL: 'http://localhost:8000' })
+
+        try {
+          await customClient.messages.create({
+            model: 'claude-3-7-sonnet-20250219',
+            messages: [{ role: 'user', content: 'Hello, world!' }],
+            max_tokens: 100,
+            temperature: 0.5,
+          })
+        } catch {
+          // expected error — no server is running
+        }
+
+        const { llmobsSpans } = await getEvents()
+
+        assert.equal(llmobsSpans[0].meta.model_provider, 'unknown', 'Model provider does not match')
+      })
+
       describe('stream', () => {
         it('creates a span', async () => {
           const stream = await client.messages.create({

packages/dd-trace/test/llmobs/plugins/openai/openaiv4.spec.js

@@ -575,6 +575,31 @@ describe('integrations', () => {
         assert.equal(llmobsSpans[0].meta.model_provider, 'deepseek', 'Model provider does not match')
       })
 
+      it('sets model_provider to unknown for unrecognized base URLs', async () => {
+        const OpenAI = require(moduleRequirePath).get()
+        const customClient = new OpenAI({
+          apiKey: 'test',
+          baseURL: 'http://localhost:8000',
+        })
+
+        try {
+          await customClient.chat.completions.create({
+            model: 'gpt-3.5-turbo',
+            messages: [
+              { role: 'user', content: 'Hello, OpenAI!' },
+            ],
+            temperature: 0.5,
+            max_tokens: 100,
+          })
+        } catch {
+          // expected error — no server is running
+        }
+
+        const { llmobsSpans } = await getEvents()
+
+        assert.equal(llmobsSpans[0].meta.model_provider, 'unknown', 'Model provider does not match')
+      })
+
       it('submits a chat completion span with cached token metrics', async () => {
         const baseMessages = [{ role: 'system', content: 'You are an expert software engineer '.repeat(200) }]