feat(llmobs): add reasoning token metrics in openai plugin

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

@@ -34,6 +34,7 @@ module.exports = {
   TOTAL_TOKENS_METRIC_KEY: 'total_tokens',
   CACHE_READ_INPUT_TOKENS_METRIC_KEY: 'cache_read_input_tokens',
   CACHE_WRITE_INPUT_TOKENS_METRIC_KEY: 'cache_write_input_tokens',
+  REASONING_OUTPUT_TOKENS_METRIC_KEY: 'reasoning_output_tokens',
 
   DROPPED_IO_COLLECTION_ERROR: 'dropped_io'
 }

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

@@ -116,6 +116,10 @@ class OpenAiLLMObsPlugin extends LLMObsPlugin {
           metrics.cacheReadTokens = cacheReadTokens
         }
       }
+      // Reasoning tokens - Responses API returns `output_tokens_details`, `completion_tokens_details`
+      const reasoningOutputObject = tokenUsage.output_tokens_details ?? tokenUsage.completion_tokens_details
+      const reasoningOutputTokens = reasoningOutputObject?.reasoning_tokens ?? 0
+      if (reasoningOutputTokens !== undefined) metrics.reasoningOutputTokens = reasoningOutputTokens
     }
 
     return metrics
@@ -429,9 +433,6 @@ class OpenAiLLMObsPlugin extends LLMObsPlugin {
     if (response.tool_choice !== undefined) outputMetadata.tool_choice = response.tool_choice
     if (response.truncation !== undefined) outputMetadata.truncation = response.truncation
     if (response.text !== undefined) outputMetadata.text = response.text
-    if (response.usage?.output_tokens_details?.reasoning_tokens !== undefined) {
-      outputMetadata.reasoning_tokens = response.usage.output_tokens_details.reasoning_tokens
-    }
 
     this._tagger.tagMetadata(span, outputMetadata) // update the metadata with the output metadata
   }

packages/dd-trace/src/llmobs/tagger.js

@@ -25,6 +25,7 @@ const {
   INPUT_TOKENS_METRIC_KEY,
   OUTPUT_TOKENS_METRIC_KEY,
   TOTAL_TOKENS_METRIC_KEY,
+  REASONING_OUTPUT_TOKENS_METRIC_KEY,
   INTEGRATION,
   DECORATOR,
   PROPAGATED_ML_APP_KEY
@@ -164,6 +165,9 @@ class LLMObsTagger {
         case 'cacheWriteTokens':
           processedKey = CACHE_WRITE_INPUT_TOKENS_METRIC_KEY
           break
+        case 'reasoningOutputTokens':
+          processedKey = REASONING_OUTPUT_TOKENS_METRIC_KEY
+          break
       }
 
       if (typeof value === 'number') {

packages/dd-trace/test/llmobs/cassettes/openai/openai_responses_post_01aba65f.yaml

@@ -0,0 +1,135 @@
+interactions:
+- request:
+    body: '{"model":"gpt-5-mini","input":"Solve this step by step: What is 15 * 24?","max_output_tokens":500,"stream":false}'
+    headers:
+      ? !!python/object/apply:multidict._multidict.istr
+      - Accept
+      : - application/json
+      ? !!python/object/apply:multidict._multidict.istr
+      - Accept-Encoding
+      : - gzip, deflate
+      ? !!python/object/apply:multidict._multidict.istr
+      - Accept-Language
+      : - '*'
+      ? !!python/object/apply:multidict._multidict.istr
+      - Connection
+      : - keep-alive
+      Content-Length:
+      - '113'
+      ? !!python/object/apply:multidict._multidict.istr
+      - Content-Type
+      : - application/json
+      ? !!python/object/apply:multidict._multidict.istr
+      - User-Agent
+      : - OpenAI/JS 6.8.1
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Arch
+      : - arm64
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Lang
+      : - js
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-OS
+      : - MacOS
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Package-Version
+      : - 6.8.1
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Retry-Count
+      : - '0'
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Runtime
+      : - node
+      ? !!python/object/apply:multidict._multidict.istr
+      - X-Stainless-Runtime-Version
+      : - v22.21.1
+      ? !!python/object/apply:multidict._multidict.istr
+      - sec-fetch-mode
+      : - cors
+    method: POST
+    uri: https://api.openai.com/v1/responses
+  response:
+    body:
+      string: "{\n  \"id\": \"resp_0e28e5318868c83a0169308b613bdc819184e3bca92ada65e6\",\n
+        \ \"object\": \"response\",\n  \"created_at\": 1764789089,\n  \"status\":
+        \"completed\",\n  \"background\": false,\n  \"billing\": {\n    \"payer\":
+        \"developer\"\n  },\n  \"error\": null,\n  \"incomplete_details\": null,\n
+        \ \"instructions\": null,\n  \"max_output_tokens\": 500,\n  \"max_tool_calls\":
+        null,\n  \"model\": \"gpt-5-mini-2025-08-07\",\n  \"output\": [\n    {\n      \"id\":
+        \"rs_0e28e5318868c83a0169308b617fa08191b6726155d23d010d\",\n      \"type\":
+        \"reasoning\",\n      \"summary\": []\n    },\n    {\n      \"id\": \"msg_0e28e5318868c83a0169308b63623c819198d6e2c9b0eb26b0\",\n
+        \     \"type\": \"message\",\n      \"status\": \"completed\",\n      \"content\":
+        [\n        {\n          \"type\": \"output_text\",\n          \"annotations\":
+        [],\n          \"logprobs\": [],\n          \"text\": \"Method 1 \\u2014 distributive
+        property:\\n15 \\u00d7 24 = 15 \\u00d7 (20 + 4)\\n= 15 \\u00d7 20 + 15 \\u00d7
+        4\\n= 300 + 60\\n= 360\\n\\nMethod 2 \\u2014 split the other way:\\n24 \\u00d7
+        15 = 24 \\u00d7 (10 + 5)\\n= 24 \\u00d7 10 + 24 \\u00d7 5\\n= 240 + 120\\n=
+        360\\n\\nAnswer: 360.\"\n        }\n      ],\n      \"role\": \"assistant\"\n
+        \   }\n  ],\n  \"parallel_tool_calls\": true,\n  \"previous_response_id\":
+        null,\n  \"prompt_cache_key\": null,\n  \"prompt_cache_retention\": null,\n
+        \ \"reasoning\": {\n    \"effort\": \"medium\",\n    \"summary\": null\n  },\n
+        \ \"safety_identifier\": null,\n  \"service_tier\": \"default\",\n  \"store\":
+        false,\n  \"temperature\": 1.0,\n  \"text\": {\n    \"format\": {\n      \"type\":
+        \"text\"\n    },\n    \"verbosity\": \"medium\"\n  },\n  \"tool_choice\":
+        \"auto\",\n  \"tools\": [],\n  \"top_logprobs\": 0,\n  \"top_p\": 1.0,\n  \"truncation\":
+        \"disabled\",\n  \"usage\": {\n    \"input_tokens\": 20,\n    \"input_tokens_details\":
+        {\n      \"cached_tokens\": 0\n    },\n    \"output_tokens\": 232,\n    \"output_tokens_details\":
+        {\n      \"reasoning_tokens\": 128\n    },\n    \"total_tokens\": 252\n  },\n
+        \ \"user\": null,\n  \"metadata\": {}\n}"
+    headers:
+      CF-RAY:
+      - 9a855ebe0d40cb6a-BOS
+      Connection:
+      - keep-alive
+      Content-Encoding:
+      - gzip
+      Content-Type:
+      - application/json
+      Date:
+      - Wed, 03 Dec 2025 19:11:33 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - __cf_bm=tGToqitV1eFrOrI1DCsm6GfeeW0ajMWiagCcgm7EZ84-1764789093-1.0.1.1-d.Bwuc72eKGJd7bZL0hQRAOgBqhp15Rk0H9FzUo.0s8hqVVPBeKHE39I5EaaTi2YZdtQyCFyHmd3iILpZJKskSfEdJ3MtfEHr_4Ktk7yDw8;
+        path=/; expires=Wed, 03-Dec-25 19:41:33 GMT; domain=.api.openai.com; HttpOnly;
+        Secure; SameSite=None
+      - _cfuvid=NO0A21ruRpV_Vd_9ghyhNxI_FfJlolr0EqD7FHB3tOs-1764789093127-0.0.1.1-604800000;
+        path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
+      Strict-Transport-Security:
+      - max-age=31536000; includeSubDomains; preload
+      Transfer-Encoding:
+      - chunked
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      openai-organization:
+      - datadog-staging
+      openai-processing-ms:
+      - '3895'
+      openai-project:
+      - proj_gt6TQZPRbZfoY2J9AQlEJMpd
+      openai-version:
+      - '2020-10-01'
+      x-envoy-upstream-service-time:
+      - '3900'
+      x-ratelimit-limit-requests:
+      - '30000'
+      x-ratelimit-limit-tokens:
+      - '180000000'
+      x-ratelimit-remaining-requests:
+      - '29999'
+      x-ratelimit-remaining-tokens:
+      - '179999985'
+      x-ratelimit-reset-requests:
+      - 2ms
+      x-ratelimit-reset-tokens:
+      - 0s
+      x-request-id:
+      - req_d13dea53ffc94a628dfa63719a5bcbb9
+    status:
+      code: 200
+      message: OK
+version: 1

packages/dd-trace/test/llmobs/plugins/openai/openaiv3.spec.js

@@ -61,7 +61,8 @@ describe('integrations', () => {
           metrics: {
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
-            total_tokens: MOCK_NUMBER
+            total_tokens: MOCK_NUMBER,
+            reasoning_output_tokens: 0
           },
           modelName: 'gpt-3.5-turbo-instruct:20230824-v2',
           modelProvider: 'openai',
@@ -113,6 +114,7 @@ describe('integrations', () => {
           ],
           metrics: {
             cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0,
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
             total_tokens: MOCK_NUMBER
@@ -146,7 +148,9 @@ describe('integrations', () => {
             { text: 'hello world' }
           ],
           outputValue: '[1 embedding(s) returned]',
-          metrics: { input_tokens: MOCK_NUMBER, output_tokens: 0, total_tokens: MOCK_NUMBER },
+          metrics: {
+            input_tokens: MOCK_NUMBER, output_tokens: 0, total_tokens: MOCK_NUMBER, reasoning_output_tokens: 0
+          },
           modelName: 'text-embedding-ada-002-v2',
           modelProvider: 'openai',
           metadata: { encoding_format: 'base64' },

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

@@ -81,7 +81,9 @@ describe('integrations', () => {
           outputMessages: [
             { content: MOCK_STRING, role: '' }
           ],
-          metrics: { input_tokens: MOCK_NUMBER, output_tokens: MOCK_NUMBER, total_tokens: MOCK_NUMBER },
+          metrics: {
+            input_tokens: MOCK_NUMBER, output_tokens: MOCK_NUMBER, total_tokens: MOCK_NUMBER, reasoning_output_tokens: 0
+          },
           modelName: 'gpt-3.5-turbo-instruct:20230824-v2',
           modelProvider: 'openai',
           metadata: {
@@ -128,6 +130,7 @@ describe('integrations', () => {
           ],
           metrics: {
             cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0,
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
             total_tokens: MOCK_NUMBER
@@ -161,7 +164,9 @@ describe('integrations', () => {
             { text: 'hello world' }
           ],
           outputValue: '[1 embedding(s) returned]',
-          metrics: { input_tokens: MOCK_NUMBER, output_tokens: 0, total_tokens: MOCK_NUMBER },
+          metrics: {
+            input_tokens: MOCK_NUMBER, output_tokens: 0, total_tokens: MOCK_NUMBER, reasoning_output_tokens: 0
+          },
           modelName: 'text-embedding-ada-002-v2',
           modelProvider: 'openai',
           metadata: { encoding_format: 'base64' },
@@ -220,6 +225,7 @@ describe('integrations', () => {
           tags: { ml_app: 'test', integration: 'openai' },
           metrics: {
             cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0,
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
             total_tokens: MOCK_NUMBER
@@ -268,7 +274,12 @@ describe('integrations', () => {
             outputMessages: [
               { content: '\n\nHello! How can I assist you?', role: '' }
             ],
-            metrics: { input_tokens: MOCK_NUMBER, output_tokens: MOCK_NUMBER, total_tokens: MOCK_NUMBER },
+            metrics: {
+              input_tokens: MOCK_NUMBER,
+              output_tokens: MOCK_NUMBER,
+              total_tokens: MOCK_NUMBER,
+              reasoning_output_tokens: 0
+            },
             modelName: 'gpt-3.5-turbo-instruct:20230824-v2',
             modelProvider: 'openai',
             metadata: {
@@ -329,6 +340,7 @@ describe('integrations', () => {
             ],
             metrics: {
               cache_read_input_tokens: 0,
+              reasoning_output_tokens: 0,
               input_tokens: MOCK_NUMBER,
               output_tokens: MOCK_NUMBER,
               total_tokens: MOCK_NUMBER
@@ -413,6 +425,7 @@ describe('integrations', () => {
             tags: { ml_app: 'test', integration: 'openai' },
             metrics: {
               cache_read_input_tokens: 0,
+              reasoning_output_tokens: 0,
               input_tokens: MOCK_NUMBER,
               output_tokens: MOCK_NUMBER,
               total_tokens: MOCK_NUMBER
@@ -599,6 +612,7 @@ describe('integrations', () => {
           ],
           metrics: {
             cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0,
             input_tokens: 1221,
             output_tokens: 100,
             total_tokens: 1321
@@ -647,6 +661,7 @@ describe('integrations', () => {
             output_tokens: 100,
             total_tokens: 1320,
             cache_read_input_tokens: 1152,
+            reasoning_output_tokens: 0
           },
           modelName: 'gpt-4o-2024-08-06',
           modelProvider: 'openai',
@@ -689,7 +704,8 @@ describe('integrations', () => {
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
             total_tokens: MOCK_NUMBER,
-            cache_read_input_tokens: 0
+            cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0
           },
           modelName: 'gpt-4o-mini-2024-07-18',
           modelProvider: 'openai',
@@ -700,7 +716,6 @@ describe('integrations', () => {
             tool_choice: 'auto',
             truncation: 'disabled',
             text: { format: { type: 'text' }, verbosity: 'medium' },
-            reasoning_tokens: 0,
             stream: false
           },
           tags: { ml_app: 'test', integration: 'openai' }
@@ -739,7 +754,8 @@ describe('integrations', () => {
             input_tokens: MOCK_NUMBER,
             output_tokens: MOCK_NUMBER,
             total_tokens: MOCK_NUMBER,
-            cache_read_input_tokens: 0
+            cache_read_input_tokens: 0,
+            reasoning_output_tokens: 0
           },
           modelName: 'gpt-4o-mini-2024-07-18',
           modelProvider: 'openai',
@@ -750,7 +766,6 @@ describe('integrations', () => {
             tool_choice: 'auto',
             truncation: 'disabled',
             text: { format: { type: 'text' }, verbosity: 'medium' },
-            reasoning_tokens: 0,
             stream: true
           },
           tags: { ml_app: 'test', integration: 'openai' }
@@ -957,6 +972,52 @@ describe('integrations', () => {
           ])
         })
       })
+
+      it('submits a response span with reasoning tokens', async function () {
+        if (semifies(realVersion, '<4.87.0')) {
+          this.skip()
+        }
+
+        await openai.responses.create({
+          model: 'gpt-5-mini',
+          input: 'Solve this step by step: What is 15 * 24?',
+          max_output_tokens: 500,
+          stream: false
+        })
+
+        const { apmSpans, llmobsSpans } = await getEvents()
+        assertLlmObsSpanEvent(llmobsSpans[0], {
+          span: apmSpans[0],
+          spanKind: 'llm',
+          name: 'OpenAI.createResponse',
+          inputMessages: [
+            { role: 'user', content: 'Solve this step by step: What is 15 * 24?' }
+          ],
+          outputMessages: [
+            { role: 'reasoning', content: MOCK_STRING },
+            { role: 'assistant', content: MOCK_STRING }
+          ],
+          metrics: {
+            input_tokens: MOCK_NUMBER,
+            output_tokens: MOCK_NUMBER,
+            total_tokens: MOCK_NUMBER,
+            cache_read_input_tokens: MOCK_NUMBER,
+            reasoning_output_tokens: 128
+          },
+          modelName: 'gpt-5-mini-2025-08-07',
+          modelProvider: 'openai',
+          metadata: {
+            max_output_tokens: 500,
+            top_p: 1,
+            temperature: 1,
+            tool_choice: 'auto',
+            truncation: 'disabled',
+            text: { format: { type: 'text' }, verbosity: 'medium' },
+            stream: false
+          },
+          tags: { ml_app: 'test', integration: 'openai' }
+        })
+      })
     })
   })
 })