feat: add completion tokens details for anthropic models

go.mod

@@ -9,7 +9,7 @@ require (
 	github.com/a8m/envsubst v1.4.3
 	github.com/alecthomas/kong v1.15.0
 	github.com/andybalholm/brotli v1.2.1
-	github.com/anthropics/anthropic-sdk-go v1.45.0
+	github.com/anthropics/anthropic-sdk-go v1.46.0
 	github.com/aws/aws-sdk-go-v2 v1.41.7
 	github.com/aws/aws-sdk-go-v2/aws/protocol/eventstream v1.7.10
 	github.com/aws/aws-sdk-go-v2/config v1.32.18

go.sum

@@ -43,8 +43,8 @@ github.com/alecthomas/repr v0.5.2 h1:SU73FTI9D1P5UNtvseffFSGmdNci/O6RsqzeXJtP0Qs
 github.com/alecthomas/repr v0.5.2/go.mod h1:Fr0507jx4eOXV7AlPV6AVZLYrLIuIeSOWtW57eE/O/4=
 github.com/andybalholm/brotli v1.2.1 h1:R+f5xP285VArJDRgowrfb9DqL18yVK0gKAW/F+eTWro=
 github.com/andybalholm/brotli v1.2.1/go.mod h1:rzTDkvFWvIrjDXZHkuS16NPggd91W3kUSvPlQ1pLaKY=
-github.com/anthropics/anthropic-sdk-go v1.45.0 h1:rWnpyBpm9OAm97jyH5bi6W4SRCwJeNY/RyhaJ7CHSUI=
-github.com/anthropics/anthropic-sdk-go v1.45.0/go.mod h1:bx5vWuHFuGPkELH8Z4KUiNSohFnUwScdpTyr+50myPo=
+github.com/anthropics/anthropic-sdk-go v1.46.0 h1:yl3n+el5ZfNgiCtQ7zQ7s/NXxB11YbrKXdc3uLPNWlU=
+github.com/anthropics/anthropic-sdk-go v1.46.0/go.mod h1:bx5vWuHFuGPkELH8Z4KUiNSohFnUwScdpTyr+50myPo=
 github.com/antlr4-go/antlr/v4 v4.13.1 h1:SqQKkuVZ+zWkMMNkjy5FZe5mr5WURWnlpmOuzYWrPrQ=
 github.com/antlr4-go/antlr/v4 v4.13.1/go.mod h1:GKmUxMtwp6ZgGwZSva4eWPC5mS6vUAmOABFgjdkM7Nw=
 github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5 h1:0CwZNZbxp69SHPdPJAN/hZIm0C4OItdklCFmMRWYpio=

internal/translator/anthropic_helper.go

@@ -864,6 +864,7 @@ func (p *anthropicStreamParser) Process(body io.Reader, endOfStream bool, span t
 		totalTokens, _ := p.tokenUsage.TotalTokens()
 		cachedTokens, _ := p.tokenUsage.CachedInputTokens()
 		cacheCreationTokens, _ := p.tokenUsage.CacheCreationInputTokens()
+		reasoningTokens, _ := p.tokenUsage.ReasoningTokens()
 		finalChunk := openai.ChatCompletionResponseChunk{
 			ID:      p.activeMessageID,
 			Created: p.created,
@@ -877,6 +878,9 @@ func (p *anthropicStreamParser) Process(body io.Reader, endOfStream bool, span t
 					CachedTokens:        int(cachedTokens),
 					CacheCreationTokens: int(cacheCreationTokens),
 				},
+				CompletionTokensDetails: &openai.CompletionTokensDetails{
+					ReasoningTokens: int(reasoningTokens),
+				},
 			},
 			Model: p.requestModel,
 		}
@@ -1064,6 +1068,7 @@ func (p *anthropicStreamParser) handleAnthropicStreamEvent(eventType []byte, dat
 			// Accumulate cache creation tokens
 			p.tokenUsage.AddCacheCreationInputTokens(cacheCreation)
 		}
+		p.tokenUsage.SetReasoningTokens(uint32(u.OutputTokensDetails.ThinkingTokens)) //nolint:gosec
 		if event.Delta.StopReason != "" {
 			p.stopReason = event.Delta.StopReason
 		}
@@ -1178,11 +1183,13 @@ func messageToChatCompletion(anthropicResp *anthropic.Message, responseModel int
 		&usage.CacheReadInputTokens,
 		&usage.CacheCreationInputTokens,
 	)
+	tokenUsage.SetReasoningTokens(uint32(usage.OutputTokensDetails.ThinkingTokens)) //nolint:gosec
 	inputTokens, _ := tokenUsage.InputTokens()
 	outputTokens, _ := tokenUsage.OutputTokens()
 	totalTokens, _ := tokenUsage.TotalTokens()
 	cachedTokens, _ := tokenUsage.CachedInputTokens()
 	cacheCreationTokens, _ := tokenUsage.CacheCreationInputTokens()
+	reasoningTokens, _ := tokenUsage.ReasoningTokens()
 	openAIResp.Usage = openai.Usage{
 		CompletionTokens: int(outputTokens),
 		PromptTokens:     int(inputTokens),
@@ -1191,6 +1198,9 @@ func messageToChatCompletion(anthropicResp *anthropic.Message, responseModel int
 			CachedTokens:        int(cachedTokens),
 			CacheCreationTokens: int(cacheCreationTokens),
 		},
+		CompletionTokensDetails: &openai.CompletionTokensDetails{
+			ReasoningTokens: int(reasoningTokens),
+		},
 	}
 
 	finishReason, err := anthropicToOpenAIFinishReason(anthropicResp.StopReason)

internal/translator/openai_awsanthropic_test.go

@@ -333,6 +333,7 @@ func TestOpenAIToAWSAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 					PromptTokensDetails: &openai.PromptTokensDetails{
 						CachedTokens: 5,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -367,6 +368,7 @@ func TestOpenAIToAWSAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 					PromptTokensDetails: &openai.PromptTokensDetails{
 						CachedTokens: 10,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -420,7 +422,7 @@ func TestOpenAIToAWSAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 				int32(tt.expectedOpenAIResponse.Usage.PromptTokensDetails.CacheCreationTokens), // nolint:gosec
 				int32(tt.expectedOpenAIResponse.Usage.CompletionTokens),                        // nolint:gosec
 				int32(tt.expectedOpenAIResponse.Usage.TotalTokens),                             // nolint:gosec
-				-1,
+				int32(tt.expectedOpenAIResponse.Usage.CompletionTokensDetails.ReasoningTokens), // nolint:gosec
 			)
 			require.Equal(t, expectedTokenUsage, usedToken)
 

internal/translator/openai_gcpanthropic_test.go

@@ -443,6 +443,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 						CachedTokens:        5,
 						CacheCreationTokens: 3,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -478,6 +479,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 						CachedTokens:        10,
 						CacheCreationTokens: 7,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -524,6 +526,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 					PromptTokensDetails: &openai.PromptTokensDetails{
 						CachedTokens: 2,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -557,6 +560,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 					PromptTokensDetails: &openai.PromptTokensDetails{
 						CachedTokens: 3,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -602,6 +606,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 					PromptTokensDetails: &openai.PromptTokensDetails{
 						CachedTokens: 1,
 					},
+					CompletionTokensDetails: &openai.CompletionTokensDetails{},
 				},
 				Choices: []openai.ChatCompletionResponseChoice{
 					{
@@ -651,7 +656,7 @@ func TestOpenAIToGCPAnthropicTranslatorV1ChatCompletion_ResponseBody(t *testing.
 				int32(tt.expectedOpenAIResponse.Usage.PromptTokensDetails.CacheCreationTokens), // nolint:gosec
 				int32(tt.expectedOpenAIResponse.Usage.CompletionTokens),                        // nolint:gosec
 				int32(tt.expectedOpenAIResponse.Usage.TotalTokens),                             // nolint:gosec
-				-1,
+				int32(tt.expectedOpenAIResponse.Usage.CompletionTokensDetails.ReasoningTokens), // nolint:gosec
 			)
 			require.Equal(t, expectedTokenUsage, usedToken)
 

tests/data-plane/testupstream_test.go

@@ -290,7 +290,7 @@ func TestWithTestUpstream(t *testing.T) {
 			requestBody:     toolCallResultsRequestBody,
 			expRequestBody:  `{"max_tokens":1024,"messages":[{"content":[{"text":"List the files in the /tmp directory","type":"text"}],"role":"user"},{"content":[{"id":"call_abc123","input":{"path":"/tmp"},"name":"list_files","type":"tool_use"}],"role":"assistant"},{"content":[{"tool_use_id":"call_abc123","is_error":false,"content":[{"text":"[\"foo.txt\", \"bar.log\", \"data.csv\"]","type":"text"}],"type":"tool_result"}],"role":"user"}],"anthropic_version":"vertex-2023-10-16"}`,
 			responseBody:    `{"id":"msg_123","type":"message","role":"assistant","stop_reason": "end_turn", "content":[{"type":"text","text":"Hello from Anthropic!"}],"usage":{"input_tokens":10,"output_tokens":25,"cache_read_input_tokens":10}}`,
-			expResponseBody: `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from Anthropic!","role":"assistant"}}],"created":123, "id":"msg_123","model":"gpt-4-0613","object":"chat.completion","usage":{"completion_tokens":25,"prompt_tokens":20,"total_tokens":45,"prompt_tokens_details":{"cached_tokens":10}}}`,
+			expResponseBody: `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from Anthropic!","role":"assistant"}}],"created":123, "id":"msg_123","model":"gpt-4-0613","object":"chat.completion","usage":{"completion_tokens":25,"completion_tokens_details":{},"prompt_tokens":20,"total_tokens":45,"prompt_tokens_details":{"cached_tokens":10}}}`,
 			expStatus:       http.StatusOK,
 		},
 		{
@@ -358,7 +358,7 @@ func TestWithTestUpstream(t *testing.T) {
 			responseStatus:    strconv.Itoa(http.StatusOK),
 			responseBody:      `{"id":"msg_123","type":"message","role":"assistant","stop_reason": "end_turn", "content":[{"type":"text","text":"Hello from Anthropic!"}],"usage":{"input_tokens":10,"output_tokens":25}}`,
 			expStatus:         http.StatusOK,
-			expResponseBody:   `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from Anthropic!","role":"assistant"}}],"created":123, "id":"msg_123","model":"claude-3-sonnet","object":"chat.completion","usage":{"completion_tokens":25,"prompt_tokens":10,"total_tokens":35,"prompt_tokens_details":{}}}`,
+			expResponseBody:   `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from Anthropic!","role":"assistant"}}],"created":123, "id":"msg_123","model":"claude-3-sonnet","object":"chat.completion","usage":{"completion_tokens":25,"completion_tokens_details":{},"prompt_tokens":10,"total_tokens":35,"prompt_tokens_details":{}}}`,
 		},
 		{
 			name:              "gcp-anthropicai - /v1/chat/completions - with cache",
@@ -372,7 +372,7 @@ func TestWithTestUpstream(t *testing.T) {
 			responseStatus:    strconv.Itoa(http.StatusOK),
 			responseBody:      `{"id":"msg_123","type":"message","role":"assistant","stop_reason": "end_turn", "content":[{"type":"text","text":"Hello from cached Anthropic!"}],"usage":{"input_tokens":10,"output_tokens":25, "cache_read_input_tokens": 8}}`,
 			expStatus:         http.StatusOK,
-			expResponseBody:   `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from cached Anthropic!","role":"assistant"}}], "created":123, "id":"msg_123", "model":"claude-3-sonnet","object":"chat.completion","usage":{"completion_tokens":25,"prompt_tokens":18,"total_tokens":43,"prompt_tokens_details":{"cached_tokens":8}}}`,
+			expResponseBody:   `{"choices":[{"finish_reason":"stop","index":0,"message":{"content":"Hello from cached Anthropic!","role":"assistant"}}], "created":123, "id":"msg_123", "model":"claude-3-sonnet","object":"chat.completion","usage":{"completion_tokens":25,"completion_tokens_details":{},"prompt_tokens":18,"total_tokens":43,"prompt_tokens_details":{"cached_tokens":8}}}`,
 		},
 		{
 			name:            "modelname-override - /v1/chat/completions",
@@ -578,7 +578,7 @@ data: {"id":"msg_123","choices":[{"index":0,"delta":{"content":" due to Rayleigh
 
 data: {"id":"msg_123","choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk"}
 
-data: {"id":"msg_123","choices":[],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk","usage":{"prompt_tokens":25,"completion_tokens":12,"total_tokens":37,"prompt_tokens_details":{"cached_tokens":10}}}
+data: {"id":"msg_123","choices":[],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk","usage":{"prompt_tokens":25,"completion_tokens":12,"total_tokens":37,"completion_tokens_details":{},"prompt_tokens_details":{"cached_tokens":10}}}
 
 data: [DONE]
 
@@ -642,7 +642,7 @@ data: {"id":"msg_123","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"i
 
 data: {"id":"msg_123","choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk"}
 
-data: {"id":"msg_123","choices":[],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk","usage":{"prompt_tokens":50,"completion_tokens":20,"total_tokens":70,"prompt_tokens_details":{}}}
+data: {"id":"msg_123","choices":[],"created":123,"model":"claude-3-sonnet","object":"chat.completion.chunk","usage":{"prompt_tokens":50,"completion_tokens":20,"total_tokens":70,"completion_tokens_details":{},"prompt_tokens_details":{}}}
 
 data: [DONE]