From 42b730a9470db500ffc98545ca038563e5238f30 Mon Sep 17 00:00:00 2001
From: Roo Code <roomote@roocode.com>
Date: Sun, 23 Nov 2025 15:26:50 +0000
Subject: [PATCH] fix: improve token usage extraction for OpenAI-compatible
 proxies

- Add support for alternative field names used by various OpenAI-compatible proxies
- Check for usage_metadata field in addition to usage field
- Try multiple token field variations (prompt_tokens, input_tokens, promptTokens, etc.)
- Add debug logging when tokens are present but not extracted correctly
- Fixes issue where Open WebUI proxy responses show 0 tokens despite having valid usage data

Fixes #9514
---
 .../base-openai-compatible-provider.ts        | 32 ++++++++++++++++---
 1 file changed, 28 insertions(+), 4 deletions(-)

diff --git a/src/api/providers/base-openai-compatible-provider.ts b/src/api/providers/base-openai-compatible-provider.ts
index 3d78ef75d1..dc2631166a 100644
--- a/src/api/providers/base-openai-compatible-provider.ts
+++ b/src/api/providers/base-openai-compatible-provider.ts
@@ -188,8 +188,13 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 				toolCallAccumulator.clear()
 			}
 
+			// Check for usage data in the chunk
+			// Open WebUI and other proxies might include usage data differently
 			if (chunk.usage) {
 				lastUsage = chunk.usage
+			} else if ((chunk as any).usage_metadata) {
+				// Some proxies use usage_metadata instead of usage
+				lastUsage = (chunk as any).usage_metadata
 			}
 		}
 
@@ -218,15 +223,34 @@ export abstract class BaseOpenAiCompatibleProvider<ModelName extends string>
 	}
 
 	protected processUsageMetrics(usage: any, modelInfo?: any): ApiStreamUsageChunk {
-		const inputTokens = usage?.prompt_tokens || 0
-		const outputTokens = usage?.completion_tokens || 0
-		const cacheWriteTokens = usage?.prompt_tokens_details?.cache_write_tokens || 0
-		const cacheReadTokens = usage?.prompt_tokens_details?.cached_tokens || 0
+		// Try different field names that various OpenAI-compatible proxies might use
+		const inputTokens = usage?.prompt_tokens || usage?.input_tokens || usage?.promptTokens || 0
+		const outputTokens = usage?.completion_tokens || usage?.output_tokens || usage?.completionTokens || 0
+
+		// Cache tokens might be in different locations
+		const cacheWriteTokens =
+			usage?.prompt_tokens_details?.cache_write_tokens ||
+			usage?.prompt_tokens_details?.cache_creation_input_tokens ||
+			usage?.cache_creation_input_tokens ||
+			0
+		const cacheReadTokens =
+			usage?.prompt_tokens_details?.cached_tokens ||
+			usage?.prompt_tokens_details?.cache_read_input_tokens ||
+			usage?.cache_read_input_tokens ||
+			0
 
 		const { totalCost } = modelInfo
 			? calculateApiCostOpenAI(modelInfo, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
 			: { totalCost: 0 }
 
+		// Log for debugging Open WebUI integration
+		if (usage && inputTokens === 0 && outputTokens === 0 && usage.total_tokens > 0) {
+			console.debug(
+				`[BaseOpenAiCompatibleProvider] Usage data present but tokens not extracted correctly:`,
+				usage,
+			)
+		}
+
 		return {
 			type: "usage",
 			inputTokens,