From 7ad44b91d08e0dc1c2b358ad0c1e89c19a63e0b8 Mon Sep 17 00:00:00 2001
From: "kiloconnect[bot]" <240665456+kiloconnect[bot]@users.noreply.github.com>
Date: Sat, 18 Apr 2026 14:15:44 +0000
Subject: [PATCH 1/2] fix(fim): set has_error on microdollar_usage for upstream
 errors

FIM parse functions never checked the HTTP status code, so 4xx/5xx
responses from Mistral/Inception never set has_error=true in
microdollar_usage. Mirror the pattern used by regular chat completions:
initialise reportedError from statusCode>=400 in the streaming path,
and include statusCode>=400 in the hasError expression for non-streaming.
---
 apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts b/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
index 5a84550be1..3fad67cc9c 100644
--- a/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
+++ b/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
@@ -494,7 +494,8 @@ function computeFimMicrodollarCost(usage: FimUsage, provider: ProviderId): numbe
 
 function parseMistralFimUsageFromString(
   response: string,
-  provider: ProviderId
+  provider: ProviderId,
+  statusCode: number
 ): MicrodollarUsageStats {
   const json: MistralFimCompletion = JSON.parse(response);
   const cost_mUsd = computeFimMicrodollarCost(json.usage, provider);
@@ -503,7 +504,7 @@ function parseMistralFimUsageFromString(
     messageId: json.id,
     model: json.model,
     responseContent: json.choices[0]?.text || '',
-    hasError: !json.model,
+    hasError: !json.model || statusCode >= 400,
     inference_provider: provider,
     inputTokens: json.usage.prompt_tokens,
     outputTokens: json.usage.completion_tokens,
@@ -524,7 +525,8 @@ function parseMistralFimUsageFromString(
 async function parseMistralFimUsageFromStream(
   stream: ReadableStream,
   requestSpan: Span | undefined,
-  provider: ProviderId
+  provider: ProviderId,
+  statusCode: number
 ): Promise<MicrodollarUsageStats> {
   requestSpan?.end();
   const streamProcessingSpan = startInactiveSpan({
@@ -539,7 +541,7 @@ async function parseMistralFimUsageFromStream(
   let messageId: string | null = null;
   let model: string | null = null;
   let responseContent = '';
-  let reportedError = false;
+  let reportedError = statusCode >= 400;
   const startedAt = performance.now();
   let firstTokenReceived = false;
   let usage: FimUsage | undefined;
@@ -627,13 +629,14 @@ export function countAndStoreFimUsage(
   const logFileExtension = usageContext.isStreaming ? '.log.resp.sse' : '.log.resp.json';
   debugSaveProxyResponseStream(clonedResponse, logFileExtension);
 
+  const statusCode = usageContext.status_code ?? 0;
   const usageStatsPromise = !clonedResponse.body
     ? Promise.resolve(null)
     : usageContext.isStreaming
-      ? parseMistralFimUsageFromStream(clonedResponse.body, requestSpan, usageContext.provider)
+      ? parseMistralFimUsageFromStream(clonedResponse.body, requestSpan, usageContext.provider, statusCode)
       : clonedResponse
           .text()
-          .then(content => parseMistralFimUsageFromString(content, usageContext.provider));
+          .then(content => parseMistralFimUsageFromString(content, usageContext.provider, statusCode));
 
   after(
     usageStatsPromise.then(usageStats => {

From d5ace23648479510ad441f38c496c7efe95fcd1f Mon Sep 17 00:00:00 2001
From: "kiloconnect[bot]" <240665456+kiloconnect[bot]@users.noreply.github.com>
Date: Sat, 18 Apr 2026 14:18:03 +0000
Subject: [PATCH 2/2] style: apply oxfmt formatting

---
 apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts b/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
index 3fad67cc9c..527063d99c 100644
--- a/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
+++ b/apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts
@@ -633,10 +633,17 @@ export function countAndStoreFimUsage(
   const usageStatsPromise = !clonedResponse.body
     ? Promise.resolve(null)
     : usageContext.isStreaming
-      ? parseMistralFimUsageFromStream(clonedResponse.body, requestSpan, usageContext.provider, statusCode)
+      ? parseMistralFimUsageFromStream(
+          clonedResponse.body,
+          requestSpan,
+          usageContext.provider,
+          statusCode
+        )
       : clonedResponse
           .text()
-          .then(content => parseMistralFimUsageFromString(content, usageContext.provider, statusCode));
+          .then(content =>
+            parseMistralFimUsageFromString(content, usageContext.provider, statusCode)
+          );
 
   after(
     usageStatsPromise.then(usageStats => {