refactor: simplify Langfuse integration with AI SDK 6 (#375)

- Remove manual token attribute setting (AI SDK 6 telemetry auto-reports)
- Use totalTokens directly instead of inputTokens + outputTokens calculation
- Fix sessionId bug in log-save/log-feedback (prevents wrong trace attachment)
- Hash IP addresses for privacy instead of storing raw IPs
- Fix isLangfuseEnabled() to check both keys for consistency

app/api/chat/route.ts

@@ -162,9 +162,13 @@ async function handleChatRequest(req: Request): Promise<Response> {
 
     const { messages, xml, previousXml, sessionId } = await req.json()
 
-    // Get user IP for Langfuse tracking
+    // Get user IP for Langfuse tracking (hashed for privacy)
     const forwardedFor = req.headers.get("x-forwarded-for")
-    const userId = forwardedFor?.split(",")[0]?.trim() || "anonymous"
+    const rawIp = forwardedFor?.split(",")[0]?.trim() || "anonymous"
+    const userId =
+        rawIp === "anonymous"
+            ? rawIp
+            : `user-${Buffer.from(rawIp).toString("base64url").slice(0, 8)}`
 
     // Validate sessionId for Langfuse (must be string, max 200 chars)
     const validSessionId =
@@ -506,12 +510,9 @@ ${userInputText}
                 userId,
             }),
         }),
-        onFinish: ({ text, usage }) => {
-            // Pass usage to Langfuse (Bedrock streaming doesn't auto-report tokens to telemetry)
-            setTraceOutput(text, {
-                promptTokens: usage?.inputTokens,
-                completionTokens: usage?.outputTokens,
-            })
+        onFinish: ({ text }) => {
+            // AI SDK 6 telemetry auto-reports token usage on its spans
+            setTraceOutput(text)
         },
         tools: {
             // Client-side tool that will be executed on the client
@@ -681,20 +682,9 @@ Call this tool to get shape names and usage syntax for a specific library.`,
         messageMetadata: ({ part }) => {
             if (part.type === "finish") {
                 const usage = (part as any).totalUsage
-                if (!usage) {
-                    console.warn(
-                        "[messageMetadata] No usage data in finish part",
-                    )
-                    return undefined
-                }
-                // Total input = non-cached + cached (these are separate counts)
-                // Note: cacheWriteInputTokens is not available on finish part
-                const totalInputTokens =
-                    (usage.inputTokens ?? 0) +
-                    (usage.inputTokenDetails?.cacheReadTokens ?? 0)
+                // AI SDK 6 provides totalTokens directly
                 return {
-                    inputTokens: totalInputTokens,
-                    outputTokens: usage.outputTokens ?? 0,
+                    totalTokens: usage?.totalTokens ?? 0,
                     finishReason: (part as any).finishReason,
                 }
             }

app/api/log-feedback/route.ts

@@ -27,9 +27,18 @@ export async function POST(req: Request) {
 
     const { messageId, feedback, sessionId } = data
 
-    // Get user IP for tracking
+    // Skip logging if no sessionId - prevents attaching to wrong user's trace
+    if (!sessionId) {
+        return Response.json({ success: true, logged: false })
+    }
+
+    // Get user IP for tracking (hashed for privacy)
     const forwardedFor = req.headers.get("x-forwarded-for")
-    const userId = forwardedFor?.split(",")[0]?.trim() || "anonymous"
+    const rawIp = forwardedFor?.split(",")[0]?.trim() || "anonymous"
+    const userId =
+        rawIp === "anonymous"
+            ? rawIp
+            : `user-${Buffer.from(rawIp).toString("base64url").slice(0, 8)}`
 
     try {
         // Find the most recent chat trace for this session to attach the score to

app/api/log-save/route.ts

@@ -27,6 +27,11 @@ export async function POST(req: Request) {
 
     const { filename, format, sessionId } = data
 
+    // Skip logging if no sessionId - prevents attaching to wrong user's trace
+    if (!sessionId) {
+        return Response.json({ success: true, logged: false })
+    }
+
     try {
         const timestamp = new Date().toISOString()
 

components/chat-panel.tsx

@@ -632,21 +632,15 @@ Continue from EXACTLY where you stopped.`,
 
             // DEBUG: Log finish reason to diagnose truncation
             console.log("[onFinish] finishReason:", metadata?.finishReason)
-            console.log("[onFinish] metadata:", metadata)
 
-            if (metadata) {
-                // Use Number.isFinite to guard against NaN (typeof NaN === 'number' is true)
-                const inputTokens = Number.isFinite(metadata.inputTokens)
-                    ? (metadata.inputTokens as number)
+            // AI SDK 6 provides totalTokens directly
+            const totalTokens =
+                metadata && Number.isFinite(metadata.totalTokens)
+                    ? (metadata.totalTokens as number)
                     : 0
-                const outputTokens = Number.isFinite(metadata.outputTokens)
-                    ? (metadata.outputTokens as number)
-                    : 0
-                const actualTokens = inputTokens + outputTokens
-                if (actualTokens > 0) {
-                    quotaManager.incrementTokenCount(actualTokens)
-                    quotaManager.incrementTPMCount(actualTokens)
-                }
+            if (totalTokens > 0) {
+                quotaManager.incrementTokenCount(totalTokens)
+                quotaManager.incrementTPMCount(totalTokens)
             }
         },
         sendAutomaticallyWhen: ({ messages }) => {

lib/langfuse.ts

@@ -21,9 +21,11 @@ export function getLangfuseClient(): LangfuseClient | null {
     return langfuseClient
 }
 
-// Check if Langfuse is configured
+// Check if Langfuse is configured (both keys required)
 export function isLangfuseEnabled(): boolean {
-    return !!process.env.LANGFUSE_PUBLIC_KEY
+    return !!(
+        process.env.LANGFUSE_PUBLIC_KEY && process.env.LANGFUSE_SECRET_KEY
+    )
 }
 
 // Update trace with input data at the start of request
@@ -43,34 +45,16 @@ export function setTraceInput(params: {
 }
 
 // Update trace with output and end the span
-export function setTraceOutput(
-    output: string,
-    usage?: { promptTokens?: number; completionTokens?: number },
-) {
+// Note: AI SDK 6 telemetry automatically reports token usage on its spans,
+// so we only need to set the output text and close our wrapper span
+export function setTraceOutput(output: string) {
     if (!isLangfuseEnabled()) return
 
     updateActiveTrace({ output })
 
+    // End the observe() wrapper span (AI SDK creates its own child spans with usage)
     const activeSpan = api.trace.getActiveSpan()
     if (activeSpan) {
-        // Manually set usage attributes since AI SDK Bedrock streaming doesn't provide them
-        if (usage?.promptTokens) {
-            activeSpan.setAttribute("ai.usage.promptTokens", usage.promptTokens)
-            activeSpan.setAttribute(
-                "gen_ai.usage.input_tokens",
-                usage.promptTokens,
-            )
-        }
-        if (usage?.completionTokens) {
-            activeSpan.setAttribute(
-                "ai.usage.completionTokens",
-                usage.completionTokens,
-            )
-            activeSpan.setAttribute(
-                "gen_ai.usage.output_tokens",
-                usage.completionTokens,
-            )
-        }
         activeSpan.end()
     }
 }