🐛 Fix token usage cost calculation with granular input/output breakdown

OrbitDockNative/OrbitDock/Models/RootShell/RootSessionNode.swift

@@ -148,6 +148,9 @@ struct RootSessionNode: Identifiable, Sendable {
   let missionId: String?
   let issueIdentifier: String?
   let totalTokens: Int
+  let inputTokens: Int
+  let outputTokens: Int
+  let cachedTokens: Int
   let totalCostUSD: Double
   let isActive: Bool
   let showsInMissionControl: Bool
@@ -281,6 +284,9 @@ extension RootSessionNode: Equatable {
       && lhs.missionId == rhs.missionId
       && lhs.issueIdentifier == rhs.issueIdentifier
       && lhs.totalTokens == rhs.totalTokens
+      && lhs.inputTokens == rhs.inputTokens
+      && lhs.outputTokens == rhs.outputTokens
+      && lhs.cachedTokens == rhs.cachedTokens
       && lhs.totalCostUSD == rhs.totalCostUSD
       && lhs.isActive == rhs.isActive
       && lhs.showsInMissionControl == rhs.showsInMissionControl
@@ -376,6 +382,9 @@ extension RootSessionNode {
     self.missionId = session.missionId
     self.issueIdentifier = session.issueIdentifier
     self.totalTokens = Int(session.totalTokens ?? 0)
+    self.inputTokens = Int(session.inputTokens ?? 0)
+    self.outputTokens = Int(session.outputTokens ?? 0)
+    self.cachedTokens = Int(session.cachedTokens ?? 0)
     self.totalCostUSD = session.totalCostUSD ?? 0
     self.isActive = status == .active
     self.showsInMissionControl = RootSessionNode.showsInMissionControl(
@@ -435,6 +444,9 @@ extension RootSessionNode {
       missionId: missionId,
       issueIdentifier: issueIdentifier,
       totalTokens: totalTokens,
+      inputTokens: inputTokens,
+      outputTokens: outputTokens,
+      cachedTokens: cachedTokens,
       totalCostUSD: totalCostUSD,
       isActive: self.status == .active,
       showsInMissionControl: RootSessionNode.showsInMissionControl(
@@ -486,6 +498,9 @@ extension RootSessionNode {
       missionId: missionId,
       issueIdentifier: issueIdentifier,
       totalTokens: totalTokens,
+      inputTokens: inputTokens,
+      outputTokens: outputTokens,
+      cachedTokens: cachedTokens,
       totalCostUSD: totalCostUSD,
       isActive: false,
       showsInMissionControl: RootSessionNode.showsInMissionControl(

OrbitDockNative/OrbitDock/Services/Server/Protocol/ServerSessionContracts.swift

@@ -207,6 +207,9 @@ struct ServerSessionListItem: Codable, Identifiable {
   let worktreeId: String?
   let totalTokens: UInt64?
   let totalCostUSD: Double?
+  let inputTokens: UInt64?
+  let outputTokens: UInt64?
+  let cachedTokens: UInt64?
   let displayTitle: String?
   let displayTitleSortKey: String?
   let displaySearchText: String?
@@ -241,6 +244,9 @@ struct ServerSessionListItem: Codable, Identifiable {
     worktreeId: String?,
     totalTokens: UInt64?,
     totalCostUSD: Double?,
+    inputTokens: UInt64? = nil,
+    outputTokens: UInt64? = nil,
+    cachedTokens: UInt64? = nil,
     displayTitle: String?,
     displayTitleSortKey: String?,
     displaySearchText: String?,
@@ -274,6 +280,9 @@ struct ServerSessionListItem: Codable, Identifiable {
     self.worktreeId = worktreeId
     self.totalTokens = totalTokens
     self.totalCostUSD = totalCostUSD
+    self.inputTokens = inputTokens
+    self.outputTokens = outputTokens
+    self.cachedTokens = cachedTokens
     self.displayTitle = displayTitle
     self.displayTitleSortKey = displayTitleSortKey
     self.displaySearchText = displaySearchText
@@ -309,6 +318,9 @@ struct ServerSessionListItem: Codable, Identifiable {
     case worktreeId = "worktree_id"
     case totalTokens = "total_tokens"
     case totalCostUSD = "total_cost_usd"
+    case inputTokens = "input_tokens"
+    case outputTokens = "output_tokens"
+    case cachedTokens = "cached_tokens"
     case displayTitle = "display_title"
     case displayTitleSortKey = "display_title_sort_key"
     case displaySearchText = "display_search_text"

OrbitDockNative/OrbitDock/Views/Dashboard/Usage/StatsPopoverContent.swift

@@ -273,6 +273,18 @@ struct StatusBarStats {
       return session.totalCostUSD
     }
     guard session.totalTokens > 0 else { return 0 }
+    // Use granular token breakdown when available for accurate cost.
+    // Cached tokens are billed at the cache-read rate, not the full input rate.
+    let hasBreakdown = session.inputTokens > 0 || session.outputTokens > 0
+    if hasBreakdown {
+      return costCalculator.calculateCost(
+        model: session.model,
+        inputTokens: session.inputTokens,
+        outputTokens: session.outputTokens,
+        cacheReadTokens: session.cachedTokens
+      )
+    }
+    // Legacy fallback: treat totalTokens as input (pre-breakdown servers).
     return costCalculator.calculateCost(
       model: session.model,
       inputTokens: session.totalTokens,

OrbitDockNative/OrbitDock/Views/QuickSwitcher/QuickSwitcherPreview.swift

@@ -86,6 +86,9 @@ private func quickSwitcherPreviewNode(
     missionId: nil,
     issueIdentifier: nil,
     totalTokens: 0,
+    inputTokens: 0,
+    outputTokens: 0,
+    cachedTokens: 0,
     totalCostUSD: 0,
     isActive: status == .active,
     showsInMissionControl: status == .active,

OrbitDockNative/OrbitDockTests/SessionDetail/StatusBarStatsTests.swift

@@ -75,10 +75,73 @@ struct StatusBarStatsTests {
     #expect(stats.costByModel.first?.cost == stats.cost)
   }
 
+  @Test func fromUsesGranularTokenBreakdownForCost() {
+    // When input/output breakdown is available, cost should reflect
+    // the different rates for input vs output tokens.
+    let inputCount = 500_000
+    let outputCount = 500_000
+    let cachedCount = 100_000
+
+    let sessions = [
+      makeSession(
+        id: "claude-1",
+        model: "claude-sonnet-4",
+        totalTokens: inputCount + outputCount,
+        inputTokens: inputCount,
+        outputTokens: outputCount,
+        cachedTokens: cachedCount,
+        totalCostUSD: 0
+      ),
+    ]
+
+    let stats = StatusBarStats.from(
+      sessions: sessions,
+      costCalculator: costCalculator
+    )
+
+    // With the fallback calculator (no prices loaded), defaults are:
+    // input: $3/M, output: $15/M, cache_read: $0.30/M
+    let expectedInput = Double(inputCount) / 1_000_000 * 3.0
+    let expectedOutput = Double(outputCount) / 1_000_000 * 15.0
+    let expectedCache = Double(cachedCount) / 1_000_000 * 0.30
+    let expectedCost = expectedInput + expectedOutput + expectedCache
+
+    #expect(stats.tokens == inputCount + outputCount)
+    #expect(abs(stats.cost - expectedCost) < 0.001)
+  }
+
+  @Test func fromFallsBackToLegacyWhenNoBreakdown() {
+    // When input/output are both 0 (legacy server), falls back to
+    // treating totalTokens as input.
+    let sessions = [
+      makeSession(
+        id: "claude-1",
+        model: "claude-sonnet-4",
+        totalTokens: 1_000_000,
+        inputTokens: 0,
+        outputTokens: 0,
+        cachedTokens: 0,
+        totalCostUSD: 0
+      ),
+    ]
+
+    let stats = StatusBarStats.from(
+      sessions: sessions,
+      costCalculator: costCalculator
+    )
+
+    // Legacy fallback: all tokens treated as input at $3/M
+    let expectedCost = Double(1_000_000) / 1_000_000 * 3.0
+    #expect(abs(stats.cost - expectedCost) < 0.001)
+  }
+
   private func makeSession(
     id: String,
     model: String,
     totalTokens: Int,
+    inputTokens: Int = 0,
+    outputTokens: Int = 0,
+    cachedTokens: Int = 0,
     totalCostUSD: Double
   ) -> RootSessionNode {
     RootSessionNode(
@@ -116,6 +179,9 @@ struct StatusBarStatsTests {
       missionId: nil,
       issueIdentifier: nil,
       totalTokens: totalTokens,
+      inputTokens: inputTokens,
+      outputTokens: outputTokens,
+      cachedTokens: cachedTokens,
       totalCostUSD: totalCostUSD,
       isActive: true,
       showsInMissionControl: true,

orbitdock-server/crates/protocol/src/types.rs

@@ -776,6 +776,9 @@ impl SessionSummary {
             worktree_id: self.worktree_id.clone(),
             total_tokens: self.token_usage.input_tokens + self.token_usage.output_tokens,
             total_cost_usd: 0.0,
+            input_tokens: self.token_usage.input_tokens,
+            output_tokens: self.token_usage.output_tokens,
+            cached_tokens: self.token_usage.cached_tokens,
             display_title: self.display_title.clone(),
             context_line: self.context_line.clone(),
             list_status: self.list_status,
@@ -813,6 +816,9 @@ impl From<SessionSummary> for SessionListItem {
             worktree_id: summary.worktree_id,
             total_tokens: summary.token_usage.input_tokens + summary.token_usage.output_tokens,
             total_cost_usd: 0.0,
+            input_tokens: summary.token_usage.input_tokens,
+            output_tokens: summary.token_usage.output_tokens,
+            cached_tokens: summary.token_usage.cached_tokens,
             display_title: summary.display_title,
             context_line: summary.context_line,
             list_status: summary.list_status,
@@ -1056,6 +1062,13 @@ pub struct SessionListItem {
     pub total_tokens: u64,
     #[serde(default)]
     pub total_cost_usd: f64,
+    /// Granular token breakdown for accurate cost calculation.
+    #[serde(default)]
+    pub input_tokens: u64,
+    #[serde(default)]
+    pub output_tokens: u64,
+    #[serde(default)]
+    pub cached_tokens: u64,
     pub display_title: String,
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub context_line: Option<String>,
@@ -1105,6 +1118,9 @@ impl SessionListItem {
             worktree_id: summary.worktree_id.clone(),
             total_tokens: summary.token_usage.input_tokens + summary.token_usage.output_tokens,
             total_cost_usd: 0.0,
+            input_tokens: summary.token_usage.input_tokens,
+            output_tokens: summary.token_usage.output_tokens,
+            cached_tokens: summary.token_usage.cached_tokens,
             display_title: summary.display_title.clone(),
             context_line: summary.context_line.clone(),
             list_status: summary.list_status,