Add GPT-5.6 Sol/Terra/Luna Codex pricing (#2023)

* Add GPT-5.6 Sol/Terra/Luna Codex pricing

Local cost scanning already counts gpt-5.6-* tokens but left day totals understated without rates. Bundle preview pricing so Sol/Terra/Luna resolve without models.dev.

* Price GPT-5.6 long-context, Priority, cache-write, and alias

Address review: official long-context (>272K) and Priority rates, 1.25x
cache-write billing for Pi disjoint writes, and route gpt-5.6 to Sol.

* Invalidate Pi cost cache and price Priority cache writes

Bump Pi session cost cache to v5 so GPT-5.6 rows reprice with 1.25x
cache-write rates, and apply the same multiplier on Priority input.

* Fall back bundled long-context rates for models.dev gaps

When models.dev has a GPT-5.6 hit without context_over_200k fields, still
apply bundled above-threshold input/output/cache-read rates with the 272K
threshold so long prompts are not under-billed at short rates.

* fix: harden GPT-5.6 cost pricing

Co-authored-by: Stephen <stephen@klu.ai>

* Use released GPT-5.6 IDs and fictitious date suffixes in tests

Match AGENTS model-name policy: price coverage uses official unsuffixed
IDs; dated normalization uses explicitly fictitious 2099-01-01 suffixes.

* test: lock GPT-5.6 Priority boundary

---------

Co-authored-by: Peter Steinberger <steipete@gmail.com>

Author: 0xSMW

CHANGELOG.md

@@ -5,6 +5,7 @@
 ### Added
 - Agent Sessions: list and focus live local or SSH-discovered Codex and Claude Code sessions from the menu and CLI.
 - Kimi K2: add a Usage Dashboard shortcut to the human-facing legacy credits page. Thanks @joeVenner!
+- Codex: add GPT-5.6 Sol, Terra, and Luna pricing, including long-context, Priority, cache-write, alias, and automatic Pi cache repricing when rates change (#2023). Thanks @0xSMW!
 - Codex: add a provider option to hide Spark quota rows without hiding credits or other extra usage (#2013). Thanks @intellectronica!
 - Wayfinder: add opt-in local gateway health, routing, savings, and latency usage with configurable loopback URL support. Thanks @tcballard!
 - Claude CLI: surface model-scoped weekly limits alongside all-model usage without duplicating matching web limits. Thanks @janpollak!

Sources/CodexBarCore/Generated/CodexParserHash.generated.swift

@@ -1,5 +1,5 @@
 // Generated by Scripts/regenerate-codex-parser-hash.sh. Do not edit by hand.
 
 enum CodexParserHash {
-    static let value = "3edc97b8139ae59e"
+    static let value = "7985c9336eb88ef7"
 }

Sources/CodexBarCore/PiSessionCostCache.swift

@@ -1,7 +1,8 @@
 import Foundation
 
 enum PiSessionCostCacheIO {
-    private static let artifactVersion = 4
+    /// Artifact schema version. Pricing changes are tracked separately by `pricingKey`.
+    private static let artifactVersion = 5
 
     private static func defaultCacheRoot() -> URL {
         let root = FileManager.default.urls(for: .cachesDirectory, in: .userDomainMask).first!
@@ -51,10 +52,11 @@ struct PiSessionCostCache: Codable {
     var lastScanUnixMs: Int64 = 0
     var scanSinceKey: String?
     var scanUntilKey: String?
+    var pricingKey: String?
     var daysByProvider: [String: [String: [String: PiPackedUsage]]] = [:]
     var files: [String: PiSessionFileUsage] = [:]
 
-    init(version: Int = 4) {
+    init(version: Int = 5) {
         self.version = version
     }
 }

Sources/CodexBarCore/PiSessionCostScanner.swift

@@ -49,6 +49,7 @@ enum PiSessionCostScanner {
     private struct ModelsDevPricingContext {
         let catalog: ModelsDevCatalog?
         let cacheRoot: URL?
+        let pricingKey: String
     }
 
     private struct ScanContext {
@@ -59,6 +60,8 @@ enum PiSessionCostScanner {
     }
 
     private static let costScale = 1_000_000_000.0
+    /// Bump for Pi-only cost formula changes not represented by the parser or pricing fingerprints.
+    private static let costFormulaVersion = 1
     private static let maxLineBytes = 16 * 1024 * 1024
     private static let maxSafeRoundedInt = Double(Int.max) - 1
     private static let sessionStartFilenameRegex = try? NSRegularExpression(
@@ -98,12 +101,12 @@ enum PiSessionCostScanner {
         var cache = PiSessionCostCacheIO.load(cacheRoot: options.cacheRoot)
         let nowMs = Int64(now.timeIntervalSince1970 * 1000)
         let refreshMs = Int64(max(0, options.refreshMinIntervalSeconds) * 1000)
-        let pricingContext = ModelsDevPricingContext(
-            catalog: CostUsagePricing.modelsDevCatalog(now: now, cacheRoot: options.cacheRoot),
-            cacheRoot: options.cacheRoot)
+        let pricingContext = self.pricingContext(now: now, cacheRoot: options.cacheRoot)
         let windowExpanded = self.requestedWindowExpandsCache(range: range, cache: cache)
+        let pricingChanged = cache.pricingKey != pricingContext.pricingKey
         let shouldRefresh = options.forceRescan
             || windowExpanded
+            || pricingChanged
             || refreshMs == 0
             || cache.lastScanUnixMs == 0
             || nowMs - cache.lastScanUnixMs > refreshMs
@@ -121,7 +124,7 @@ enum PiSessionCostScanner {
                     cache: &cache,
                     context: ScanContext(
                         range: range,
-                        forceRescan: options.forceRescan || windowExpanded,
+                        forceRescan: options.forceRescan || windowExpanded || pricingChanged,
                         pricingContext: pricingContext,
                         checkCancellation: checkCancellation))
             }
@@ -139,6 +142,7 @@ enum PiSessionCostScanner {
 
             cache.scanSinceKey = range.scanSinceKey
             cache.scanUntilKey = range.scanUntilKey
+            cache.pricingKey = pricingContext.pricingKey
             cache.lastScanUnixMs = nowMs
             try checkCancellation?()
             PiSessionCostCacheIO.save(cache: cache, cacheRoot: options.cacheRoot)
@@ -185,9 +189,8 @@ enum PiSessionCostScanner {
         guard !cache.daysByProvider.isEmpty else { return nil }
         guard !self.requestedWindowExpandsCache(range: range, cache: cache) else { return nil }
 
-        let pricingContext = ModelsDevPricingContext(
-            catalog: CostUsagePricing.modelsDevCatalog(now: now, cacheRoot: cacheRoot),
-            cacheRoot: cacheRoot)
+        let pricingContext = self.pricingContext(now: now, cacheRoot: cacheRoot)
+        guard cache.pricingKey == pricingContext.pricingKey else { return nil }
         let report = self.buildReport(
             provider: provider,
             cache: cache,
@@ -200,6 +203,17 @@ enum PiSessionCostScanner {
         return CachedDailyReportResult(report: report, lastScanAt: lastScanAt)
     }
 
+    private static func pricingContext(now: Date, cacheRoot: URL?) -> ModelsDevPricingContext {
+        let modelsDevArtifact = ModelsDevCache.load(now: now, cacheRoot: cacheRoot).artifact
+        return ModelsDevPricingContext(
+            catalog: modelsDevArtifact?.catalog,
+            cacheRoot: cacheRoot,
+            pricingKey: CostUsagePricingKey.codex(
+                modelsDevArtifact: modelsDevArtifact,
+                formulaVersion: Self.costFormulaVersion,
+                parserHash: CodexParserHash.value))
+    }
+
     private static func requestedWindowExpandsCache(
         range: CostUsageScanner.CostUsageDayRange,
         cache: PiSessionCostCache) -> Bool
@@ -645,12 +659,14 @@ enum PiSessionCostScanner {
         switch provider {
         case .codex:
             // Pi records input, cache reads, and cache writes as disjoint counts. Codex pricing
-            // expects cached input to be a subset of total input, so reconstruct that total here.
+            // expects cached/write tokens to be subsets of total input, so reconstruct that total
+            // here and pass writes separately (1.25x input for GPT-5.6 when rates are known).
             CostUsagePricing.codexCostUSD(
                 model: modelName,
                 inputTokens: usage.inputTokens + usage.cacheReadTokens + usage.cacheWriteTokens,
                 cachedInputTokens: usage.cacheReadTokens,
                 outputTokens: usage.outputTokens,
+                cacheWriteInputTokens: usage.cacheWriteTokens,
                 modelsDevCatalog: pricingContext?.catalog,
                 modelsDevCacheRoot: pricingContext?.cacheRoot)
         case .claude: