fix(warp): price conversations from total_billed_cost_in_cents

src/daily-cache.ts

@@ -361,7 +361,10 @@ import type { DateRange, ProjectSummary } from './types.js'
 // copilotCredits, and a chat session that also ran on the agent host is left
 // to it. That twin drops calls, so copilot joins
 // PENDING_REDERIVE_PROVIDER_VERSIONS at 75.
-export const DAILY_CACHE_VERSION = 75
+// v77: Warp conversations that carry only total_billed_cost_in_cents are
+// priced from it instead of credits or the token floor. Calls are unchanged,
+// so no PENDING_REDERIVE_PROVIDER_VERSIONS entry is needed.
+export const DAILY_CACHE_VERSION = 77
 const MIN_SUPPORTED_VERSION = 28
 
 /// Providers whose per-day CALL COUNT means something different at

src/providers/warp.ts

@@ -88,6 +88,9 @@ type WarpConversationData = {
     total_provider_cost_in_cents?: number
     // Cumulative per-category charged usage across the conversation, in cents.
     total_charged_usage?: WarpChargedUsageTotals
+    // Server snapshot of what the customer was billed, in US cents. Replaced
+    // total_provider_cost_in_cents in Warp 0.2026.10 (warpdotdev/warp#16327).
+    total_billed_cost_in_cents?: number
     // Credits Warp deducted for the conversation (always populated today).
     credits_spent?: number
   }
@@ -216,8 +219,9 @@ function extractTokenBudget(rawConversationData: string): { tokenBudget: number;
 // least authoritative:
 //   (a) total_provider_cost_in_cents  — server-authoritative USD  (exact)
 //   (b) sum of total_charged_usage    — server per-category USD   (exact)
-//   (c) credits_spent × CREDIT_USD_RATE — credits at published rate (heuristic)
-//   (d) null → caller falls back to the token × LiteLLM estimate  (floor)
+//   (c) total_billed_cost_in_cents    — server billed snapshot USD (exact)
+//   (d) credits_spent × CREDIT_USD_RATE — credits at published rate (heuristic)
+//   (e) null → caller falls back to the token × LiteLLM estimate  (floor)
 // A value of 0/absent at a rung is treated as "not recorded here" and defers to
 // the next rung, so a zero never masks a real cost recorded elsewhere.
 function conversationCostUsd(rawConversationData: string): WarpCost | null {
@@ -244,6 +248,9 @@ function conversationCostUsd(rawConversationData: string): WarpCost | null {
     if (cents > 0) return { usd: cents / 100, estimated: false }
   }
 
+  const billedCents = safeNumber(meta.total_billed_cost_in_cents)
+  if (billedCents > 0) return { usd: billedCents / 100, estimated: false }
+
   const credits = safeNumber(meta.credits_spent)
   if (credits > 0) return { usd: credits * CREDIT_USD_RATE, estimated: true }
 

src/session-cache.ts

@@ -612,7 +612,9 @@ export const PROVIDER_PARSE_VERSIONS: Record<string, string> = {
   // cached before this hold costUSD: undefined and are re-priced from the
   // token floor on every read, so they must re-parse once for the real dollars
   // to land.
-  warp: 'worktree-project-grouping-v1-est-cost-billing-cost-v1',
+  // billed-cost-v1: total_billed_cost_in_cents prices conversations that carry
+  // no provider cost or charged usage.
+  warp: 'worktree-project-grouping-v1-est-cost-billing-cost-v1-billed-cost-v1',
   // cache-read-v1-est-cost: gen_metadata and RPC usage now carry cache-read
   // tokens, fields 9/10 read as thinking/response (they were swapped), and
   // placeholder-only models are priced and flagged costIsEstimated.

tests/providers/warp.test.ts

@@ -469,4 +469,41 @@ skipUnlessSqlite('warp provider', () => {
     expect(totalCost).toBeCloseTo(0.05, 6)
     expect(calls[0]!.costIsEstimated).toBe(false)
   })
+
+  it.each([
+    ['only total_billed_cost_in_cents', { total_billed_cost_in_cents: 3.5 }, 0.035],
+    ['only legacy total_provider_cost_in_cents', { total_provider_cost_in_cents: 2.5 }, 0.025],
+    ['both cost fields, legacy provider cost first', { total_provider_cost_in_cents: 2.5, total_billed_cost_in_cents: 3.5 }, 0.025],
+    ['charged usage ahead of the billed snapshot', { total_billed_cost_in_cents: 3.5, total_charged_usage: { input_cost_in_cents: 4 } }, 0.04],
+  ])('prices from %s', async (_label, costFields, expectedUsd) => {
+    const dbPath = createWarpDb(tmpDir)
+    withTestDb(dbPath, (db) => {
+      insertConversation(db, 'conv-billed', {
+        conversation_usage_metadata: {
+          ...costFields,
+          credits_spent: 1.44165,
+          token_usage: [
+            {
+              model_id: 'Claude Haiku 4.5',
+              warp_tokens: 51385,
+              byok_tokens: 0,
+              warp_token_usage_by_category: { primary_agent: 51385 },
+              byok_token_usage_by_category: {},
+            },
+          ],
+        },
+      })
+      insertQuery(db, {
+        exchangeId: 'ex-1',
+        conversationId: 'conv-billed',
+        startTs: '2026-10-18 10:00:00.000000',
+        input: JSON.stringify([{ Query: { text: 'do the thing' } }]),
+      })
+    })
+
+    const calls = await collectCalls(dbPath, 'conv-billed')
+    expect(calls).toHaveLength(1)
+    expect(calls[0]!.costUSD).toBeCloseTo(expectedUsd, 6)
+    expect(calls[0]!.costIsEstimated).toBe(false)
+  })
 })