Cloud stats: the CLI now reports Anthropic prompt-caching breakdown (cache-creation and cache-read token counts) alongside the existing input/output/cost totals, so the dashboard can show "saved $X with caching" for CLI and VS Code (ACP) sessions.
Changed
- Cache token reporting in cloud stats.
StatsPayload
(src/utils/codeepCloud.ts) now carries optional cacheCreationTokens / cacheReadTokens. The three reportStats call sites — src/renderer/main.ts (sync and async paths) and src/renderer/agentExecution.ts — pass through the per-model buckets from getCostBreakdown(), and getSessionStats() exposes session-totals via new totalCacheCreationTokens / totalCacheReadTokens fields. The ACP server (src/acp/server.ts) — the path VS Code uses — was updated in lockstep so both clients report the same shape.
Why: estimatedCost already folded cache multipliers into the dollar total, but the raw cache counts were dropped on the floor, so the dashboard couldn't break out "this session read 500k tokens from cache." Now it can.
ProviderCostBreakdownandSessionTokenStatsextended. Both
interfaces gained cacheCreationTokens / cacheReadTokens (required on the breakdown, since every provider reports something — 0 for non-caching). getCostBreakdown() accumulates them per provider/model group; getSessionStats() sums across all records. Two new tests cover the accumulation paths.