Skip to content

phsl / new-api

Go to a project
Commit 92d3c9d1 by CaIon
Options

fix: bound uncached remainder by prompt-max(cached,write) and forward compact prompt_cache_key

parent 48068ce9
Showing with 34 additions and 18 deletions
dto/openai_response.go
...@@ -270,12 +270,17 @@ type InputTokenDetails struct { ...@@ -270,12 +270,17 @@ type InputTokenDetails struct {
// which field the upstream reported it in: Claude-derived conversions populate // which field the upstream reported it in: Claude-derived conversions populate
// CachedCreationTokens while OpenAI reports cache_write_tokens natively. Both // CachedCreationTokens while OpenAI reports cache_write_tokens natively. Both
// are billed at the cache-creation price; when both are present the larger // are billed at the cache-creation price; when both are present the larger
// value wins so the same tokens are never double-counted. // value wins so the same tokens are never double-counted. Negative upstream
// values are clamped to zero so they can never lower a charge.
func (d InputTokenDetails) CacheCreationTokensTotal() int { func (d InputTokenDetails) CacheCreationTokensTotal() int {
if d.CacheWriteTokens > d.CachedCreationTokens { total := d.CachedCreationTokens
return d.CacheWriteTokens if d.CacheWriteTokens > total {
total = d.CacheWriteTokens
} }
return d.CachedCreationTokens if total < 0 {
return 0
}
return total
} }
type OutputTokenDetails struct { type OutputTokenDetails struct {
......
relay/channel/openai/relay_image.go
...@@ -80,6 +80,7 @@ func normalizeOpenAIUsage(usage *dto.Usage) { ...@@ -80,6 +80,7 @@ func normalizeOpenAIUsage(usage *dto.Usage) {
if usage.InputTokensDetails != nil { if usage.InputTokensDetails != nil {
usage.PromptTokensDetails.CachedTokens = usage.InputTokensDetails.CachedTokens usage.PromptTokensDetails.CachedTokens = usage.InputTokensDetails.CachedTokens
usage.PromptTokensDetails.CachedCreationTokens = usage.InputTokensDetails.CachedCreationTokens usage.PromptTokensDetails.CachedCreationTokens = usage.InputTokensDetails.CachedCreationTokens
usage.PromptTokensDetails.CacheWriteTokens = usage.InputTokensDetails.CacheWriteTokens
usage.PromptTokensDetails.ImageTokens = usage.InputTokensDetails.ImageTokens usage.PromptTokensDetails.ImageTokens = usage.InputTokensDetails.ImageTokens
usage.PromptTokensDetails.TextTokens = usage.InputTokensDetails.TextTokens usage.PromptTokensDetails.TextTokens = usage.InputTokensDetails.TextTokens
usage.PromptTokensDetails.AudioTokens = usage.InputTokensDetails.AudioTokens usage.PromptTokensDetails.AudioTokens = usage.InputTokensDetails.AudioTokens
......
relay/responses_handler.go
...@@ -40,6 +40,11 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError * ...@@ -40,6 +40,11 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
case *dto.OpenAIResponsesRequest: case *dto.OpenAIResponsesRequest:
responsesReq = req responsesReq = req
case *dto.OpenAIResponsesCompactionRequest: case *dto.OpenAIResponsesCompactionRequest:
// Only fields documented for POST /v1/responses/compact are forwarded:
// model, input, instructions, previous_response_id, prompt_cache_key,
// prompt_cache_options, prompt_cache_retention, service_tier.
// Undocumented Codex-parity fields (tools, reasoning, text) are parsed
// for client compatibility but intentionally not sent upstream.
responsesReq = &dto.OpenAIResponsesRequest{ responsesReq = &dto.OpenAIResponsesRequest{
Model: req.Model, Model: req.Model,
Input: req.Input, Input: req.Input,
...@@ -47,6 +52,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError * ...@@ -47,6 +52,7 @@ func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *
PreviousResponseID: req.PreviousResponseID, PreviousResponseID: req.PreviousResponseID,
ParallelToolCalls: req.ParallelToolCalls, ParallelToolCalls: req.ParallelToolCalls,
ServiceTier: req.ServiceTier, ServiceTier: req.ServiceTier,
PromptCacheKey: req.PromptCacheKey,
PromptCacheOptions: req.PromptCacheOptions, PromptCacheOptions: req.PromptCacheOptions,
PromptCacheRetention: req.PromptCacheRetention, PromptCacheRetention: req.PromptCacheRetention,
} }
......
service/relayconvert/internal/oai_chat/to_claude_messages_resp.go
...@@ -44,7 +44,8 @@ func buildClaudeUsageFromOpenAIUsage(oaiUsage *dto.Usage) *dto.ClaudeUsage { ...@@ -44,7 +44,8 @@ func buildClaudeUsageFromOpenAIUsage(oaiUsage *dto.Usage) *dto.ClaudeUsage {
if oaiUsage.PromptTokensDetails.CacheWriteTokens > 0 { if oaiUsage.PromptTokensDetails.CacheWriteTokens > 0 {
// OpenAI native cache-write usage counts cached and cache-write tokens // OpenAI native cache-write usage counts cached and cache-write tokens
// inside prompt_tokens, while Claude semantics reports input_tokens // inside prompt_tokens, while Claude semantics reports input_tokens
// excluding both; the uncached remainder clamps at zero. // excluding both. Both counts are unadjusted prefixes and may overlap,
// so clamp a negative remainder at zero.
inputTokens = oaiUsage.PromptTokens - oaiUsage.PromptTokensDetails.CachedTokens - cacheCreationTokens inputTokens = oaiUsage.PromptTokens - oaiUsage.PromptTokensDetails.CachedTokens - cacheCreationTokens
if inputTokens < 0 { if inputTokens < 0 {
inputTokens = 0 inputTokens = 0
......
service/relayconvert/internal/oai_chat/to_claude_messages_resp_test.go
...@@ -92,7 +92,7 @@ func TestBuildClaudeUsageFromOpenAICacheWriteUsage(t *testing.T) { ...@@ -92,7 +92,7 @@ func TestBuildClaudeUsageFromOpenAICacheWriteUsage(t *testing.T) {
require.NotNil(t, usage) require.NotNil(t, usage)
// Claude semantics reports input_tokens excluding cache read/write; the // Claude semantics reports input_tokens excluding cache read/write; the
// remainder 3619-2921-3616 clamps to 0. // overlapping unadjusted prefixes drive the remainder negative, clamp to 0.
assert.Equal(t, 0, usage.InputTokens) assert.Equal(t, 0, usage.InputTokens)
assert.Equal(t, 2921, usage.CacheReadInputTokens) assert.Equal(t, 2921, usage.CacheReadInputTokens)
assert.Equal(t, 3616, usage.CacheCreationInputTokens) assert.Equal(t, 3616, usage.CacheCreationInputTokens)
......
service/text_quota.go
...@@ -294,9 +294,10 @@ func calculateTextQuotaSummary(ctx *gin.Context, relayInfo *relaycommon.RelayInf ...@@ -294,9 +294,10 @@ func calculateTextQuotaSummary(ctx *gin.Context, relayInfo *relaycommon.RelayInf
} }
} }
// OpenAI cache-write usage can report cached_tokens + cache_write_tokens // OpenAI cache-write usage reports unadjusted prefix counts, so
// exceeding prompt_tokens; the uncached remainder must clamp at zero so // cached_tokens + cache_write_tokens can exceed prompt_tokens and the
// billing never subtracts more than the reported input. // remainder can go negative. Clamp at zero so overlap never turns into
// a negative base charge.
if baseTokens.IsNegative() { if baseTokens.IsNegative() {
baseTokens = decimal.Zero baseTokens = decimal.Zero
} }
......
service/text_quota_test.go
...@@ -411,8 +411,8 @@ func TestCalculateTextQuotaSummaryBillsOpenAICacheWriteTokens(t *testing.T) { ...@@ -411,8 +411,8 @@ func TestCalculateTextQuotaSummaryBillsOpenAICacheWriteTokens(t *testing.T) {
t.Run("uncached remainder clamps to zero", func(t *testing.T) { t.Run("uncached remainder clamps to zero", func(t *testing.T) {
// Real OpenAI payload shape: cached_tokens + cache_write_tokens exceeds // Real OpenAI payload shape: cached_tokens + cache_write_tokens exceeds
// prompt_tokens, so the uncached remainder must clamp to 0 instead of // prompt_tokens because both are unadjusted prefix counts. The negative
// producing a negative charge component. // remainder must clamp to zero, never turn into a negative base charge.
usage := &dto.Usage{ usage := &dto.Usage{
PromptTokens: 3619, PromptTokens: 3619,
CompletionTokens: 36, CompletionTokens: 36,
......
service/tiered_settle.go
...@@ -67,6 +67,8 @@ func BuildTieredTokenParams(usage *dto.Usage, isClaudeUsageSemantic bool, usedVa ...@@ -67,6 +67,8 @@ func BuildTieredTokenParams(usage *dto.Usage, isClaudeUsageSemantic bool, usedVa
} }
} }
// OpenAI cache-write usage reports unadjusted prefix counts, so cr + cc can
// exceed the prompt and drive the remainder negative. Clamp at zero.
if p < 0 { if p < 0 {
p = 0 p = 0
} }
......
web/default/src/components/ui/sidebar.tsx
...@@ -444,7 +444,7 @@ function SidebarGroupAction({ ...@@ -444,7 +444,7 @@ function SidebarGroupAction({
props: mergeProps<'button'>( props: mergeProps<'button'>(
{ {
className: cn( className: cn(
'absolute top-3.5 right-3 flex aspect-square w-5 items-center justify-center rounded-md p-0 text-sidebar-foreground ring-sidebar-ring outline-hidden transition-transform group-data-[collapsible=icon]:hidden after:absolute after:-inset-2 hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:ring-2 md:after:hidden [&>svg]:size-4 [&>svg]:shrink-0', 'absolute top-3.5 right-3 flex aspect-square w-5 items-center justify-center rounded-md p-0 text-sidebar-foreground ring-sidebar-ring outline-hidden transition-transform group-data-[collapsible=icon]:hidden after:absolute after:-inset-2 hover:bg-muted hover:text-sidebar-foreground focus-visible:ring-2 md:after:hidden [&>svg]:size-4 [&>svg]:shrink-0',
className className
), ),
}, },
...@@ -495,13 +495,13 @@ function SidebarMenuItem({ className, ...props }: React.ComponentProps<'li'>) { ...@@ -495,13 +495,13 @@ function SidebarMenuItem({ className, ...props }: React.ComponentProps<'li'>) {
} }
const sidebarMenuButtonVariants = cva( const sidebarMenuButtonVariants = cva(
'peer/menu-button group/menu-button flex w-full items-center gap-2 overflow-hidden rounded-md p-2 text-left text-sm ring-sidebar-ring outline-hidden transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:size-8! group-data-[collapsible=icon]:p-2! hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:ring-2 active:bg-sidebar-accent active:text-sidebar-accent-foreground disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 data-open:hover:bg-sidebar-accent data-open:hover:text-sidebar-accent-foreground data-active:bg-sidebar-accent data-active:font-medium data-active:text-sidebar-accent-foreground [&_svg]:size-4 [&_svg]:shrink-0 [&>span:last-child]:truncate', 'peer/menu-button group/menu-button flex w-full items-center gap-2 overflow-hidden rounded-md p-2 text-left text-sm ring-sidebar-ring outline-hidden transition-[width,height,padding] group-has-data-[sidebar=menu-action]/menu-item:pr-8 group-data-[collapsible=icon]:size-8! group-data-[collapsible=icon]:p-2! hover:bg-muted hover:text-sidebar-foreground focus-visible:ring-2 active:bg-sidebar-accent active:text-sidebar-accent-foreground disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 data-open:hover:bg-muted data-open:hover:text-sidebar-foreground data-active:bg-sidebar-accent data-active:font-medium data-active:text-sidebar-accent-foreground data-active:hover:bg-sidebar-accent data-active:hover:text-sidebar-accent-foreground [&_svg]:size-4 [&_svg]:shrink-0 [&>span:last-child]:truncate',
{ {
variants: { variants: {
variant: { variant: {
default: 'hover:bg-sidebar-accent hover:text-sidebar-accent-foreground', default: 'hover:bg-muted hover:text-sidebar-foreground',
outline: outline:
'bg-background shadow-[0_0_0_1px_var(--sidebar-border)] hover:bg-sidebar-accent hover:text-sidebar-accent-foreground hover:shadow-[0_0_0_1px_var(--sidebar-accent)]', 'bg-background shadow-[0_0_0_1px_var(--sidebar-border)] hover:bg-muted hover:text-sidebar-foreground hover:shadow-[0_0_0_1px_var(--muted)]',
}, },
size: { size: {
default: 'h-8 text-sm', default: 'h-8 text-sm',
...@@ -586,7 +586,7 @@ function SidebarMenuAction({ ...@@ -586,7 +586,7 @@ function SidebarMenuAction({
props: mergeProps<'button'>( props: mergeProps<'button'>(
{ {
className: cn( className: cn(
'absolute top-1.5 right-1 flex aspect-square w-5 items-center justify-center rounded-md p-0 text-sidebar-foreground ring-sidebar-ring outline-hidden transition-transform group-data-[collapsible=icon]:hidden peer-hover/menu-button:text-sidebar-accent-foreground peer-data-[size=default]/menu-button:top-1.5 peer-data-[size=lg]/menu-button:top-2.5 peer-data-[size=sm]/menu-button:top-1 after:absolute after:-inset-2 hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:ring-2 md:after:hidden [&>svg]:size-4 [&>svg]:shrink-0', 'absolute top-1.5 right-1 flex aspect-square w-5 items-center justify-center rounded-md p-0 text-sidebar-foreground ring-sidebar-ring outline-hidden transition-transform group-data-[collapsible=icon]:hidden peer-hover/menu-button:text-sidebar-foreground peer-data-[size=default]/menu-button:top-1.5 peer-data-[size=lg]/menu-button:top-2.5 peer-data-[size=sm]/menu-button:top-1 after:absolute after:-inset-2 hover:bg-muted hover:text-sidebar-foreground focus-visible:ring-2 md:after:hidden [&>svg]:size-4 [&>svg]:shrink-0',
showOnHover && showOnHover &&
'group-focus-within/menu-item:opacity-100 group-hover/menu-item:opacity-100 peer-data-active/menu-button:text-sidebar-accent-foreground aria-expanded:opacity-100 md:opacity-0', 'group-focus-within/menu-item:opacity-100 group-hover/menu-item:opacity-100 peer-data-active/menu-button:text-sidebar-accent-foreground aria-expanded:opacity-100 md:opacity-0',
className className
...@@ -611,7 +611,7 @@ function SidebarMenuBadge({ ...@@ -611,7 +611,7 @@ function SidebarMenuBadge({
data-slot='sidebar-menu-badge' data-slot='sidebar-menu-badge'
data-sidebar='menu-badge' data-sidebar='menu-badge'
className={cn( className={cn(
'text-sidebar-foreground peer-hover/menu-button:text-sidebar-accent-foreground peer-data-active/menu-button:text-sidebar-accent-foreground pointer-events-none absolute right-1 flex h-5 min-w-5 items-center justify-center rounded-md px-1 text-xs font-medium tabular-nums select-none group-data-[collapsible=icon]:hidden peer-data-[size=default]/menu-button:top-1.5 peer-data-[size=lg]/menu-button:top-2.5 peer-data-[size=sm]/menu-button:top-1', 'text-sidebar-foreground peer-hover/menu-button:text-sidebar-foreground peer-data-active/menu-button:text-sidebar-accent-foreground pointer-events-none absolute right-1 flex h-5 min-w-5 items-center justify-center rounded-md px-1 text-xs font-medium tabular-nums select-none group-data-[collapsible=icon]:hidden peer-data-[size=default]/menu-button:top-1.5 peer-data-[size=lg]/menu-button:top-2.5 peer-data-[size=sm]/menu-button:top-1',
className className
)} )}
{...props} {...props}
...@@ -701,7 +701,7 @@ function SidebarMenuSubButton({ ...@@ -701,7 +701,7 @@ function SidebarMenuSubButton({
props: mergeProps<'a'>( props: mergeProps<'a'>(
{ {
className: cn( className: cn(
'flex h-7 min-w-0 -translate-x-px items-center gap-2 overflow-hidden rounded-md px-2 text-sidebar-foreground ring-sidebar-ring outline-hidden group-data-[collapsible=icon]:hidden hover:bg-sidebar-accent hover:text-sidebar-accent-foreground focus-visible:ring-2 active:bg-sidebar-accent active:text-sidebar-accent-foreground disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 data-[size=md]:text-sm data-[size=sm]:text-xs data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground [&>span:last-child]:truncate [&>svg]:size-4 [&>svg]:shrink-0 [&>svg]:text-sidebar-accent-foreground', 'flex h-7 min-w-0 -translate-x-px items-center gap-2 overflow-hidden rounded-md px-2 text-sidebar-foreground ring-sidebar-ring outline-hidden group-data-[collapsible=icon]:hidden hover:bg-muted hover:text-sidebar-foreground focus-visible:ring-2 active:bg-sidebar-accent active:text-sidebar-accent-foreground disabled:pointer-events-none disabled:opacity-50 aria-disabled:pointer-events-none aria-disabled:opacity-50 data-[size=md]:text-sm data-[size=sm]:text-xs data-active:bg-sidebar-accent data-active:text-sidebar-accent-foreground data-active:hover:bg-sidebar-accent data-active:hover:text-sidebar-accent-foreground [&>span:last-child]:truncate [&>svg]:size-4 [&>svg]:shrink-0 [&>svg]:text-sidebar-accent-foreground',
className className
), ),
}, },
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or sign in to comment