Files
new-api/setting/ratio_setting/cache_ratio.go
T
Calcium-Ion 86ac0f7745 refactor: extract protocol conversion layer into standalone relaykit module (#6369)
* test(relayconvert): add golden snapshot matrix and relaykit boundary guard

Phase 0 of the relaykit extraction plan: pin byte-level output of every
registered (from,to) request/response/stream conversion route, and
forbid kit-bound packages from growing host-only imports.

* wip(relayconvert): drop gin.Context from converter signatures; add convmeta draft

Phase 1 in progress: relayconvert now takes context.Context; host media
resolver adapts gin.Context back at the service boundary.

* refactor(relayconvert): decouple converters from RelayInfo, gin, and settings

Phase 1 of the relaykit extraction plan:
- converters now depend on convmeta.Meta (implemented by RelayInfo) instead
  of *relaycommon.RelayInfo; ClaudeConvertInfo and the format guesser move
  to convmeta with aliases left behind
- host settings reach converters via a convmeta.Options snapshot built in
  RelayInfo.ConvOptions; no more model_setting/reasoning global reads inside
  the conversion layer
- effort-suffix helpers move to service/relayconvert/reasoning (old package
  forwards); chat-to-responses upgrade policy moves to service (host routing
  logic, not conversion)
- golden conversion matrix unchanged

* test(relayconvert): tighten boundary — kit packages now free of gin/setting imports

* refactor(dto): drop gin and logger dependencies

Phase 2 (part 1): dto.Request.IsStream now takes *http.Request instead of
*gin.Context (Gemini's impl reads query/path off the std request); dto's
three logger calls become common.SysError. Boundary test allowlist is now
empty — kit-bound packages import no gin/setting/logger/model.

* refactor(kit): extract dependency-free kitutil; dto/types/relayconvert stop importing common

Phase 2 of the relaykit extraction plan:
- new service/relayconvert/kitutil holds the pure helpers the kit needs
  (JSON wrappers, pointer/string/uuid/timestamp utils, MaskSensitiveInfo,
  pluggable LogInfo/LogError hooks, Debug flag)
- dto, types, and all relayconvert packages now use kitutil; their only
  remaining internal deps are dto/types/constant
- common keeps every original symbol (MaskSensitiveInfo delegates to
  kitutil) so host code is untouched; main.go routes kit logging into
  common.SysLog/SysError and mirrors DebugEnabled
- golden conversion matrix unchanged

* refactor(kit): move EndpointType/FinishReason to types; OpenRouter dialect via Options

Kit packages (dto/types/relayconvert/reasonmap) no longer import constant:
- EndpointType and finish-reason values live in types; constant re-exports
- the OpenRouter special-case in claude->openai request conversion reads
  Options.OpenRouterDialect, set by the host from the channel type;
  InitChannelMeta invalidates the cached snapshot on channel switch

* refactor: extract relaykit submodule (dto/types/relayconvert/reasonmap)

Phase 3 of the relaykit extraction plan:
- new go module github.com/QuantumNous/new-api/relaykit containing dto
  (minus task family), types, relayconvert (with convmeta/kitutil/reasoning),
  and reasonmap; host consumes it via require + replace, go.work for dev
- task-family dto (task/suno/midjourney/video) stays in the host dto
  package; dual-consumer host files alias it as taskdto
- relaykit builds and tests standalone (GOWORK=off): no host imports,
  no gin, no DB, no settings
- golden conversion matrix unchanged

* build(docker): copy relaykit/go.mod before go mod download

The local-replace submodule's go.mod must exist inside the build context
for the main module graph to resolve.

* fix: address relaykit extraction regressions

* fix: address relaykit review regressions

* docs: document Meta nil receiver contract

* fix(relaykit): fail OpenAI→Claude conversion without max_tokens; reject negative default_max_tokens

The Claude Messages API requires max_tokens (omitting it is a 400
"Field required"), but with a nil Options.Claude.DefaultMaxTokens hook
the converters silently emitted a request the upstream is guaranteed to
reject. Both OpenAI Chat and Responses → Claude conversions now return
sharedclaude.ErrMissingMaxTokens when no path (client value, default
hook, thinking-adapter floor) supplied one. Unreachable in the host,
which always configures the hook.

Host side, claude.default_max_tokens now rejects negative values at the
option API before persisting — they would wrap into huge unsigned values
during conversion. Zero stays allowed: the current API treats
max_tokens: 0 as cache pre-warming.

* fix: make Gemini safety settings read path race-free
2026-07-27 15:56:21 +08:00

182 lines
7.0 KiB
Go

package ratio_setting
import (
"github.com/QuantumNous/new-api/relaykit/types"
)
var defaultCacheRatio = map[string]float64{
"gemini-3-flash-preview": 0.1,
"gemini-3-pro-preview": 0.1,
"gemini-3.1-pro-preview": 0.1,
"gpt-4": 0.5,
"o1": 0.5,
"o1-2024-12-17": 0.5,
"o1-preview-2024-09-12": 0.5,
"o1-preview": 0.5,
"o1-mini-2024-09-12": 0.5,
"o1-mini": 0.5,
"o3-mini": 0.5,
"o3-mini-2025-01-31": 0.5,
"gpt-4o-2024-11-20": 0.5,
"gpt-4o-2024-08-06": 0.5,
"gpt-4o": 0.5,
"gpt-4o-mini-2024-07-18": 0.5,
"gpt-4o-mini": 0.5,
"gpt-4o-realtime-preview": 0.5,
"gpt-4o-mini-realtime-preview": 0.5,
"gpt-4.5-preview": 0.5,
"gpt-4.5-preview-2025-02-27": 0.5,
"gpt-4.1": 0.25,
"gpt-4.1-mini": 0.25,
"gpt-4.1-nano": 0.25,
"gpt-5": 0.1,
"gpt-5-2025-08-07": 0.1,
"gpt-5-chat-latest": 0.1,
"gpt-5-mini": 0.1,
"gpt-5-mini-2025-08-07": 0.1,
"gpt-5-nano": 0.1,
"gpt-5-nano-2025-08-07": 0.1,
"deepseek-chat": 0.25,
"deepseek-reasoner": 0.25,
"deepseek-coder": 0.25,
"claude-3-sonnet-20240229": 0.1,
"claude-3-opus-20240229": 0.1,
"claude-3-haiku-20240307": 0.1,
"claude-3-5-haiku-20241022": 0.1,
"claude-haiku-4-5-20251001": 0.1,
"claude-3-5-sonnet-20240620": 0.1,
"claude-3-5-sonnet-20241022": 0.1,
"claude-3-7-sonnet-20250219": 0.1,
"claude-3-7-sonnet-20250219-thinking": 0.1,
"claude-sonnet-4-20250514": 0.1,
"claude-sonnet-4-20250514-thinking": 0.1,
"claude-opus-4-20250514": 0.1,
"claude-opus-4-20250514-thinking": 0.1,
"claude-opus-4-1-20250805": 0.1,
"claude-opus-4-1-20250805-thinking": 0.1,
"claude-sonnet-4-5-20250929": 0.1,
"claude-sonnet-4-5-20250929-thinking": 0.1,
"claude-opus-4-5-20251101": 0.1,
"claude-opus-4-5-20251101-thinking": 0.1,
"claude-opus-4-6": 0.1,
"claude-opus-4-6-thinking": 0.1,
"claude-opus-4-6-max": 0.1,
"claude-opus-4-6-high": 0.1,
"claude-opus-4-6-medium": 0.1,
"claude-opus-4-6-low": 0.1,
"claude-opus-4-7": 0.1,
"claude-opus-4-7-thinking": 0.1,
"claude-opus-4-7-max": 0.1,
"claude-opus-4-7-xhigh": 0.1,
"claude-opus-4-7-high": 0.1,
"claude-opus-4-7-medium": 0.1,
"claude-opus-4-7-low": 0.1,
"claude-opus-4-8": 0.1,
"claude-opus-4-8-thinking": 0.1,
"claude-opus-4-8-max": 0.1,
"claude-opus-4-8-xhigh": 0.1,
"claude-opus-4-8-high": 0.1,
"claude-opus-4-8-medium": 0.1,
"claude-opus-4-8-low": 0.1,
}
var defaultCreateCacheRatio = map[string]float64{
"gpt-5.6-sol": 1.25,
"gpt-5.6-terra": 1.25,
"gpt-5.6-luna": 1.25,
"claude-3-sonnet-20240229": 1.25,
"claude-3-opus-20240229": 1.25,
"claude-3-haiku-20240307": 1.25,
"claude-3-5-haiku-20241022": 1.25,
"claude-haiku-4-5-20251001": 1.25,
"claude-3-5-sonnet-20240620": 1.25,
"claude-3-5-sonnet-20241022": 1.25,
"claude-3-7-sonnet-20250219": 1.25,
"claude-3-7-sonnet-20250219-thinking": 1.25,
"claude-sonnet-4-20250514": 1.25,
"claude-sonnet-4-20250514-thinking": 1.25,
"claude-opus-4-20250514": 1.25,
"claude-opus-4-20250514-thinking": 1.25,
"claude-opus-4-1-20250805": 1.25,
"claude-opus-4-1-20250805-thinking": 1.25,
"claude-sonnet-4-5-20250929": 1.25,
"claude-sonnet-4-5-20250929-thinking": 1.25,
"claude-opus-4-5-20251101": 1.25,
"claude-opus-4-5-20251101-thinking": 1.25,
"claude-opus-4-6": 1.25,
"claude-opus-4-6-thinking": 1.25,
"claude-opus-4-6-max": 1.25,
"claude-opus-4-6-high": 1.25,
"claude-opus-4-6-medium": 1.25,
"claude-opus-4-6-low": 1.25,
"claude-opus-4-7": 1.25,
"claude-opus-4-7-thinking": 1.25,
"claude-opus-4-7-max": 1.25,
"claude-opus-4-7-xhigh": 1.25,
"claude-opus-4-7-high": 1.25,
"claude-opus-4-7-medium": 1.25,
"claude-opus-4-7-low": 1.25,
"claude-opus-4-8": 1.25,
"claude-opus-4-8-thinking": 1.25,
"claude-opus-4-8-max": 1.25,
"claude-opus-4-8-xhigh": 1.25,
"claude-opus-4-8-high": 1.25,
"claude-opus-4-8-medium": 1.25,
"claude-opus-4-8-low": 1.25,
}
//var defaultCreateCacheRatio = map[string]float64{}
var cacheRatioMap = types.NewRWMap[string, float64]()
var createCacheRatioMap = types.NewRWMap[string, float64]()
// GetCacheRatioMap returns a copy of the cache ratio map
func GetCacheRatioMap() map[string]float64 {
return cacheRatioMap.ReadAll()
}
// CacheRatio2JSONString converts the cache ratio map to a JSON string
func CacheRatio2JSONString() string {
return cacheRatioMap.MarshalJSONString()
}
// CreateCacheRatio2JSONString converts the create cache ratio map to a JSON string
func CreateCacheRatio2JSONString() string {
return createCacheRatioMap.MarshalJSONString()
}
// UpdateCacheRatioByJSONString updates the cache ratio map from a JSON string
func UpdateCacheRatioByJSONString(jsonStr string) error {
return types.LoadFromJsonStringWithCallback(cacheRatioMap, jsonStr, InvalidateExposedDataCache)
}
// UpdateCreateCacheRatioByJSONString updates the create cache ratio map from a JSON string
func UpdateCreateCacheRatioByJSONString(jsonStr string) error {
return types.LoadFromJsonStringWithCallback(createCacheRatioMap, jsonStr, InvalidateExposedDataCache)
}
// GetCacheRatio returns the cache ratio for a model
func GetCacheRatio(name string) (float64, bool) {
ratio, ok := cacheRatioMap.Get(name)
if !ok {
return 1, false // Default to 1 if not found
}
return ratio, true
}
func GetCreateCacheRatio(name string) (float64, bool) {
ratio, ok := createCacheRatioMap.Get(name)
if !ok {
return 1.25, false // Default to 1.25 if not found
}
return ratio, true
}
func GetCacheRatioCopy() map[string]float64 {
return cacheRatioMap.ReadAll()
}
func GetCreateCacheRatioCopy() map[string]float64 {
return createCacheRatioMap.ReadAll()
}