feat(thinking): enhance adaptive thinking support across models and update test cases

This commit is contained in:
hkfires
2026-03-03 13:00:24 +08:00
parent 532107b4fa
commit d2e5857b82
5 changed files with 607 additions and 93 deletions
@@ -14,6 +14,7 @@ import (
"strings"
"github.com/google/uuid"
"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
"github.com/router-for-me/CLIProxyAPI/v6/internal/thinking"
"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
"github.com/tidwall/gjson"
@@ -115,24 +116,73 @@ func ConvertGeminiRequestToClaude(modelName string, inputRawJSON []byte, stream
// Include thoughts configuration for reasoning process visibility
// Translator only does format conversion, ApplyThinking handles model capability validation.
if thinkingConfig := genConfig.Get("thinkingConfig"); thinkingConfig.Exists() && thinkingConfig.IsObject() {
hasLevel := func(levels []string, target string) bool {
for _, level := range levels {
if strings.EqualFold(strings.TrimSpace(level), target) {
return true
}
}
return false
}
mi := registry.LookupModelInfo(modelName, "claude")
supportsAdaptive := mi != nil && mi.Thinking != nil && len(mi.Thinking.Levels) > 0
supportsMax := supportsAdaptive && hasLevel(mi.Thinking.Levels, "max")
mapToEffort := func(level string) (string, bool) {
level = strings.ToLower(strings.TrimSpace(level))
switch level {
case "":
return "", false
case "minimal":
return "low", true
case "low", "medium", "high":
return level, true
case "xhigh", "max":
if supportsMax {
return "max", true
}
return "high", true
case "auto":
return "high", true
default:
return "", false
}
}
thinkingLevel := thinkingConfig.Get("thinkingLevel")
if !thinkingLevel.Exists() {
thinkingLevel = thinkingConfig.Get("thinking_level")
}
if thinkingLevel.Exists() {
level := strings.ToLower(strings.TrimSpace(thinkingLevel.String()))
switch level {
case "":
case "none":
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
case "auto":
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
default:
if budget, ok := thinking.ConvertLevelToBudget(level); ok {
if supportsAdaptive {
switch level {
case "":
case "none":
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
out, _ = sjson.Delete(out, "output_config.effort")
default:
effort, ok := mapToEffort(level)
if ok {
out, _ = sjson.Set(out, "thinking.type", "adaptive")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
out, _ = sjson.Set(out, "output_config.effort", effort)
}
}
} else {
switch level {
case "":
case "none":
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
case "auto":
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Set(out, "thinking.budget_tokens", budget)
out, _ = sjson.Delete(out, "thinking.budget_tokens")
default:
if budget, ok := thinking.ConvertLevelToBudget(level); ok {
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Set(out, "thinking.budget_tokens", budget)
}
}
}
} else {
@@ -142,16 +192,35 @@ func ConvertGeminiRequestToClaude(modelName string, inputRawJSON []byte, stream
}
if thinkingBudget.Exists() {
budget := int(thinkingBudget.Int())
switch budget {
case 0:
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
case -1:
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
default:
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Set(out, "thinking.budget_tokens", budget)
if supportsAdaptive {
switch budget {
case 0:
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
out, _ = sjson.Delete(out, "output_config.effort")
default:
level, ok := thinking.ConvertBudgetToLevel(budget)
if ok {
effort, ok := mapToEffort(level)
if ok {
out, _ = sjson.Set(out, "thinking.type", "adaptive")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
out, _ = sjson.Set(out, "output_config.effort", effort)
}
}
}
} else {
switch budget {
case 0:
out, _ = sjson.Set(out, "thinking.type", "disabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
case -1:
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Delete(out, "thinking.budget_tokens")
default:
out, _ = sjson.Set(out, "thinking.type", "enabled")
out, _ = sjson.Set(out, "thinking.budget_tokens", budget)
}
}
} else if includeThoughts := thinkingConfig.Get("includeThoughts"); includeThoughts.Exists() && includeThoughts.Type == gjson.True {
out, _ = sjson.Set(out, "thinking.type", "enabled")
@@ -238,7 +238,7 @@ func ConvertClaudeRequestToCodex(modelName string, inputRawJSON []byte, _ bool)
effort = strings.ToLower(strings.TrimSpace(v.String()))
}
switch effort {
case "low", "medium", "high":
case "minimal", "low", "medium", "high":
reasoningEffort = effort
case "max":
reasoningEffort = string(thinking.LevelXHigh)
@@ -9,6 +9,7 @@ import (
"bytes"
"strings"
"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
"github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/common"
"github.com/tidwall/gjson"
"github.com/tidwall/sjson"
@@ -151,7 +152,7 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
}
}
// Map Anthropic thinking -> Gemini thinkingBudget/include_thoughts when enabled
// Map Anthropic thinking -> Gemini thinking config when enabled
// Translator only does format conversion, ApplyThinking handles model capability validation.
if t := gjson.GetBytes(rawJSON, "thinking"); t.Exists() && t.IsObject() {
switch t.Get("type").String() {
@@ -162,9 +163,31 @@ func ConvertClaudeRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.includeThoughts", true)
}
case "adaptive", "auto":
// Keep adaptive/auto as a high level sentinel; ApplyThinking resolves it
// to model-specific max capability.
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingLevel", "high")
// For adaptive thinking:
// - If output_config.effort is explicitly present, map it to thinkingLevel.
// - Otherwise, treat it as "enabled with target-model maximum" and emit thinkingBudget=max.
effort := ""
if v := gjson.GetBytes(rawJSON, "output_config.effort"); v.Exists() && v.Type == gjson.String {
effort = strings.ToLower(strings.TrimSpace(v.String()))
}
if effort != "" {
level := effort
switch level {
case "xhigh", "max":
level = "high"
}
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingLevel", level)
} else {
maxBudget := 0
if mi := registry.LookupModelInfo(modelName, "gemini"); mi != nil && mi.Thinking != nil {
maxBudget = mi.Thinking.Max
}
if maxBudget > 0 {
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingBudget", maxBudget)
} else {
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.thinkingLevel", "high")
}
}
out, _ = sjson.Set(out, "generationConfig.thinkingConfig.includeThoughts", true)
}
}
@@ -83,7 +83,7 @@ func ConvertClaudeRequestToOpenAI(modelName string, inputRawJSON []byte, stream
effort = strings.ToLower(strings.TrimSpace(v.String()))
}
switch effort {
case "low", "medium", "high":
case "minimal", "low", "medium", "high":
out, _ = sjson.Set(out, "reasoning_effort", effort)
case "max":
out, _ = sjson.Set(out, "reasoning_effort", string(thinking.LevelXHigh))