fix(kimi): update base URL and integrate ClaudeExecutor fallback

- Updated `KimiAPIBaseURL` to remove versioning from the root path.
- Integrated `ClaudeExecutor` fallback in `KimiExecutor` methods for compatibility with Claude requests.
- Simplified token counting by delegating to `ClaudeExecutor`.

internal/auth/kimi/kimi.go

@@ -30,7 +30,7 @@ const (
 	// kimiTokenURL is the endpoint for exchanging device codes for tokens.
 	kimiTokenURL = kimiOAuthHost + "/api/oauth/token"
 	// KimiAPIBaseURL is the base URL for Kimi API requests.
-	KimiAPIBaseURL = "https://api.kimi.com/coding/v1"
+	KimiAPIBaseURL = "https://api.kimi.com/coding"
 	// defaultPollInterval is the default interval for polling token endpoint.
 	defaultPollInterval = 5 * time.Second
 	// maxPollDuration is the maximum time to wait for user authorization.

internal/runtime/executor/kimi_executor.go

@@ -25,6 +25,7 @@ import (
 
 // KimiExecutor is a stateless executor for Kimi API using OpenAI-compatible chat completions.
 type KimiExecutor struct {
+	ClaudeExecutor
 	cfg *config.Config
 }
 
@@ -64,6 +65,12 @@ func (e *KimiExecutor) HttpRequest(ctx context.Context, auth *cliproxyauth.Auth,
 
 // Execute performs a non-streaming chat completion request to Kimi.
 func (e *KimiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (resp cliproxyexecutor.Response, err error) {
+	from := opts.SourceFormat
+	if from.String() == "claude" {
+		auth.Attributes["base_url"] = kimiauth.KimiAPIBaseURL
+		return e.ClaudeExecutor.Execute(ctx, auth, req, opts)
+	}
+
 	baseModel := thinking.ParseSuffix(req.Model).ModelName
 
 	token := kimiCreds(auth)
@@ -71,7 +78,6 @@ func (e *KimiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	reporter := newUsageReporter(ctx, e.Identifier(), baseModel, auth)
 	defer reporter.trackFailure(ctx, &err)
 
-	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
 	originalPayload := bytes.Clone(req.Payload)
 	if len(opts.OriginalRequest) > 0 {
@@ -95,7 +101,7 @@ func (e *KimiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	requestedModel := payloadRequestedModel(opts, req.Model)
 	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
-	url := kimiauth.KimiAPIBaseURL + "/chat/completions"
+	url := kimiauth.KimiAPIBaseURL + "/v1/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
 	if err != nil {
 		return resp, err
@@ -155,14 +161,18 @@ func (e *KimiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 
 // ExecuteStream performs a streaming chat completion request to Kimi.
 func (e *KimiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (stream <-chan cliproxyexecutor.StreamChunk, err error) {
-	baseModel := thinking.ParseSuffix(req.Model).ModelName
+	from := opts.SourceFormat
+	if from.String() == "claude" {
+		auth.Attributes["base_url"] = kimiauth.KimiAPIBaseURL
+		return e.ClaudeExecutor.ExecuteStream(ctx, auth, req, opts)
+	}
 
+	baseModel := thinking.ParseSuffix(req.Model).ModelName
 	token := kimiCreds(auth)
 
 	reporter := newUsageReporter(ctx, e.Identifier(), baseModel, auth)
 	defer reporter.trackFailure(ctx, &err)
 
-	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
 	originalPayload := bytes.Clone(req.Payload)
 	if len(opts.OriginalRequest) > 0 {
@@ -190,7 +200,7 @@ func (e *KimiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 	requestedModel := payloadRequestedModel(opts, req.Model)
 	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
-	url := kimiauth.KimiAPIBaseURL + "/chat/completions"
+	url := kimiauth.KimiAPIBaseURL + "/v1/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
 	if err != nil {
 		return nil, err
@@ -269,26 +279,8 @@ func (e *KimiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 
 // CountTokens estimates token count for Kimi requests.
 func (e *KimiExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
-	baseModel := thinking.ParseSuffix(req.Model).ModelName
+	auth.Attributes["base_url"] = kimiauth.KimiAPIBaseURL
-
+	return e.ClaudeExecutor.CountTokens(ctx, auth, req, opts)
-	from := opts.SourceFormat
-	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
-
-	// Use a generic tokenizer for estimation
-	enc, err := tokenizerForModel("gpt-4")
-	if err != nil {
-		return cliproxyexecutor.Response{}, fmt.Errorf("kimi executor: tokenizer init failed: %w", err)
-	}
-
-	count, err := countOpenAIChatTokens(enc, body)
-	if err != nil {
-		return cliproxyexecutor.Response{}, fmt.Errorf("kimi executor: token counting failed: %w", err)
-	}
-
-	usageJSON := buildOpenAIUsageJSON(count)
-	translated := sdktranslator.TranslateTokenCount(ctx, to, from, count, usageJSON)
-	return cliproxyexecutor.Response{Payload: []byte(translated)}, nil
 }
 
 // Refresh refreshes the Kimi token using the refresh token.