feat(translators): add token counting support for Claude and Gemini responses

- Implemented `TokenCount` transform method across translators to calculate token usage.
- Integrated token counting logic into executor pipelines for Claude, Gemini, and CLI translators.
- Added corresponding API endpoints and handlers (`/messages/count_tokens`) for token usage retrieval.
- Enhanced translation registry to support `TokenCount` functionality alongside existing response types.

sdk/translator/registry.go

@@ -91,6 +91,19 @@ func (r *Registry) TranslateNonStream(ctx context.Context, from, to Format, mode
 	return string(rawJSON)
 }
 
+// TranslateNonStream applies the registered non-stream response translator.
+func (r *Registry) TranslateTokenCount(ctx context.Context, from, to Format, count int64, rawJSON []byte) string {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	if byTarget, ok := r.responses[to]; ok {
+		if fn, isOk := byTarget[from]; isOk && fn.TokenCount != nil {
+			return fn.TokenCount(ctx, count)
+		}
+	}
+	return string(rawJSON)
+}
+
 var defaultRegistry = NewRegistry()
 
 // Default exposes the package-level registry for shared use.
@@ -122,3 +135,8 @@ func TranslateStream(ctx context.Context, from, to Format, model string, origina
 func TranslateNonStream(ctx context.Context, from, to Format, model string, originalRequestRawJSON, requestRawJSON, rawJSON []byte, param *any) string {
 	return defaultRegistry.TranslateNonStream(ctx, from, to, model, originalRequestRawJSON, requestRawJSON, rawJSON, param)
 }
+
+// TranslateTokenCount is a helper on the default registry.
+func TranslateTokenCount(ctx context.Context, from, to Format, count int64, rawJSON []byte) string {
+	return defaultRegistry.TranslateTokenCount(ctx, from, to, count, rawJSON)
+}

sdk/translator/types.go

@@ -11,8 +11,11 @@ type ResponseStreamTransform func(ctx context.Context, model string, originalReq
 // ResponseNonStreamTransform converts non-stream responses between schemas.
 type ResponseNonStreamTransform func(ctx context.Context, model string, originalRequestRawJSON, requestRawJSON, rawJSON []byte, param *any) string
 
+type ResponseTokenCountTransform func(ctx context.Context, count int64) string
+
 // ResponseTransform groups streaming and non-streaming transforms.
 type ResponseTransform struct {
-	Stream    ResponseStreamTransform
-	NonStream ResponseNonStreamTransform
+	Stream     ResponseStreamTransform
+	NonStream  ResponseNonStreamTransform
+	TokenCount ResponseTokenCountTransform
 }