feat: 添加 OpenAI Chat Completions 兼容端点

基于 @yulate 在 PR #648 (commit 0bb6a392) 的工作，解决了与最新
main 分支的合并冲突。

原始功能（@yulate）:
- 添加 /v1/chat/completions 和 /chat/completions 兼容端点
- 将 Chat Completions 请求转换为 Responses API 格式并转换回来
- 添加 API Key 直连转发支持
- 包含单元测试

Co-authored-by: yulate <yulate@users.noreply.github.com>

backend/internal/service/openai_gateway_service.go

@@ -12,6 +12,7 @@ import (
 	"io"
 	"math/rand"
 	"net/http"
+	"regexp"
 	"sort"
 	"strconv"
 	"strings"
@@ -36,6 +37,7 @@ const (
 	chatgptCodexURL = "https://chatgpt.com/backend-api/codex/responses"
 	// OpenAI Platform API for API Key accounts (fallback)
 	openaiPlatformAPIURL   = "https://api.openai.com/v1/responses"
+	openaiChatAPIURL       = "https://api.openai.com/v1/chat/completions"
 	openaiStickySessionTTL = time.Hour // 粘性会话TTL
 	codexCLIUserAgent      = "codex_cli_rs/0.104.0"
 	// codex_cli_only 拒绝时单个请求头日志长度上限（字符）
@@ -54,6 +56,16 @@ const (
 	codexCLIVersion                    = "0.104.0"
 )
 
+// OpenAIChatCompletionsBodyKey stores the original chat-completions payload in gin.Context.
+const OpenAIChatCompletionsBodyKey = "openai_chat_completions_body"
+
+// OpenAIChatCompletionsIncludeUsageKey stores stream_options.include_usage in gin.Context.
+const OpenAIChatCompletionsIncludeUsageKey = "openai_chat_completions_include_usage"
+
+// openaiSSEDataRe matches SSE data lines with optional whitespace after colon.
+// Some upstream APIs return non-standard "data:" without space (should be "data: ").
+var openaiSSEDataRe = regexp.MustCompile(`^data:\s*`)
+
 // OpenAI allowed headers whitelist (for non-passthrough).
 var openaiAllowedHeaders = map[string]bool{
 	"accept-language":       true,
@@ -97,6 +109,19 @@ var codexCLIOnlyDebugHeaderWhitelist = []string{
 	"X-Real-IP",
 }
 
+// OpenAI chat-completions allowed headers (extend responses whitelist).
+var openaiChatAllowedHeaders = map[string]bool{
+	"accept-language":     true,
+	"content-type":        true,
+	"conversation_id":     true,
+	"user-agent":          true,
+	"originator":          true,
+	"session_id":          true,
+	"openai-organization": true,
+	"openai-project":      true,
+	"openai-beta":         true,
+}
+
 // OpenAICodexUsageSnapshot represents Codex API usage limits from response headers
 type OpenAICodexUsageSnapshot struct {
 	PrimaryUsedPercent          *float64 `json:"primary_used_percent,omitempty"`
@@ -1577,6 +1602,23 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 		return nil, errors.New("codex_cli_only restriction: only codex official clients are allowed")
 	}
 
+	if c != nil && account != nil && account.Type == AccountTypeAPIKey {
+		if raw, ok := c.Get(OpenAIChatCompletionsBodyKey); ok {
+			if rawBody, ok := raw.([]byte); ok && len(rawBody) > 0 {
+				includeUsage := false
+				if v, ok := c.Get(OpenAIChatCompletionsIncludeUsageKey); ok {
+					if flag, ok := v.(bool); ok {
+						includeUsage = flag
+					}
+				}
+				if passthroughWriter, ok := c.Writer.(interface{ SetPassthrough() }); ok {
+					passthroughWriter.SetPassthrough()
+				}
+				return s.forwardChatCompletions(ctx, c, account, rawBody, includeUsage, startTime)
+			}
+		}
+	}
+
 	originalBody := body
 	reqModel, reqStream, promptCacheKey := extractOpenAIRequestMetaFromBody(body)
 	originalModel := reqModel