chore: customize llm

src/server/main.py

@@ -1556,14 +1556,15 @@ def set_language_api(req: LanguageRequest):
 
 class LLMConfigDTO(BaseModel):
     base_url: str
-    api_key: str
+    api_key: Optional[str] = ""
     model_name: str
     fast_model_name: str
     mode: str
+    max_concurrent_requests: Optional[int] = 10
 
 class TestConnectionRequest(BaseModel):
     base_url: str
-    api_key: str
+    api_key: Optional[str] = ""
     model_name: str
 
 @app.get("/api/config/llm")
@@ -1574,7 +1575,8 @@ def get_llm_config():
         "api_key": getattr(CONFIG.llm, "key", ""),
         "model_name": getattr(CONFIG.llm, "model_name", ""),
         "fast_model_name": getattr(CONFIG.llm, "fast_model_name", ""),
-        "mode": getattr(CONFIG.llm, "mode", "default")
+        "mode": getattr(CONFIG.llm, "mode", "default"),
+        "max_concurrent_requests": getattr(CONFIG.ai, "max_concurrent_requests", 10)
     }
 
 @app.post("/api/config/llm/test")
@@ -1619,6 +1621,12 @@ async def save_llm_config(req: LLMConfigDTO):
         CONFIG.llm.fast_model_name = req.fast_model_name
         CONFIG.llm.mode = req.mode
 
+        # 更新 ai 配置
+        if req.max_concurrent_requests:
+            if not hasattr(CONFIG, "ai"):
+                 CONFIG.ai = OmegaConf.create({})
+            CONFIG.ai.max_concurrent_requests = req.max_concurrent_requests
+
         # 2. Persist to local_config.yml
         # 使用 src/utils/config.py 中类似的路径逻辑
         # 注意：这里我们假设是在项目根目录下运行，或者静态文件路径是相对固定的
@@ -1641,6 +1649,12 @@ async def save_llm_config(req: LLMConfigDTO):
         conf.llm.model_name = req.model_name
         conf.llm.fast_model_name = req.fast_model_name
         conf.llm.mode = req.mode
+
+        # Ensure ai section exists and update
+        if req.max_concurrent_requests:
+            if "ai" not in conf:
+                conf.ai = {}
+            conf.ai.max_concurrent_requests = req.max_concurrent_requests
         
         OmegaConf.save(conf, local_config_path)
         

web/src/components/game/panels/system/LLMConfigPanel.vue

@@ -16,7 +16,8 @@ const config = ref<LLMConfigDTO>({
   api_key: '',
   model_name: '',
   fast_model_name: '',
-  mode: 'default'
+  mode: 'default',
+  max_concurrent_requests: 10
 })
 
 const modeOptions = computed(() => [
@@ -107,10 +108,6 @@ const emit = defineEmits<{
 }>()
 
 async function handleTestAndSave() {
-  if (!config.value.api_key) {
-    message.warning(t('llm.api_key_required'))
-    return
-  }
   if (!config.value.base_url) {
     message.warning(t('llm.base_url_required'))
     return
@@ -185,6 +182,19 @@ onMounted(() => {
             class="input-field"
           />
         </div>
+
+        <div class="form-item">
+          <label>{{ t('llm.labels.max_concurrent_requests') }}</label>
+          <div class="desc">{{ t('llm.descs.max_concurrent_requests') }}</div>
+          <input 
+            v-model.number="config.max_concurrent_requests" 
+            type="number" 
+            min="1"
+            max="50"
+            :placeholder="t('llm.placeholders.max_concurrent_requests')"
+            class="input-field"
+          />
+        </div>
       </div>
 
       <!-- 模型配置 -->

web/src/locales/zh-CN.json

@@ -61,17 +61,20 @@
       "base_url": "Base URL",
       "normal_model": "智能模型 (Normal)",
       "fast_model": "快速模型 (Fast)",
+      "max_concurrent_requests": "最大并发数",
       "what_is_api": "什么是 API / 如何获取?"
     },
     "descs": {
       "normal_model": "用于处理复杂逻辑、剧情生成等任务",
-      "fast_model": "用于简单判定、频繁交互等任务"
+      "fast_model": "用于简单判定、频繁交互等任务",
+      "max_concurrent_requests": "同时请求 AI 的线程数。调大可加快速度，但可能触发 API 速率限制 (Rate Limit)。建议值：5-20。"
     },
     "placeholders": {
       "api_key": "在此填入你自己的 API Key (通常以 sk- 开头)",
       "base_url": "https://api.example.com/v1",
       "normal_model": "例如: gpt-4, claude-3-opus, qwen-plus",
-      "fast_model": "例如: gpt-3.5-turbo, qwen-flash"
+      "fast_model": "例如: gpt-3.5-turbo, qwen-flash",
+      "max_concurrent_requests": "默认: 10"
     },
     "modes": {
       "default": "均衡 (Default)",

web/src/types/api.ts

@@ -123,6 +123,7 @@ export interface LLMConfigDTO {
   model_name: string;
   fast_model_name: string;
   mode: string;
+  max_concurrent_requests?: number;
 }
 
 export interface GameStartConfigDTO {