修改balance资源配置

README.md

@@ -496,7 +496,7 @@ MODEL_SERVER=https://openrouter.ai/api/v1
 API_KEY=your-api-key
 
 # 队列配置  
-AGENT_CACHE_MAX_SIZE=20
+TOOL_CACHE_MAX_SIZE=20
 
 # 其他配置
 TOKENIZERS_PARALLELISM=false

agent/agent_memory_cache.py

@@ -12,7 +12,7 @@ from collections import OrderedDict
 from datetime import datetime, timedelta
 
 import cachetools
-from utils.settings import AGENT_CACHE_MAX_SIZE, AGENT_CACHE_TTL, AGENT_CACHE_AUTO_RENEW
+from utils.settings import TOOL_CACHE_MAX_SIZE, TOOL_CACHE_TTL, TOOL_CACHE_AUTO_RENEW
 
 logger = logging.getLogger('app')
 
@@ -371,9 +371,9 @@ def get_memory_cache_manager() -> AgentMemoryCacheManager:
     if _global_cache_manager is None:
         # 从 settings 导入配置
         _global_cache_manager = AgentMemoryCacheManager(
-            max_size=AGENT_CACHE_MAX_SIZE,
-            default_ttl=AGENT_CACHE_TTL,
-            auto_renew=AGENT_CACHE_AUTO_RENEW
+            max_size=TOOL_CACHE_MAX_SIZE,
+            default_ttl=TOOL_CACHE_TTL,
+            auto_renew=TOOL_CACHE_AUTO_RENEW
         )
 
     return _global_cache_manager

docs/agent_memory_cache.md

@@ -19,15 +19,15 @@ Agent 内存缓存系统使用 cachetools 库在内存中缓存 Agent 实例，
 
 可以通过以下环境变量配置缓存行为：
 
-- `AGENT_CACHE_MAX_SIZE`：最大缓存项数（默认：`1000`）
-- `AGENT_CACHE_TTL`：默认过期时间，秒（默认：`180` = 3分钟）
-- `AGENT_CACHE_AUTO_RENEW`：是否自动续期（默认：`true`）
+- `TOOL_CACHE_MAX_SIZE`：最大缓存项数（默认：`1000`）
+- `TOOL_CACHE_TTL`：默认过期时间，秒（默认：`180` = 3分钟）
+- `TOOL_CACHE_AUTO_RENEW`：是否自动续期（默认：`true`）
 
 示例：
 ```bash
-export AGENT_CACHE_MAX_SIZE="500"
-export AGENT_CACHE_TTL="300"  # 5分钟
-export AGENT_CACHE_AUTO_RENEW="false"
+export TOOL_CACHE_MAX_SIZE="500"
+export TOOL_CACHE_TTL="300"  # 5分钟
+export TOOL_CACHE_AUTO_RENEW="false"
 ```
 
 ### 代码配置

markdown/PERFORMANCE_OPTIMIZATION.md

@@ -65,7 +65,7 @@
 ### 环境变量配置
 ```bash
 # Agent缓存配置
-export AGENT_CACHE_MAX_SIZE=50
+export TOOL_CACHE_MAX_SIZE=50
 
 # Tokenizer优化
 export TOKENIZERS_PARALLELISM=true

start_all_optimized.sh

@@ -62,17 +62,17 @@ setup_environment() {
     case $PROFILE in
         "low_memory")
             export TOKENIZERS_PARALLELISM=false
-            export AGENT_CACHE_MAX_SIZE=20
+            export TOOL_CACHE_MAX_SIZE=20
             ;;
         "balanced")
             export TOKENIZERS_PARALLELISM=true
             export TOKENIZERS_FAST=1
-            export AGENT_CACHE_MAX_SIZE=50
+            export TOOL_CACHE_MAX_SIZE=50
             ;;
         "high_performance")
             export TOKENIZERS_PARALLELISM=true
             export TOKENIZERS_FAST=1
-            export AGENT_CACHE_MAX_SIZE=100
+            export TOOL_CACHE_MAX_SIZE=100
             ;;
     esac
 

start_unified.py

@@ -226,19 +226,19 @@ class ProcessManager:
         if args.profile == "low_memory":
             env_vars = {
                 'TOKENIZERS_PARALLELISM': 'false',
-                'AGENT_CACHE_MAX_SIZE': '20',
+                'TOOL_CACHE_MAX_SIZE': '10',
             }
         elif args.profile == "balanced":
             env_vars = {
                 'TOKENIZERS_PARALLELISM': 'true',
                 'TOKENIZERS_FAST': '1',
-                'AGENT_CACHE_MAX_SIZE': '50',
+                'TOOL_CACHE_MAX_SIZE': '20',
             }
         elif args.profile == "high_performance":
             env_vars = {
                 'TOKENIZERS_PARALLELISM': 'true',
                 'TOKENIZERS_FAST': '1',
-                'AGENT_CACHE_MAX_SIZE': '100',
+                'TOOL_CACHE_MAX_SIZE': '50',
             }
 
         # 通用优化
@@ -331,11 +331,11 @@ def main():
     if args.api_workers is None:
         cpu_count = multiprocessing.cpu_count()
         if args.profile == "low_memory":
-            args.api_workers = min(2, cpu_count)
+            args.api_workers = min(1, cpu_count)
         elif args.profile == "balanced":
-            args.api_workers = min(4, cpu_count + 1)
+            args.api_workers = min(2, cpu_count + 1)
         elif args.profile == "high_performance":
-            args.api_workers = min(8, cpu_count * 2)
+            args.api_workers = min(4, cpu_count * 2)
     
     # 创建进程管理器并运行
     manager = ProcessManager()

utils/settings.py

@@ -9,9 +9,9 @@ SUMMARIZATION_MAX_TOKENS = MAX_CONTEXT_TOKENS - MAX_OUTPUT_TOKENS - 1000
 SUMMARIZATION_MESSAGES_TO_KEEP = int(os.getenv("SUMMARIZATION_MESSAGES_TO_KEEP", 20))
 
 # Agent Cache Settings
-AGENT_CACHE_MAX_SIZE = int(os.getenv("AGENT_CACHE_MAX_SIZE", 20))
-AGENT_CACHE_TTL = int(os.getenv("AGENT_CACHE_TTL", 180))
-AGENT_CACHE_AUTO_RENEW = os.getenv("AGENT_CACHE_AUTO_RENEW", "true") == "true"
+TOOL_CACHE_MAX_SIZE = int(os.getenv("TOOL_CACHE_MAX_SIZE", 20))
+TOOL_CACHE_TTL = int(os.getenv("TOOL_CACHE_TTL", 180))
+TOOL_CACHE_AUTO_RENEW = os.getenv("TOOL_CACHE_AUTO_RENEW", "true") == "true"
 
 # API Settings
 BACKEND_HOST = os.getenv("BACKEND_HOST", "https://api-dev.gptbase.ai")
@@ -57,6 +57,6 @@ CHECKPOINT_BUSY_TIMEOUT = int(os.getenv("CHECKPOINT_BUSY_TIMEOUT", "10000"))
 
 # 连接池大小
 # 同时可以持有的最大连接数
-CHECKPOINT_POOL_SIZE = int(os.getenv("CHECKPOINT_POOL_SIZE", "30"))
+CHECKPOINT_POOL_SIZE = int(os.getenv("CHECKPOINT_POOL_SIZE", "15"))