脚本策略增加运行控制，并输出日志方便策略开发调试。

backend_api_python/app/routes/strategy.py

@@ -9,6 +9,7 @@
 import re
 import traceback
 import time
+import os
 
 from app.services.strategy import StrategyService
 from app.services.strategy_compiler import StrategyCompiler
@@ -1989,6 +1990,7 @@ def ai_generate_strategy():
         payload = request.get_json() or {}
         lang = _request_lang()
         prompt = payload.get('prompt', '')
+        language_prompt = f"The comments in the code should be in {payload.get('language', 'English')}"
         if not prompt.strip():
             return jsonify({'code': '', 'msg': _strategy_ai_text('prompt_empty', lang), 'params': None})
 
@@ -2185,7 +2187,7 @@ def ai_generate_strategy():
 
             content = llm.call_llm_api(
                 messages=[
-                    {"role": "system", "content": system_prompt},
+                    {"role": "system", "content": system_prompt+language_prompt},
                     {"role": "user", "content": user_content},
                 ],
                 model=llm.get_code_generation_model(),
@@ -2287,7 +2289,7 @@ def ai_generate_strategy():
 
             content = llm.call_llm_api(
                 messages=[
-                    {"role": "system", "content": system_prompt},
+                    {"role": "system", "content": system_prompt+language_prompt},
                     {"role": "user", "content": user_content},
                 ],
                 model=llm.get_code_generation_model(),
@@ -2313,8 +2315,16 @@ def ai_generate_strategy():
             if not isinstance(updates, dict):
                 return jsonify({'code': '', 'params': None, 'msg': _strategy_ai_text('invalid_json_params', lang)})
             return jsonify({'code': '', 'params': updates, 'msg': _strategy_ai_text('success', lang)})
-
-        system_prompt = """You are a quantitative trading strategy code generator.
+        try:
+            _backtest_framework_path = os.path.join(
+                os.path.dirname(os.path.abspath(__file__)),
+                '..', 'services', 'backtest.py'
+            )
+            with open(_backtest_framework_path, 'r', encoding='utf-8') as _f:
+                frame_context = _f.read()
+        except Exception:
+            frame_context = ""
+        system_prompt = f"""You are a quantitative trading strategy code generator.
 Generate Python strategy code that follows this framework:
 - def on_init(ctx): Initialize strategy parameters using ctx.param(name, default)
 - def on_bar(ctx, bar): Core logic called on each K-line bar
@@ -2342,6 +2352,8 @@ def ai_generate_strategy():
 - When sizing with ctx.equity * some_pct, keep some_pct as a 0–1 ratio.
 - Template UI may show 0–100; only the Python default literals should be ratios.
 - If user says "80% position", use ctx.param('position_pct', 0.8) and qty = ctx.equity * ctx.position_pct / price.
+Strategy runtime framework context:
+{frame_context}
 """
 
         extra = ''
@@ -2362,7 +2374,7 @@ def ai_generate_strategy():
 
         content = llm.call_llm_api(
             messages=[
-                {"role": "system", "content": system_prompt},
+                {"role": "system", "content": system_prompt+language_prompt},
                 {"role": "user", "content": user_prompt},
             ],
             model=llm.get_code_generation_model(),
@@ -2422,7 +2434,7 @@ def _repair_strategy_code_via_llm(bad_code: str, validation: dict) -> str:
             )
             repaired_content = llm.call_llm_api(
                 messages=[
-                    {"role": "system", "content": system_prompt},
+                    {"role": "system", "content": system_prompt+language_prompt},
                     {"role": "user", "content": repair_prompt},
                 ],
                 model=llm.get_code_generation_model(),

backend_api_python/app/services/backtest.py

@@ -729,6 +729,7 @@ def run_multi_timeframe(
                 mtf_requested=True,
                 mtf_active=True,
             )
+            result['logs'] = signals.get('logs', [])
             self._attach_actual_range_to_result(result, df_signal)
             logger.info("Backtest result formatted successfully")
         except Exception as e:
@@ -1773,6 +1774,7 @@ def _run_script_strategy(
         ea['fillRule'] = 'next_bar_open'
         result['executionAssumptions'] = ea
         self._attach_actual_range_to_result(result, df)
+        result['logs'] = signals.get('logs', [])
         return result
 
     def run_code_strategy(
@@ -2007,7 +2009,7 @@ def run(
         if df.empty:
             raise ValueError("No candle data available in the backtest date range")
 
-        
+
         # 2. Execute indicator code to get signals (pass backtest params)
         backtest_params = {
             'leverage': leverage,
@@ -2444,8 +2446,10 @@ def _execute_script_strategy(self, code: str, df: pd.DataFrame, runtime: Optiona
                     volume=float(row.get('volume') or 0),
                     timestamp=row.get('time')
                 )
-                on_bar(ctx, bar)
-
+                try:
+                    on_bar(ctx, bar)
+                except Exception as e:
+                    ctx.log(traceback.format_exc())
                 for order in ctx._orders:
                     action = str(order.get('action') or '').lower()
                     intent = str(order.get('intent') or 'auto').lower()
@@ -2521,6 +2525,7 @@ def _execute_script_strategy(self, code: str, df: pd.DataFrame, runtime: Optiona
                 'close_short': close_short,
                 'add_long': add_long,
                 'add_short': add_short,
+                'logs': ctx.flush_logs(),
             }
         except Exception as e:
             logger.error(f"Strategy script execution error: {e}")