来源: GitCode Issue #172 by 曹晓乐 (https://gitcode.com/developer-skill/vod-skill/issues/172)

描述问题 (Description)

CodeArts CLI生成代码耗时过长，一个简单的Flask应用（6个HTML模板+1个CSS+1个app.py）需要7分39秒。其中模板+CSS生成占57.5%（263秒），是最大瓶颈。模型排队等待也耗时36秒。

复现步骤 (To Reproduce)

• 用户意图: 使用CodeArts CLI生成教师点名系统代码
• Agent执行: Called huawei-cloud-codearts-code-gen skill with GLM-5.1 model to generate a Python Flask application
• 触发错误: performance_issue - CodeArts CLI code generation takes excessively long time (7min39s for a simple Flask app), with template+CSS generation consuming 57.5% of total time

预期行为 (Expected behavior)

简单应用的代码生成应在2-3分钟内完成，模板/CSS生成应支持并行或增量生成以减少等待时间

实际行为 (Actual behavior)

总耗时7分39秒，模板+CSS单步耗时263秒，用户等待体验差

错误堆栈 (Stack Trace)

``` CodeArts CLI code generation performance analysis:

• Total time: 7min 39s (17:17:25 -> 17:25:04)
• Step 1 (project init): ~36s (model queue wait)
• Step 2 (app.py generation): 93s (20.3% of total)
• Step 3 (templates + CSS generation): 263s (57.5% of total) - BIGGEST BOTTLENECK
• Step 4 (requirements.txt): minimal
• Total tokens: 558,612 (input: 79,821 / output: 26,706 / reasoning: 2,229)

The template+CSS generation step is the primary bottleneck, taking nearly 4.5 minutes for 6 HTML templates and 1 CSS file. ```