guantao bef63dd906 Merge branch 'main' into dev_tao 5 days ago
..
features 8915822793 fixed: add multimodal info to token estimation 1 week ago
input 3d38517655 update: automatically create experiencce.md 1 week ago
resource 7c1d3651c2 chore 1 week ago
skills cfa3bc82c1 how prompt 1 week ago
tool b16f501a01 feat:model_usage 5 days ago
README.md 325df6d774 doc: how prompt 1 week ago
analyze_images.py 8915822793 fixed: add multimodal info to token estimation 1 week ago
encode_images.py 8915822793 fixed: add multimodal info to token estimation 1 week ago
load_imgs.py 8915822793 fixed: add multimodal info to token estimation 1 week ago
presets.json 2f341faef1 feat: print subagent trace 1 week ago
production.prompt 325df6d774 doc: how prompt 1 week ago
run.py 3429e8f24a update: restore example and bash with virtual env 1 week ago
save_b64.py 8915822793 fixed: add multimodal info to token estimation 1 week ago

README.md

运行方法

  1. 输入:将原始帖子内容放到 examples/how/input/ 文件夹下
  2. 运行:在项目根目录下运行 python examples/how/run.py
  3. 输出:在 examples/how/output 中查看

prompt调试

  • 主Agent(调度与评估):修改 examples/how/production.prompt
    • 原始输入/参考资料等等都可以在这里输入文件路径
    • 可以在这里指定各类输出的保存路径
  • 制作表解构Agent:修改 examples/how/skills/deconstruct.md , 这部分内容会在主Agent创建子Agent时作为子Agent的system prompt
  • 还原Agent:修改 examples/how/skills/construct.md , 这部分内容会在主Agent创建子Agent时作为子Agent的system prompt