Agently icon indicating copy to clipboard operation
Agently copied to clipboard

[AI Agent Application Development Framework] - 🚀 Build AI agent native application in very few code 💬 Easy to interact with AI agent in code using structure data and chained-calls syntax 🧩 Enhance...

Results 24 Agently issues
Sort by recently updated
recently updated
newest added

测试程序: https://gist.github.com/twinsant/e99141a75f8fce8be2ea88c5d2f675c2 如果我将本地Agently仓库路径加入sys.path ``` def use_local_agently(): sys.path.append('/Users/ant/GitHub/Agently') ``` 会出现下面的bug ``` (Agently) ➜ agents3 git:(main) ✗ python agently_tool.py /Users/ant/GitHub/Agently/Agently/__init__.py Exception in thread Thread-2 (start_in_theard): Traceback (most recent call last): File "/Users/ant/miniconda3/envs/Agently/lib/python3.10/threading.py",...

场景: 用户输入input问题,根据问题Agently使用工具进行解决。 工具 : 数据库向量查询。 预期: 根据input完整信息embedding后,查询数据库中与其最相似的问题然后返回。例如:”input(”测试环境登录不上去帮忙看看呢 http://127.0.0.1:9384/login“)“ 实际: 在调用工具时,Agently对input的原始信息进行了一次提炼,导致用户原始问题的信息丢失,例如:(测试环境登录不上去帮忙看看呢),缺失了环境接口信息(http://127.0.0.1:9384/login) 请问这种情况该怎么处理? 此问题呈现一定规律,并不是每次都会缺失信息,概率在50% ![image](https://github.com/user-attachments/assets/d47dd5b6-9f77-4d7d-9d07-cd229b63050d) ![image](https://github.com/user-attachments/assets/c48d39b1-3f8d-4ef7-8151-a324600bc115)

目前我本地安装的版本 Agently==3.0.4 agent = ( Agently.create_agent() .set_settings("current_model", "OAIClient") .set_settings("model.OAIClient.auth.api_key", "sk-3f10906a8c054a5a9cf0a7a4abe27e8a") # 使用DashScope专门适配的API Base URL .set_settings("model.OAIClient.url", "https://dashscope.aliyuncs.com/compatible-mode/v1") # 如果需要切换模型,可参考模型清单:https://help.aliyun.com/zh/dashscope/developer-reference/openai-file-interface .set_settings("model.OAIClient.options.model", "qwen-turbo") ) result = ( agent # general: agent应该知道的全局信息,通常被视作system prompt或类似位置的信息 .general("输出规定",...