Guess

Are you talking about using the [metagpt](https://docs.deepwisdom.ai/main/en/guide/get_started/quickstart.html#usage) command?

1. The incremental capabilities of metagpt are targeted at the requirements level， you can "write a snake game" first, and add "add moving enemy" as the incremental requirement. 2. If...

Sorry, I didn't understand what you wanted to do-_-||. - `requirement.txt` is an input for metagpt, and the old and incremental user requirements are merged and saved in PRD, not...

Sorry for not replying for so long. If you choose the DI solution, all SOPs can serve as tools for DI. Choosing multiple DI agents is certainly because they need...

I've consolidated all the incremental development-related issues into #1498 to make it easier to follow up. Any new issues will be discussed in this newly opened issue, and the old...

实现业务逻辑有三种方案： 1. 用SOP(standard operating procedure)形式实现业务流程。这种方案下流程是人工规划好的，operating（action）的输入输出都是设计好的，预期内的格式。由于人工规划好了流程，因此不需要agent理解各个action是干嘛的。 2. 用角色扮演agent的形式来串联业务流程。这种方案下流程也是人工规划好的，每个角色agent的输入输出只是在语义层面时预期内的，数据格式则是可以灵活调整的。 3. 用`RoleZero` agent的形式来串联业务流程。这种方案下流程是由agent规划的，只有这种情况下，agent才需要知道tool api的功能描述和使用方法。如果想了解更多的细节，可以参考：[DataInterpreter](https://github.com/geekan/MetaGPT/blob/main/metagpt/roles/di/data_interpreter.py)的代码。

用角色扮演agent的形式来串联业务流程时： 1. 每个agent的角色是规划好的，固定的角色关注固定的目标、采用固定的方式。 2. 任务相关的数据的提取和格式化由agent自己负责，不需要上下游agent负责。 方案2是介于方案1和方案3之间的方案。跟SOP相比，方案2对输入数据的格式没有强制要求，对输入的容错能力更高；跟方案3相比，方案2的功能（你可以理解成像方案3那样的tool api）是固定的。 方案2适用于SOP场景，且可接收人工输入（人做为SOP的一个环节）。

If there are no further questions, this issue will be closed. If you have any other questions, you can reopen this issue or create a new one.

```yaml llm: api_type: "ollama" model: "llama3.2" # or gpt-3.5-turbo base_url: "http://localhost:11434/api" api_key: "any string will be ok" ``` Sovled by: #1710

Please follow these steps: 1. [install metagpt](https://docs.deepwisdom.ai/main/en/guide/get_started/installation.html) 2. [config LLM](https://docs.deepwisdom.ai/main/en/guide/get_started/configuration/llm_api_configuration.html) 3. [run software company](https://docs.deepwisdom.ai/main/en/guide/get_started/quickstart.html): `metagpt "write a cli blackjack game"`