我的系统是Windows10 ，使用的IDEA，对acp的配置如下：
"iflow": {
"command": "C:\\\\Users\\\\cwsong\\\\AppData\\\\Local\\\\nvm\\\\v22.21.1\\\\iflow.cmd",
"args": \["--experimental-acp","-y"\]
}
然后新建聊天切换到iflow的时候没有显示可选模型，发送消息会收到错误提示：
Failed to initialize ACP session. Error: Field 'modelId' is required for type with serial name 'com.agentclientprotocol.model.ModelInfo', but it was missing
修改步骤
args 数组中，除了已有的 --experimental-acp 和 -y 外，新增两行 ：
"--model-id""具体的模型ID" （例如："gpt-4" 、"qwen-long" 或你服务端支持的模型名称）。我用的–model-id
发送消息就报错了：
Failed to initialize ACP process. Process terminated with exit code: 1. Process output:
Usage: iflow [options] [command]
iFlow CLI - Launch an interactive CLI, use -p/--prompt for non-interactive mode
Commands:
iflow [query] Launch iFlow CLI [default]
iflow mcp Manage MCP servers
iflow commands Manage marketplace commands
iflow agent Manage agents
iflow workflow workflowCommand.description
iflow skill Manage skills
Positionals:
query Optional query to process (equivalent to -p for non-interactive or -i for interactive mode) [string]
Options:
-m, --model Model [string]
-p, --prompt Prompt. Appended to input on stdin (if any). [string]
-i, --prompt-interactive Execute the provided prompt and continue in interactive mode [string]
-c, --continue Start the application and load the most recent conversation from current directory [boolean] [default: false]
-r, --resume Resume conversation from a specific session file. If no session ID is provided, shows interactive session selector.
-s, --sandbox Run in sandbox? [boolean]
--sandbox-image Sandbox image URI. [string]
-d, --debug Run in debug mode? [boolean] [default: false]
-a, --all-files Include ALL files in context? [boolean] [default: false]
--all_files Include ALL files in context? [deprecated: Use --all-files instead. We will be removing --all_files in the coming weeks.] [boolean] [default: false]
--show-memory-usage Show memory usage in status bar [boolean] [default: false]
--show_memory_usage Show memory usage in status bar [deprecated: Use --show-memory-usage instead. We will be removing --show_memory_usage in the coming weeks.] [boolean] [default: false]
-y, --yolo Automatically accept all actions (aka YOLO mode, see https://www.youtube.com/watch?v=xvFZjo5PgG0 for more details)? [boolean] [default: false]
--default Use default mode (manual approval for actions) [boolean]
--plan Use plan mode (planning without execution) [boolean]
--thinking Enable thinking mode when supported by the selected model [boolean]
--autoEdit Use auto-edit mode [boolean]
--telemetry Enable telemetry? This flag specifically controls if telemetry is sent. Other --telemetry-* flags set specific values but do not enable telemetry on their own. [boolean]
--telemetry-target Set the telemetry target (local or gcp). Overrides settings files. [string] [choices: "local", "gcp"]
--telemetry-otlp-endpoint Set the OTLP endpoint for telemetry. Overrides environment variables and settings files. [string]
--telemetry-log-prompts Enable or disable logging of user prompts for telemetry. Overrides settings files. [boolean]
--telemetry-outfile Redirect all telemetry output to the specified file. [string]
--checkpointing Enables checkpointing of file edits [boolean] [default: true]
--experimental-acp Starts the agent in ACP mode [boolean]
--stream Enable or disable streaming mode [boolean]
--port Port number for ACP server (used with --experimental-acp) [number]
--allowed-mcp-server-names Allowed MCP server names [array]
--include-directories, --add-dir Additional directories to include in the workspace (comma-separated or multiple --include-directories) [array]
--max-turns Maximum number of model calls (assistant rounds) before terminating [number]
--max_turns Maximum number of model calls (assistant rounds) before terminating [deprecated: Use --max-turns instead. We will be removing --max_turns in the coming weeks.] [number]
--max-tokens Maximum total token usage (input + output) before terminating [number]
--max_tokens Maximum total token usage (input + output) before terminating [deprecated: Use --max-tokens instead. We will be removing --max_tokens in the coming weeks.] [number]
--timeout Maximum execution time in seconds before terminating [number]
-o, --output-file Output file path to save execution information (non-interactive mode only) [string]
--output_file Output file path to save execution information (non-interactive mode only) [deprecated: Use --output-file instead. We will be removing --output_file in the coming weeks.] [string]
-v, --version Show version number [boolean]
-h, --help Show help [boolean]
Unknown arguments: model-id, modelId
嘿嘿，等等明天官方看看吧，兄弟，先用命令行终端
不知道官方啥时候管管
排查后判断是个bug，之后会修复，很抱歉影响您的使用 
不懂就问，ACP是啥。。。？
用于连接代码编辑器（如IntelliJ IDEA, VS Code）与AI Agent（如Claude Code, Codex）的协议。
配置目的： 让AI助手直接在IDE中运行，而不是在终端里闪屏；实现更流畅的代码生成和工具调用（如MCP）。
okkkk,期待修复
ACP (Agent Client Protocol) 是 Zed 编辑器推出的一种通用智能体协议，使智能体的核心功能（服务端）和用户界面（客户端）解耦，允许用户自由选择不同的智能体服务端和客户端进行搭配使用。
翻译成人话：如果你的IDE/编辑器支持ACP协议，就可以在IDE原生AI功能中使用支持ACP协议的agent。
例如：假设IDEA、VSCODE支持ACP协议，就可以直接在IDEA/VSCODE自己的AI功能中使用codex、cc、opencode，而不需要对应的插件或者cli中使用。
新版本将修复这个问题，这两天发布最新版本哦～
刚刚又遇到个新问题：
agent选择iflow后提示Starting iflow ,过了很久之后提示：
Failed to initialize ACP session. Error: Process cancelled
是最新版本吗，方便的话可以提供下具体日志？
刚刚更新到0.5.4，依旧是这个问题
好的，技术同学在看了
我从别人那抄了个acp配置就可以了，希望可以在IDEA内切换模型
"iFlow": {
"command": "C:\\nvm4w\\nodejs\\node.exe",
"args": [
"C:\\Users\\cwsong\\AppData\\Local\\nvm\\v22.21.1\\node_modules\\@iflow-ai\\iflow-cli\\bundle\\iflow.js",
"--experimental-acp"
]
}
切换模型这个功能需要等 ACP 规范中明确指出，然后等各个客户端适配才能用上。
如果在 terminal 里面可以直接运行 iflow 命令，可以尝试下面的配置。--stream 可以开启流式输出
{
"default_mcp_settings": {},
"agent_servers": {
"iFlow": {
"command": "iflow.cmd",
"args": [
"--experimental-acp",
"--stream"
],
"env": {
}
}
}
}
我看QwenCode和OpenCode都可以进行切换，他们是怎么实现的呢？