modelcontextprotocol / modelcontextprotocol/python-sdk
ClientSession: add public API for updating callbacks after initialization
还没有人认领这个 Issue。
- 主要语言
- Python
- 星标
- 24.3k
- 派生
- 4k
- 平均合并
- 1 天 1 小时
- 30 天内合并 PR
- 31
描述
Problem
ClientSession accepts callback parameters (list_roots_callback, sampling_callback, elicitation_callback) at initialization, but provides no public API to update them after the session is created.
This means any client that needs to change callbacks at runtime (e.g., updating roots in response to user action) must mutate private attributes like _list_roots_callback directly — which is fragile and couples consumers to implementation details.
Use case
A client connects to a server with initial roots, then the user changes the working directory or project context. The client needs to update its roots callback so that the next roots/list request from the server reflects the new roots. Today this requires:
session._list_roots_callback = new_callback # private attribute
The same issue applies to _sampling_callback and _elicitation_callback.
Proposed solution
Add public setter methods on ClientSession for updating callbacks after initialization. For example:
session.set_list_roots_callback(callback)
session.set_sampling_callback(callback)
session.set_elicitation_callback(callback)
Or alternatively, make the callback attributes public (without the leading underscore).
Context
This came up while fixing PrefectHQ/fastmcp#326 — Client.set_roots() wasn't updating the live session because it only modified pending kwargs. The fix (PrefectHQ/fastmcp#3714) mutates _list_roots_callback directly with a comment noting the fragility. A public API would make this safe and stable.
贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
调研方向
定位 ClientSession,检查现有的 _list_roots_callback、_sampling_callback 和 _elicitation_callback 属性及其使用方式。查看 ClientSession 附近的测试,然后添加并测试一种公开方式来更新每个 callback,使更改影响已经初始化的 session。
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- python
- 领域
- api
- Issue 类型
- 功能
- 难度
- 2/5
- 预计耗时
- 1-3 小时
- 活跃度
- 活跃
- 描述清晰度
- 基本清楚
- 新手友好度
- 70/100