Roadmap: where CoreCoder goes after v0.8.0
维护者通常 1 天内回复
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 12/100
- Issue 类型
- 功能
- 描述清晰度
- 需要澄清
- 活跃度
- 活跃
- 技术栈
- python
- 领域
- ai, cli, documentation
调研方向
This is a maintainer roadmap, not a task, and it names no single file to change. The closest concrete starting point is the eight essays in article/, whose line pins the issue says have drifted as MCP, hooks and plan mode landed, and the 227-test suite that backs the README numbers. There is no defined done state here; the issue stays open and is edited as items land, so a newcomer should read it for context and comment before starting any item.
由索引模型根据 Issue 内容生成。
描述
v0.8.0 is out and the engine is where I want it: 1,322 net lines covering the loop, the model interface, context compression, tools and sessions, with plan mode, hooks, checkpoints and MCP stdio on top, 227 tests green. This issue is the running note for what comes next, and just as much for what does not.
The promise that shapes the roadmap. CoreCoder's value is that one person can read all of it in an afternoon and trust every number on the README. Anything that trades readability for features is a regression here, no matter how useful that feature would be in a daily driver.
Next up, roughly in order:
-
Keep the reading series honest. The eight essays in
article/pin specific lines, and the engine kept growing after the first ones (MCP, hooks, plan mode all landed later). The pins need a sweep, and MCP plus plan mode probably deserve essays of their own. -
MCP streamable HTTP, maybe. Stdio covers local servers; the HTTP transport is how hosted servers talk, and people hit it in the wild. It goes in only if it stays around a hundred lines. If it starts growing an error taxonomy and auth branches, it does not belong in this repo.
-
Provider dialects as they surface. The 400 adaptation (
max_tokenstomax_completion_tokens, droppingtemperature) handles everything seen in the wild so far. New dialect cases get added only with a real 400 message in hand, never on speculation. -
Windows paths. Every Windows report so far got a real fix (doc pins, session paths). That bar stays.
Explicitly not on the roadmap: a TUI framework, more built-in tools beyond the current eight, multi-agent orchestration. Those are fine features for a product and poison for a reading reference.
This issue stays open and gets edited as things land. Suggestions welcome, especially the kind that end with "and it only costs twenty lines".
- 主要语言
- Python
- 星标
- 1.8k
- 派生
- 453
- 平均合并
- 6 小时 16 分钟
- 30 天内合并 PR
- 15
环境准备
这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。
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。
he-yufeng/CoreCoder 的其他 Issue
-
One non-object JSON record makes the session catalog exit可能已有人在做 @he-yufeng 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 68/100
维护者通常 1 天内回复
-
难度 1/5 1 小时以内 新手友好度 30/100
维护者通常 1 天内回复
-
难度 3/5 1-3 小时 新手友好度 25/100
he-yufeng/CoreCoder#62 · 1 条评论 ·
维护者通常 1 天内回复
-
学完之后的实践未关闭
难度 1/5 1 小时以内 新手友好度 25/100
he-yufeng/CoreCoder#49 · 2 条评论 ·
维护者通常 1 天内回复
查看 he-yufeng/CoreCoder 的全部 Issue
相似的 Issue
-
enhancement
难度 2/5 1-3 小时 新手友好度 72/100
维护者通常 1 天内回复
-
bug
难度 1/5 1 小时以内 新手友好度 89/100
维护者通常 1 天内回复
-
DoRA magnitude vectors are not loaded by Transformers load_adapter / from_pretrained(adapter_dir)未关闭post-v1
难度 2/5 1-3 小时 新手友好度 72/100
huggingface/peft#4208 · 1 条评论 ·
维护者通常 1 天内回复
-
approved correction metadata
难度 1/5 1 小时以内 新手友好度 76/100
acl-org/acl-anthology#10996 · 2 条评论 ·
维护者通常 1 天内回复
-
needs-triage
难度 1/5 1-3 小时 新手友好度 85/100
gke-labs/kube-agents#6620 ·
维护者通常 1 天内回复