@milisp/keke 0.1.8 → 0.1.9
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +46 -68
- package/README.zh-CN.md +35 -58
- package/package.json +6 -6
package/README.md
CHANGED
|
@@ -1,10 +1,15 @@
|
|
|
1
1
|
# keke
|
|
2
2
|
|
|
3
|
-
[中文文档](README.zh-CN.md) | [Architecture](docs/architecture.md) | [Roadmap](docs/ROADMAP.md)
|
|
3
|
+
[中文文档](README.zh-CN.md) | [Architecture](docs/architecture.md) | [Config](docs/config.md) | [Roadmap](docs/ROADMAP.md)
|
|
4
4
|
|
|
5
5
|
keke is a local terminal coding agent built for zero-vendor lock-in.
|
|
6
6
|
Works with subscriptions you already have, standard API keys, or self-hosted local models.
|
|
7
7
|
|
|
8
|
+
**Status: early (v0.1.x), usable day to day.** Good fit for personal daily
|
|
9
|
+
driving, scripted/CI use, and small-team trials. Not yet the right choice for
|
|
10
|
+
a regulated or uninterruptible production pipeline — see
|
|
11
|
+
[Production & safety](#production--safety).
|
|
12
|
+
|
|
8
13
|
## Why keke?
|
|
9
14
|
|
|
10
15
|
- **Protocol: ACP for every client**
|
|
@@ -23,7 +28,9 @@ curl -fsSL https://raw.githubusercontent.com/milisp/keke-agent/main/scripts/inst
|
|
|
23
28
|
```
|
|
24
29
|
|
|
25
30
|
This downloads the latest prebuilt binary for your platform into
|
|
26
|
-
`~/.local/bin` (override with `KEKE_INSTALL_DIR`).
|
|
31
|
+
`~/.local/bin` (override with `KEKE_INSTALL_DIR`). Piping a remote script into
|
|
32
|
+
`sh` runs it with your privileges — inspect it first if that matters to you:
|
|
33
|
+
`curl -fsSL .../install.sh | less`.
|
|
27
34
|
|
|
28
35
|
### npm
|
|
29
36
|
|
|
@@ -35,92 +42,63 @@ You can also grab a binary directly from the
|
|
|
35
42
|
[latest release](https://github.com/milisp/keke-agent/releases/latest), or build
|
|
36
43
|
from source with `cargo build --release`.
|
|
37
44
|
|
|
38
|
-
## Try it
|
|
45
|
+
## Try it (30 seconds)
|
|
39
46
|
|
|
40
47
|
```sh
|
|
41
|
-
#
|
|
48
|
+
keke doctor # see which providers/logins resolve, before you rely on one
|
|
49
|
+
|
|
50
|
+
# Sign in with a subscription you already pay for...
|
|
42
51
|
keke login codex
|
|
43
52
|
keke login grok
|
|
44
|
-
|
|
45
53
|
# ...or bring a key
|
|
46
|
-
export
|
|
54
|
+
export ANTHROPIC_API_KEY=sk-ant-...
|
|
47
55
|
|
|
48
56
|
keke exec "what does this project do?" # one-shot, for scripts and CI
|
|
49
57
|
keke # interactive TUI
|
|
50
58
|
keke resume # pick the last conversation back up
|
|
51
|
-
keke doctor # which providers and logins resolve
|
|
52
|
-
```
|
|
53
|
-
|
|
54
|
-
## What you can point it at
|
|
55
|
-
|
|
56
|
-
| Provider | How you authenticate | Notes |
|
|
57
|
-
| --- | --- | --- |
|
|
58
|
-
| OpenAI / ChatGPT | `keke login codex`, or `OPENAI_API_KEY` | OAuth flow ported from codex |
|
|
59
|
-
| Anthropic Claude | `ANTHROPIC_API_KEY` | Built in; the default provider. API key only — no subscription login |
|
|
60
|
-
| xAI Grok | `keke login grok`, or `XAI_API_KEY` | Built in |
|
|
61
|
-
| Local (Ollama, vLLM, …) | none | Your code never leaves the machine |
|
|
62
|
-
| Any OpenAI-compatible gateway | `env_key` in `config.toml` | Company proxies, NVIDIA NIM, routers |
|
|
63
|
-
|
|
64
|
-
Anything not built in is a few lines of `$KEKE_HOME/config.toml`, not a code
|
|
65
|
-
change:
|
|
66
|
-
|
|
67
|
-
```toml
|
|
68
|
-
[providers.ollama]
|
|
69
|
-
base_url = "http://localhost:11434/v1"
|
|
70
|
-
default_model = "gpt-oss:20b"
|
|
71
|
-
|
|
72
|
-
[providers.anthropic]
|
|
73
|
-
base_url = "https://api.anthropic.com"
|
|
74
|
-
env_key = "ANTHROPIC_API_KEY"
|
|
75
|
-
wire = "messages"
|
|
76
|
-
```
|
|
77
|
-
|
|
78
|
-
`wire` picks the request format — `chat_completions` (the default),
|
|
79
|
-
`responses`, or `messages` — so a new endpoint is a config entry rather than a release.
|
|
80
|
-
|
|
81
|
-
Behind a corporate proxy or a TLS-intercepting gateway, a declared provider
|
|
82
|
-
can also carry a CA certificate, a proxy, and headers your gateway wants for
|
|
83
|
-
routing or audit:
|
|
84
|
-
|
|
85
|
-
```toml
|
|
86
|
-
[providers.internal-gateway]
|
|
87
|
-
base_url = "https://gateway.corp.internal/v1"
|
|
88
|
-
env_key = "INTERNAL_GATEWAY_API_KEY"
|
|
89
|
-
ca_cert_path = "/etc/ssl/certs/corp-root-ca.pem"
|
|
90
|
-
proxy = "http://proxy.corp.internal:8080"
|
|
91
|
-
proxy_username = "svc-keke"
|
|
92
|
-
proxy_password_env_key = "CORP_PROXY_PASSWORD"
|
|
93
|
-
|
|
94
|
-
[providers.internal-gateway.headers]
|
|
95
|
-
X-Department-Token = "env:DEPT_TOKEN"
|
|
96
|
-
X-Company-User-Id = "milisp-labs"
|
|
97
59
|
```
|
|
98
60
|
|
|
99
|
-
|
|
100
|
-
|
|
101
|
-
file in the clear. `authorization` is reserved for the provider's own
|
|
102
|
-
credential and cannot be set here.
|
|
61
|
+
Provider routes, API keys, local models, gateways, per-directory accounts, and
|
|
62
|
+
all other settings are documented in [`docs/config.md`](docs/config.md).
|
|
103
63
|
|
|
104
64
|
## Status
|
|
105
65
|
|
|
106
66
|
Usable day to day. `keke exec`, the TUI, and the ACP server all run real
|
|
107
67
|
sessions end to end; runtime plugins (skills, commands, hooks, MCP servers)
|
|
108
68
|
install in the Claude Code format, and repository-provided ones stay inert
|
|
109
|
-
until you approve them.
|
|
110
|
-
|
|
69
|
+
until you approve them. `/model` switches models inside a running session
|
|
70
|
+
(tied to their provider so config can't persist an invalid pairing), and the
|
|
71
|
+
agent can spawn subagents — isolated child sessions given one task that
|
|
72
|
+
report back a single answer instead of their whole search trace. See
|
|
73
|
+
[`docs/ROADMAP.md`](docs/ROADMAP.md) for what's next.
|
|
74
|
+
|
|
75
|
+
## Production & safety
|
|
76
|
+
|
|
77
|
+
Fine for personal use, local/self-hosted models, and CI one-shots today.
|
|
78
|
+
Before trusting it with anything you can't afford to break, weigh in:
|
|
79
|
+
|
|
80
|
+
- **Sandboxing & approvals** — default `approval_policy` and `sandbox_mode`
|
|
81
|
+
are conservative, but you should confirm they match what you're running
|
|
82
|
+
([`docs/config.md`](docs/config.md)).
|
|
83
|
+
- **Plugin trust** — repository-provided plugins (hooks, MCP servers) never
|
|
84
|
+
execute on `git clone` alone; a person must approve them, keyed to their
|
|
85
|
+
exact contents, not their path. There's no flag to turn that gate off.
|
|
86
|
+
- **Maturity** — early (v0.1.x), single-maintainer, no formal security audit
|
|
87
|
+
or SLA. Remaining roadmap gaps (full end-to-end MCP tool-call verification)
|
|
88
|
+
are tracked in [`docs/ROADMAP.md`](docs/ROADMAP.md).
|
|
89
|
+
|
|
90
|
+
**Good fit:** you want no vendor lock-in, script/CI-friendly one-shot runs,
|
|
91
|
+
local models, or ACP integration into your own editor/client.
|
|
92
|
+
**Not yet a fit:** you need turnkey OAuth for every vendor, an established
|
|
93
|
+
plugin ecosystem, or vendor support contracts.
|
|
111
94
|
|
|
112
95
|
## Where it comes from
|
|
113
96
|
|
|
114
|
-
keke is a fresh implementation
|
|
115
|
-
|
|
116
|
-
|
|
117
|
-
|
|
118
|
-
|
|
119
|
-
flow, for instance); most of it is keke's own code, shaped by what worked and
|
|
120
|
-
what didn't in those three. What keke does differently — and why that
|
|
121
|
-
matters enough to enforce in CI — is in
|
|
122
|
-
[`docs/architecture.md`](docs/architecture.md); the invariants themselves are
|
|
123
|
-
in [`AGENTS.md`](AGENTS.md).
|
|
97
|
+
keke is a fresh implementation informed by OpenAI's **codex**, xAI's
|
|
98
|
+
**grok-build**, and **deepseek-harness**. A few pieces are ported and
|
|
99
|
+
attributed in-crate; most of the code is original. Design rationale and
|
|
100
|
+
CI-enforced invariants: [`docs/architecture.md`](docs/architecture.md),
|
|
101
|
+
[`AGENTS.md`](AGENTS.md).
|
|
124
102
|
|
|
125
103
|
## License
|
|
126
104
|
|
package/README.zh-CN.md
CHANGED
|
@@ -1,16 +1,20 @@
|
|
|
1
1
|
# keke
|
|
2
2
|
|
|
3
|
-
[
|
|
3
|
+
[English](README.md) | [架构设计](docs/architecture.md) | [配置文档](docs/config.md) | [路线图](docs/ROADMAP.md)
|
|
4
4
|
|
|
5
|
-
|
|
5
|
+
keke 是一个专为“零厂商锁定”打造的本地终端编码 agent。
|
|
6
|
+
可直接搭配你已有的付费订阅、标准 API Key 或自托管的本地模型使用。
|
|
6
7
|
|
|
7
|
-
|
|
8
|
-
服务、一个 API key，或者你自己机器上跑的模型。
|
|
8
|
+
**当前状态：早期阶段（v0.1.x），已可用于日常开发。** 适合个人日常主力使用、脚本/CI 自动化及小团队试用。目前尚不适合用于强监管或要求不可中断的生产流水线——详见[生产与安全性](#生产与安全性)。
|
|
9
9
|
|
|
10
|
-
|
|
11
|
-
|
|
12
|
-
|
|
13
|
-
|
|
10
|
+
## 为什么选择 keke？
|
|
11
|
+
|
|
12
|
+
- **协议：为所有客户端提供 ACP 支持**
|
|
13
|
+
无论是外部客户端集成，还是内部 TUI 与 agent 之间的解耦边界，统一基于开放的 Agent Client Protocol (ACP)。
|
|
14
|
+
- **脚本与 CI 优先（`keke exec`）**
|
|
15
|
+
开箱即用支持非交互式一次性执行，专为脚本自动化与 CI 流水线设计。
|
|
16
|
+
- **与厂商隔离的引擎核心**
|
|
17
|
+
`keke-core` 内部没有任何厂商特定的逻辑。添加标准模型端点无需修改任何引擎代码——只需在 `config.toml` 中简单声明配置即可。
|
|
14
18
|
|
|
15
19
|
## 安装
|
|
16
20
|
|
|
@@ -20,8 +24,7 @@ GUI（编辑器、Zed 之类）都能像终端一样驱动它。如果想在脚
|
|
|
20
24
|
curl -fsSL https://raw.githubusercontent.com/milisp/keke-agent/main/scripts/install.sh | sh
|
|
21
25
|
```
|
|
22
26
|
|
|
23
|
-
|
|
24
|
-
（可通过 `KEKE_INSTALL_DIR` 覆盖）。
|
|
27
|
+
该脚本会下载适合你平台的最新预编译二进制文件至 `~/.local/bin`（可通过 `KEKE_INSTALL_DIR` 覆盖）。将远程脚本管道传输到 `sh` 会以你的权限执行——如介意可先检查脚本：`curl -fsSL .../install.sh | less`。
|
|
25
28
|
|
|
26
29
|
### npm 安装
|
|
27
30
|
|
|
@@ -29,71 +32,45 @@ curl -fsSL https://raw.githubusercontent.com/milisp/keke-agent/main/scripts/inst
|
|
|
29
32
|
npm install -g @milisp/keke
|
|
30
33
|
```
|
|
31
34
|
|
|
32
|
-
你也可以直接从 [
|
|
33
|
-
下载预编译二进制文件，或者用 `cargo build --release` 从源码构建。
|
|
35
|
+
你也可以直接从 [最新 Release](https://github.com/milisp/keke-agent/releases/latest) 下载二进制文件，或使用 `cargo build --release` 从源码构建。
|
|
34
36
|
|
|
35
|
-
##
|
|
37
|
+
## 快速试用（30 秒）
|
|
36
38
|
|
|
37
39
|
```sh
|
|
38
|
-
#
|
|
40
|
+
keke doctor # 在使用前检查哪些 provider/登录配置能正常解析
|
|
41
|
+
|
|
42
|
+
# 使用已有付费订阅登录……
|
|
39
43
|
keke login codex
|
|
40
44
|
keke login grok
|
|
45
|
+
# ……或直接配置 API Key
|
|
46
|
+
export ANTHROPIC_API_KEY=sk-ant-...
|
|
41
47
|
|
|
42
|
-
#
|
|
43
|
-
export XAI_API_KEY=xai-...
|
|
44
|
-
|
|
45
|
-
keke exec "what does this project do?" # 一次性运行，适合脚本和 CI
|
|
48
|
+
keke exec "what does this project do?" # 一次性运行，适合脚本与 CI
|
|
46
49
|
keke # 交互式 TUI
|
|
47
|
-
keke resume #
|
|
48
|
-
keke doctor # 查看哪些 provider 和登录能解析成功
|
|
50
|
+
keke resume # 恢复上一次的对话
|
|
49
51
|
```
|
|
50
52
|
|
|
51
|
-
|
|
52
|
-
|
|
53
|
-
| Provider | 如何认证 | 说明 |
|
|
54
|
-
| --- | --- | --- |
|
|
55
|
-
| OpenAI / ChatGPT | `keke login codex`，或 `OPENAI_API_KEY` | OAuth 流程从 codex 移植而来 |
|
|
56
|
-
| xAI Grok | `keke login grok`，或 `XAI_API_KEY` | 内置，默认 provider |
|
|
57
|
-
| Anthropic | `config.toml` 中的 `env_key` | 声明时用 `wire = "messages"` |
|
|
58
|
-
| 本地模型（Ollama、vLLM 等） | 无需认证 | 代码不会离开这台机器 |
|
|
59
|
-
| 任意 OpenAI 兼容网关 | `config.toml` 中的 `env_key` | 公司代理、NVIDIA NIM、路由器等 |
|
|
53
|
+
配置文件位置、Provider 声明、API Key、本地模型、多账号、按目录切换账号及其他设置，详见 [`docs/config.md`](docs/config.md)。
|
|
60
54
|
|
|
61
|
-
|
|
55
|
+
## 现状
|
|
62
56
|
|
|
63
|
-
|
|
64
|
-
[providers.ollama]
|
|
65
|
-
base_url = "http://localhost:11434/v1"
|
|
66
|
-
default_model = "gpt-oss:20b"
|
|
57
|
+
日常开发已完全可用。`keke exec`、TUI 和 ACP 服务器均支持端到端完整运行会话；运行时插件（skills、commands、hooks、MCP servers）遵循 Claude Code 格式安装，仓库自带的插件在你显式批准前会保持非激活状态。在运行中的会话内可通过 `/model` 切换模型（模型与 provider 绑定，避免配置持久化无效组合），并且 agent 支持生成子 agent——分配独立任务的子会话，仅汇总返回最终结果而不会污染完整的搜索 trace。后续规划请参见 [`docs/ROADMAP.md`](docs/ROADMAP.md)。
|
|
67
58
|
|
|
68
|
-
|
|
69
|
-
base_url = "https://api.anthropic.com"
|
|
70
|
-
env_key = "ANTHROPIC_API_KEY"
|
|
71
|
-
wire = "messages"
|
|
72
|
-
```
|
|
59
|
+
## 生产与安全性
|
|
73
60
|
|
|
74
|
-
|
|
75
|
-
`messages`——所以接入新接口只是一条配置，而不是一次发版。
|
|
61
|
+
目前非常适合个人使用、本地/自托管模型以及 CI 一次性任务。在将其托付给任何不容出错的关键场景前，请评估以下事项：
|
|
76
62
|
|
|
77
|
-
|
|
63
|
+
- **沙盒与审批** — 默认的 `approval_policy` 和 `sandbox_mode` 策略较为保守，但建议确认其符合你的运行环境与需求（[`docs/config.md`](docs/config.md)）。
|
|
64
|
+
- **插件信任机制** — 仓库自带的插件（hooks、MCP servers）在单纯 `git clone` 后绝不会自动执行；必须由人工确认批准，且该授权基于命令的具体内容而非文件路径。系统未提供任何可全局绕过此门控的开关。
|
|
65
|
+
- **成熟度** — 项目处于早期阶段（v0.1.x），目前由单人维护，暂无正式的安全审计或 SLA。路线图中的待补齐项（完整的端到端 MCP 工具调用校验）记录在 [`docs/ROADMAP.md`](docs/ROADMAP.md) 中。
|
|
78
66
|
|
|
79
|
-
|
|
80
|
-
|
|
81
|
-
安装，仓库自带的插件在你批准之前始终处于未激活状态。在运行中的会话里切换
|
|
82
|
-
模型这个功能还没实现——详见 [`docs/ROADMAP.md`](docs/ROADMAP.md)。
|
|
67
|
+
**适合的场景：** 希望摆脱厂商锁定、需要脚本/CI 友好的一次性运行模式、使用本地模型，或希望将 ACP 集成至自己的编辑器/客户端中。
|
|
68
|
+
**暂不适合的场景：** 需要针对所有厂商的即插即用式开箱即用 OAuth、成熟的插件生态市场，或需要商业支持合同。
|
|
83
69
|
|
|
84
|
-
##
|
|
70
|
+
## 灵感与来源
|
|
85
71
|
|
|
86
|
-
keke
|
|
87
|
-
**grok-build**，以及 **deepseek-harness**——写出来的全新实现，其中
|
|
88
|
-
deepseek-harness 提倡的 seam-first 架构（引擎和厂商之间是硬边界，而不是
|
|
89
|
-
约定）是 keke 最依赖的一个想法。少数部分是直接移植过来的，并在承载它的
|
|
90
|
-
crate 中注明来源（比如 codex 的 OAuth 登录流程）；其余大部分都是 keke
|
|
91
|
-
自己的代码，只是被这三者的经验（包括踩过的坑）塑造出来的。keke 的不同
|
|
92
|
-
之处、以及为什么值得在 CI 里强制执行，详见
|
|
93
|
-
[`docs/architecture.md`](docs/architecture.md)；不变量本身在
|
|
94
|
-
[`AGENTS.md`](AGENTS.md)。
|
|
72
|
+
keke 是一个全新的实现，参考并吸取了 OpenAI 的 **codex**、xAI 的 **grok-build** 以及 **deepseek-harness** 的设计思路。少数模块直接移植并在对应 crate 中注明来源；大部分代码均为原创。设计理念与 CI 强制执行的不变量详见：[`docs/architecture.md`](docs/architecture.md)、[`AGENTS.md`](AGENTS.md)。
|
|
95
73
|
|
|
96
74
|
## 许可证
|
|
97
75
|
|
|
98
|
-
Apache-2.0。详见 [`LICENSE`](LICENSE) 与 [`NOTICE`](NOTICE)
|
|
99
|
-
的代码会在承载它的 crate 的 `THIRD_PARTY_NOTICES.md` 中注明来源。
|
|
76
|
+
Apache-2.0。详见 [`LICENSE`](LICENSE) 与 [`NOTICE`](NOTICE)；从其他项目移植的代码在相应 crate 的 `THIRD_PARTY_NOTICES.md` 中注明来源。
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@milisp/keke",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.9",
|
|
4
4
|
"description": "keke is a multi-vendor terminal coding agent.",
|
|
5
5
|
"license": "Apache-2.0",
|
|
6
6
|
"bin": {
|
|
@@ -18,11 +18,11 @@
|
|
|
18
18
|
"win32"
|
|
19
19
|
],
|
|
20
20
|
"optionalDependencies": {
|
|
21
|
-
"@milisp/keke-darwin-x64": "0.1.
|
|
22
|
-
"@milisp/keke-darwin-arm64": "0.1.
|
|
23
|
-
"@milisp/keke-linux-x64": "0.1.
|
|
24
|
-
"@milisp/keke-linux-arm64": "0.1.
|
|
25
|
-
"@milisp/keke-win32-x64": "0.1.
|
|
21
|
+
"@milisp/keke-darwin-x64": "0.1.9",
|
|
22
|
+
"@milisp/keke-darwin-arm64": "0.1.9",
|
|
23
|
+
"@milisp/keke-linux-x64": "0.1.9",
|
|
24
|
+
"@milisp/keke-linux-arm64": "0.1.9",
|
|
25
|
+
"@milisp/keke-win32-x64": "0.1.9"
|
|
26
26
|
},
|
|
27
27
|
"repository": {
|
|
28
28
|
"type": "git",
|