Skip to main content

AIComposer 3.0.0

业务可见性优先的新范式框架——框架 = 可选能力层(opt-in), 不是强制约束层。 打开一个应用, 30 秒内知道它在干嘛; 开发者写应用 = 传统写法 + 两个 API(ctx/mount)。

教程(aic/tutorial/ 随 pip 包; aic skills 分发到 .claude/.codex/.agent)

01-quickstart.md     3 分钟跑起来(init → 打开 app.py → 写业务 → 跑)
02-plugin.md         插件 = 普通包(apply(ctx))
03-ctx-events.md     ctx 用法 + 事件四模式
04-toolchain.md      工具链(init/graph/template/skills)
05-real-app.md       真实业务对照(review / constplan 迁移形态)
06-sdk-reference.md  SDK 参考(3.0.0 完整 API)

怎么跑

python verify.py             # 内核机制: 装配/事件/卸载零残留/覆盖可恢复(9/9)
python verify_events.py      # 事件四模式: EMIT/WATERFALL/PARALLEL/SERIAL(5/5)
python verify_review.py      # review 真实业务: 提取→检索→引擎→报告→事件(9/9)
python verify_constplan.py   # ConstPlan 真实业务: 创建→并行章节→合并→渲染→版本化(11/11)
python verify_toolchain.py   # 工具链端到端: init→运行→graph→template→skills(8/8)

# 工具链(aic 命令; 或 pip install -e . 后 aic 全局可用)
python -m aic.cli init my_app --root <dir>          # 生成应用骨架
python -m aic.cli graph --app <app_dir>             # 计算图 graph-viz.html
python -m aic.cli template <src_app> --out <dir>    # 插件模板提取
python -m aic.cli skills --root <dir>               # skill 分发

全量验证 42/42(5 套, 经真实 FastAPI TestClient)。

框架包结构(aic/)

aic/
├── __init__.py   统一入口(ctx/mount/unmount/mounted/Task/EventMode)
├── context.py    能力容器: register/get/effect/on/emit + 事件四模式
├── assembly.py   装配: mount/unmount(可逆, 零残留/不误伤)+ mounted()
├── task.py       会话信封 Task(可选, 不约束)
├── cli.py        命令入口: init / graph / template / skills(UTF-8 输出)
├── init.py       骨架生成(app.py + plugins/ + 示例插件 + README)
├── graph.py      计算图(AST 扫描: 装配/能力/依赖/事件 → graph-viz.html)
├── template.py   插件模板提取(plugins/ 全带走, 代码可见可改)
└── skills.py     aic 范式 skill 分发(三平台)

开发者只需要知道两个概念

ctx    能力容器: register(提供) / get(取用) / effect(清理) / on/emit(事件)
mount  装配插件(可逆: 卸载零残留——后台机制)

机制(全部后台, 不要求理解——框架替你做的)

  • 可逆装配: mount/unmount 零残留、不误伤
  • 覆盖可恢复: ctx.register 同 key 覆盖 → 撤销恢复前一个(换实现零改动)
  • 事件四模式: EMIT(观察)/ WATERFALL(中间件)/ PARALLEL(并发)/ SERIAL(顺序加工) ——插件间通信机制, 框架职责, 不是业务胶水
  • 会话信封 Task(可选): 多插件协作统一上下文(身份/工作区/元数据/轮次)
  • 装配可计算: mounted() / aic graph(装配/依赖/能力/事件 → graph-viz.html)
  • 工具: aic init / template(骨架与模板, 代码可见可改)

示例应用(真实业务验证)

demo/          文件转换(converter 业务插件 + extract 公共插件)
review_app/    审查(extract/standard/engine 公共插件 + service 普通模块)
constplan_app/ 施工方案编写(迁移自 ConstPlan: 分章并行→合并→渲染→版本化)
  • 业务 = 普通模块普通函数, 唯一 AIC 痕迹 = ctx.get("...") 取能力(像用 NumPy)
  • 公共能力 = 插件, mount 注入; 换引擎 = 同 key 覆盖(业务零改动)
  • 落盘 = 直接写文件(无产物通道); DB/Celery/MinIO 等生产化设施 = 普通库用法

不做的事

  • 不强制业务进插件(业务默认写在 app.py/普通模块, 可复用才抽)
  • 不预设会话模型(业务自己管状态, 落盘直接写文件)
  • 无壳布局契约 / 无注入声明 / 无 promote 上浮(插件平铺, 无私有公共之分)
  • HTTP/路由不归 AIC 管(应用直接用 FastAPI, 零新概念)

打包安装

pip install -e .              # 装 ai-composer 3.0 → aic 命令全局可用
# 注意: 与旧版 ai-composer(同名 aic 包)冲突, 正式切换需先卸载旧版

Metadata

Release files for ai-composer 3.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ai-composer 3.3.0
File Size Uploaded Provenance
ai_composer-3.3.0.tar.gz 35.2 kB Without provenance Details

Built distribution (wheel)

Table of built distributions (wheels) for ai-composer 3.3.0
File Interpreter ABI Platform
ai_composer-3.3.0-py3-none-any.whl Python 3 none any Details

Total release size: 79.3 kB

Release files / ai_composer-3.3.0.tar.gz

Download URL ai_composer-3.3.0.tar.gz
Size 35.2 kB
Tags Source
SHA-256 checksum
How to use checksums
a9526fe67bcda5c5d9b728d50edbe6b84663a464d82f46272a5b5502a53c1a7c
BLAKE2b-256 checksum
How to use checksums
5b310dea77ad91d03eb332a177a964f04c0a2671a9eadace4c93ebdff66c5501
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.4

Release files / ai_composer-3.3.0-py3-none-any.whl

Download URL ai_composer-3.3.0-py3-none-any.whl
Size 44.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1a4e042b580f44c05ea31f1721ff7a91699d8f611c47d844a8c0770bb6818808
BLAKE2b-256 checksum
How to use checksums
7a01410a5e95f06696f67baac157f5eaeef4520a87a1265794a091db26043c82
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.4

Security for ai-composer 3.3.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page