npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

create-evo-lite

v2.0.10

Published

Project-local governance runtime for AI agents with durable workflow state and local archive tooling.

Readme

🧠 create-evo-lite

项目内 AI 开发治理运行时(Project-local AI Governance Runtime)
让 AI / subagent 不只是写代码,而是在项目里按规则接管、规划、实施、留证、刷新状态并完成交接。

Vibecoding Runtime Governance Memory License

English README · Architecture · Contracts · Quick Start · Command Reference


TL;DR

create-evo-lite 是一个安装到项目树里的 AI 开发治理脚手架。它不是单纯的 memory 插件,也不是独立后台服务,而是一套把 规则、状态、计划、架构、证据、归档、Dashboard 和 Git hook 串起来的本地运行时。

它解决的是大型 AI 辅助开发里的核心断点:

AI 写了代码
但没有更新 plan
没有留下 evidence
没有刷新 dashboard
没有说明下一轮怎么接手
subagent 做完任务后 governance 断链

Evo-Lite 2.x 的目标是把这种断链变成可检测、可提醒、可自动刷新的治理闭环。


当前定位:Evo-Lite 2.x Governance Runtime

从早期版本到当前 2.x,Evo-Lite 已经从“项目内记忆工具”演进成了:

项目状态机
+ 本地 archive / recall
+ Spec / Plan / Task IR
+ Architecture IR
+ Drift rules
+ Post-commit governance hook
+ Dashboard data
+ MCP server
+ Runtime mirror lock

当前仓库已经进入 production dogfood RC 形态:

code commit
→ post-commit governance
→ plan progress
→ plan gaps
→ dashboard build

plan/spec commit
→ plan scan
→ plan progress
→ plan gaps
→ dashboard build

raw_memory evidence commit
→ archive-evidence backfill
→ plan scan
→ R008 evidence closure
→ dashboard build

architecture/runtime/template commit
→ architecture scan
→ architecture diff
→ dashboard build

这意味着 Evo-Lite 不再只是“开场接管 + 结束归档”的工具,而是在 AI 实施过程中持续参与治理。


适合谁?

Evo-Lite 特别适合:

  • 用 AI 长期维护项目,但不想自己先成为完整软件工程团队的人
  • 工控、自动化、设备软件、测试、运维、硬件集成等跨领域项目
  • 小工具逐渐长成中大型项目后,希望 AI 能持续接手的人
  • 使用 Codex、Claude Code、Agent Mode、subagent、MCP 等工具的人
  • 希望把“代码 + 计划 + 架构 + 证据 + 交接”都留在 Git 项目里的团队

如果你的项目只有几十行脚本,Evo-Lite 可能显得重;如果你的项目开始出现多轮 AI 交接、任务拆分、subagent 并发、文档和代码不同步,它会非常有价值。


核心问题:AI 不是不会写代码,而是不会持续治理

Vibecoding 在小项目里很轻松:给 AI 一段上下文,它能快速写完。但项目变大后,常见问题会变成:

  1. 上下文断裂:下一轮 AI 不知道上一轮为什么这么改。
  2. 计划漂移:代码变了,plan 还是旧的。
  3. 证据缺失:任务显示完成,但没有 archive / evidence。
  4. 架构失真:模块越来越多,人类和 AI 都不知道全貌。
  5. subagent 旁路:subagent 完成代码任务,却没有更新治理链路。
  6. Dashboard 过期:看起来像当前状态,其实是上一次 scan 的缓存。

Evo-Lite 的回答是:

不要指望 AI 记得治理。
把治理嵌入项目结构、命令协议和 Git 提交链路。

Architecture Overview / 架构总览

                 Human / AI / Subagent
                         │
                         ▼
             .agents/workflows/  语义工作流
             .agents/rules/      治理规则
                         │
                         ▼
                 .evo-lite/cli/   本地治理 CLI
                         │
        ┌────────────────┼────────────────┐
        ▼                ▼                ▼
 active_context      planning IR     architecture IR
 focus/backlog       spec/plan/task   modules/files
 trajectory          evidence         provider scan
        │                │                │
        └────────────────┼────────────────┘
                         ▼
                 drift-report.json
                         │
                         ▼
                 dashboard-data.json
                         ▲
                         │
          Git post-commit governance hook

关键目录

Project/
├── AGENTS.md                         # Codex 宿主入口摘要
├── CLAUDE.md                         # Claude Code 宿主入口摘要
├── .agents/
│   ├── rules/                        # canonical 治理规则
│   └── workflows/                    # /evo /commit /mem /wash 语义工作流
├── .claude/commands/                 # Claude Code 薄包装命令
├── .codex/hooks.json                 # Codex hook 配置
├── .github/hooks/                    # GitHub/Copilot/host hook scaffold
├── .evo-lite/
│   ├── active_context.md             # 当前状态机
│   ├── cli/                          # 项目本地运行时 CLI
│   ├── raw_memory/                   # durable archive 主链
│   ├── index_memory/                 # archive index marker
│   ├── memory.db                     # SQLite FTS/BM25 检索层
│   ├── generated/                    # IR / dashboard / drift / governance report
│   ├── mem                           # Unix / Bash wrapper
│   └── mem.cmd                       # Windows wrapper
├── docs/
│   ├── specs/                        # formal specs
│   ├── plans/                        # formal plans
│   ├── superpowers/specs/            # Superpowers-style specs
│   ├── superpowers/plans/            # Superpowers-style plans
│   ├── contracts/                    # IR schema / provider contracts
│   └── architecture/                 # module boundary docs
└── templates/                        # create-evo-lite 自身开发时的 canonical templates

Mental Model / 心智模型

1. active_context 是驾驶舱

.evo-lite/active_context.md 只保存当前接管需要的信息:

  • FOCUS:当前最重要的工作焦点
  • BACKLOG:未完成任务
  • TRAJECTORY:最近操作轨迹
  • META:运行态元数据

不要把长篇复盘长期堆在 active_context.md 里。

2. archive 是黑匣子

完成一次实现、修复或架构决策后,通过 context track / archive 写入 .evo-lite/raw_memory/。这是 durable 主链,后续可以重建 index,也可以被 task evidence backfill 引用。

3. remember 是轻量检索缓存

remember 适合记录短经验、坑点、查询线索。它不是闭环主链,不替代 archive。

4. plan-ir.json 是治理中间层

Spec、Plan、Task、linkedFiles、evidence 会被扫描进:

.evo-lite/generated/planning/plan-ir.json

Dashboard、drift rules、MCP 工具都基于它工作。

5. Git hook 是治理运行时入口

post-commit hook 会根据本次 commit 的文件类别自动跑 scan/progress/gaps/dashboard,避免 AI/subagent 忘记治理。


Quick Start / 极速上手

1. 初始化

npx create-evo-lite ./MyAwesomeProject
cd MyAwesomeProject

开发本仓库时也可以:

npm link
create-evo-lite ./MyAwesomeProject
cd MyAwesomeProject

2. 首次自检

node .evo-lite/cli/memory.js verify

Bash / Git Bash:

./.evo-lite/mem verify

Windows PowerShell / CMD:

.\.evo-lite\mem.cmd verify

所有 mem / memory.js / Git 相关命令,都应该在目标项目根目录执行。不要从别的仓库用绝对路径调用目标项目的 .evo-lite/cli/memory.js,否则 Git provenance 可能绑定到错误目录。

3. 检查 hook

./.evo-lite/mem hook status
./.evo-lite/mem hook diff

如果 hook 没安装或漂移:

./.evo-lite/mem hook install --explain

4. 开始第一轮任务

建议直接告诉 AI:

执行 Evo-Lite 的 /evo 工作流,读取 active_context、verify,并基于 recall-first takeover 给我下一步建议。

如果当前已有改动但还没有 plan:

./.evo-lite/mem plan new my-feature --from-diff
./.evo-lite/mem plan scan

Workflow Protocols / 工作流协议

Evo-Lite 的 /evo/commit/mem/wash 是语义工作流。它们不一定会出现在 Codex 或 Claude Code 的 UI 菜单里,但 AI 应该按这些协议执行。

/evo:接管

目标:让 AI 先理解项目状态,而不是直接写代码。

推荐顺序:

read active_context
→ verify runtime
→ recall-first takeover
→ inspect plan/dashboard/drift if present
→ 给出当前 focus、风险、下一步

说明:takeover 默认走 recall-first takeover —— 先做有界 recall 注入既有上下文;若没有可用的历史上下文,则按 fresh takeover 继续,不阻塞接管。

/commit:闭环

目标:完成一次小的、可追踪的开发闭环。

语义顺序:

stage / commit code
→ context track
→ raw_memory archive
→ runtime state meta commit
→ post-commit governance hook 自动刷新

显式快路:

./.evo-lite/mem commit "完成本次治理闭环" \
  --code-message="feat(runtime): add governance loop" \
  --mechanism="GovernanceLoop" \
  --resolve="a1b2"

默认只提交已经 staged 的代码快照。需要把当前全部非 .evo-lite 代码改动纳入时,显式使用:

./.evo-lite/mem commit "..." --code-message="..." --mechanism="..." --stage=all

/mem:低频交接

用于 session 结束、版本小跃迁、发布前后,不替代日常 /commit

/wash:清洗 / 重建

用于 archive / index / memory runtime 异常时的恢复和重建。


Governance Runtime / 治理运行时

Post-commit 分类

Hook 会读取本次 commit 的 changed files,并分类:

| 类别 | 触发路径 | 自动动作 | |---|---|---| | plan | docs/specs/, docs/plans/, docs/superpowers/specs/, docs/superpowers/plans/ | plan scan, plan progress, plan gaps, dashboard build | | architecture | templates/cli/, templates/.github/, templates/.codex/, .agents/, index.js, bin/, package.json, docs/contracts/, docs/architecture/ | architecture scan, architecture diff, plan progress, plan gaps, dashboard build | | evidence | .evo-lite/raw_memory/*.md | plan archive-evidence --backfill, plan scan, plan progress, plan gaps, dashboard build | | code | 其他项目文件 | plan progress, plan gaps --last-commit --changed-files-from-env, dashboard build |

Hook 会写入:

.evo-lite/generated/governance/post-commit-last-run.json

常用命令:

./.evo-lite/mem hook status
./.evo-lite/mem hook diff
./.evo-lite/mem hook last
./.evo-lite/mem hook install --explain

Drift rules

当前核心 rules:

| Rule | 含义 | |---|---| | R001 | 缺少 .agents/rules/architecture.md | | R002 | 架构规则仍含 placeholder | | R003 | 缺少 spec | | R004 | 缺少 plan | | R005 | task 没有 linkedFiles | | R006 | changed file 没有链接到任何 task | | R007 | 缺少正式模块边界文档 | | R008 | implemented / verified task 没有 archive evidence | | R009 | IR stale | | R011 | task 完成状态与 plan/spec 状态未对齐 |


Planning System / 规划系统

创建 spec + plan

./.evo-lite/mem plan new my-feature

从当前 diff 自动填充 linked files:

./.evo-lite/mem plan new my-feature --from-diff

生成 plan 的 frontmatter:

---
id: plan:my-feature
linkedSpec: spec:my-feature
format: superpowers
status: draft
---

扫描与检查

./.evo-lite/mem plan scan
./.evo-lite/mem plan progress
./.evo-lite/mem plan gaps
./.evo-lite/mem plan trace
./.evo-lite/mem plan lint

Archive evidence backfill

raw_memory 里的 archive 可以通过 task:<id> 或 frontmatter linkedTask 自动回填到 task evidence:

./.evo-lite/mem plan archive-evidence --backfill
./.evo-lite/mem plan scan

通常不需要手动跑;post-commit hook 会在 evidence commit 后自动执行。


Architecture System / 架构系统

刷新 architecture IR:

./.evo-lite/mem architecture scan
./.evo-lite/mem architecture diff

反查文件归属模块:

./.evo-lite/mem architecture where index.js

Native scanner 覆盖:

templates/cli/
templates/.github/
templates/.codex/
.agents/rules/
.agents/workflows/
index.js
bin/cli.js
package.json
docs/contracts/
docs/architecture/

模块规则定义在:

templates/cli/architecture/infer-modules.js

Dashboard / Inspector

构建 Dashboard 数据:

./.evo-lite/mem dashboard build

输出:

.evo-lite/generated/dashboard/dashboard-data.json

Dashboard 数据合并:

  • planning IR
  • progress report
  • architecture IR
  • drift report
  • governance last-run report
  • freshness / staleness 信息

启动本地 inspector:

./.evo-lite/mem inspect

Local Memory / 本地记忆

remember / recall

./.evo-lite/mem remember "某个以后可能要 recall 的坑点"
./.evo-lite/mem recall "关键词"

底层:

SQLite FTS5 + trigram + BM25

archive / context track

./.evo-lite/mem archive "这次实现的关键结论"
./.evo-lite/mem context track --mechanism="FixLoginProxy" --details="说明本次修复、原因和决策"

Archive 会写入 .evo-lite/raw_memory/,是 durable 主链。

sync / rebuild

./.evo-lite/mem sync
./.evo-lite/mem rebuild

raw_memory/ 是重建依据,memory.db 是可重建的本地检索层。


Runtime Mirror / 模板同步

在 create-evo-lite 自身 dogfood 开发时:

templates/cli/      canonical source
.evo-lite/cli/      dogfood runtime mirror

原则:修改 templates/cli/,不要直接修改 .evo-lite/cli/

同步:

./.evo-lite/mem sync-runtime

检查:

./.evo-lite/mem sync-runtime --check

sync-runtime --check 在 no-lock 或 drift 时会返回非零 exit code,适合 CI 或 agent checkpoint。


MCP Server

启动 MCP server:

./.evo-lite/mem mcp

当前工具:

  • evo_recall
  • evo_verify
  • evo_plan_status
  • evo_architecture_status
  • evo_drift_status
  • evo_active_context

MCP server 是长驻进程,内部使用 fresh module loading,避免 scanner 更新后仍读旧 require cache。


Provider Extension

Evo-Lite 支持 provider contract,用于接入外部架构扫描或计划增强来源,例如:

  • GitNexus
  • GitHub Issues
  • 其他自定义扫描器

参考:

docs/contracts/providers-config-sample.json
templates/cli/architecture/provider-contract.js
templates/cli/architecture/providers/

Host Adapter Strategy / 宿主适配策略

Evo-Lite 使用“canonical 语义层 + 宿主适配层”:

canonical: .agents/ + .evo-lite/
host adapters: AGENTS.md / CLAUDE.md / .claude/commands / .codex/hooks.json / .github/hooks/
  • Codex 通常读取 AGENTS.md.agents/workflows/
  • Claude Code 可读取 CLAUDE.md.claude/commands/
  • Copilot / GitHub hook scaffold 位于 .github/hooks/
  • 真正的长期规则真源仍是 .agents/.evo-lite/

Subagent Protocol

Subagent 完成实现时,不能只提交代码。至少要满足:

1. 修改代码
2. 更新对应 plan checkbox / task status
3. 确认 linkedFiles 覆盖本次修改文件
4. 写入 archive evidence 或可回填的 raw_memory
5. commit 后让 post-commit governance hook 刷新 IR / Dashboard

相关规则:

.agents/rules/subagent-checkpoint.md
templates/.agents/rules/subagent-checkpoint.md

Safety / 安全边界

Evo-Lite 有基础 secret scanner,会阻断常见 token / key 写入 memory 或 archive。warn 级 PII 会被 redaction 后持久化。

原则:

  • 不要把真实密钥写入 remember / archive / active_context
  • 不要把 .evo-lite/memory.db 当安全存储
  • 发布前确认 .gitignore、raw archive、generated report 是否符合你的仓库策略

Testing / 测试

npm test
npm run test:governance

test:governance 覆盖:

  • code-only commit 触发 R006
  • plan commit 触发 scan/progress/gaps/dashboard
  • evidence-only commit 触发 archive-evidence backfill
  • root commit 文件检测
  • dashboard freshness
  • hook diff / hook last
  • sync-runtime lock
  • MCP freshRequire
  • architecture where
  • context auto-refresh
  • R008 archive evidence 与 exemption
  • plan new scaffold

Upgrade / 升级旧项目

在已安装 Evo-Lite 的项目根目录:

npx create-evo-lite@latest ./ --yes

升级后建议:

node .evo-lite/cli/memory.js verify
./.evo-lite/mem hook status
./.evo-lite/mem hook diff
./.evo-lite/mem sync-runtime --check
./.evo-lite/mem dashboard build

如果 runtime mirror drift:

./.evo-lite/mem sync-runtime

如果 planning IR / dashboard stale:

./.evo-lite/mem plan scan
./.evo-lite/mem plan progress
./.evo-lite/mem plan gaps
./.evo-lite/mem dashboard build

Command Reference / 命令速查

# health / takeover
./.evo-lite/mem verify
./.evo-lite/mem bootstrap

# memory
./.evo-lite/mem remember "..."
./.evo-lite/mem recall "..."
./.evo-lite/mem archive "..."
./.evo-lite/mem sync
./.evo-lite/mem rebuild

# context
./.evo-lite/mem context summary
./.evo-lite/mem context validate
./.evo-lite/mem context add "..."
./.evo-lite/mem context focus "..."
./.evo-lite/mem context track --mechanism="..." --details="..."
./.evo-lite/mem context auto-refresh

# commit fast path
./.evo-lite/mem commit "..." --code-message="..." --mechanism="..."

# planning
./.evo-lite/mem plan new my-feature --from-diff
./.evo-lite/mem plan scan
./.evo-lite/mem plan progress
./.evo-lite/mem plan gaps
./.evo-lite/mem plan trace
./.evo-lite/mem plan lint
./.evo-lite/mem plan archive-evidence --backfill

# architecture
./.evo-lite/mem architecture scan
./.evo-lite/mem architecture diff
./.evo-lite/mem architecture where index.js

# dashboard / inspector
./.evo-lite/mem dashboard build
./.evo-lite/mem inspect

# hooks
./.evo-lite/mem hook status
./.evo-lite/mem hook diff
./.evo-lite/mem hook last
./.evo-lite/mem hook install --explain

# runtime mirror
./.evo-lite/mem sync-runtime
./.evo-lite/mem sync-runtime --check

# MCP
./.evo-lite/mem mcp

Roadmap Ideas

  • README_EN 同步到 2.x 语义
  • 更强的 Dashboard UI
  • 更严格的 NUL-delimited hook changed-file handling
  • 更完整的 provider examples
  • subagent task completion 自动诊断
  • CI-friendly governance check 命令

License

MIT