A
alexnodeland/lattice
🧩 Python libraries for apps where people and agents work together — shared artifacts, rules over events,…
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"alexnodeland-lattice": {
"command": "uvx",
"args": ["copier"]
}
}
}解析自我们于 2026-10-29 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"alexnodeland-lattice": {
"command": "uvx",
"args": ["copier"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio alexnodeland-lattice -- uvx copier
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 1 ★ · alexnodeland/lattice · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-10-29 · 最近见到 2026-11-14 · 共见到 17 次 · 被 1 个来源列出
- 评分构成
- 综合分 103 · 1 ★ · 1 个来源 · repo
- 安装
uvx copier
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
uvx copier。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: 开源项目。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
alexnodeland/lattice；我们收录于 2026-10-29；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。Quick start ·
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
The packages ·
How it fits ·
Docs ·
How it's built ·
Contributing
lattice is a set of Python libraries for applications in which people and agents work together, and the stack they run on.
🤝 People and agents share versioned artifacts. An agent proposes a change, and a person accepts it.
⚡ Rules over event streams run agents, graphs and functions when something happens.
💬 People can draft those rules in chat (🚧 landing now). A rule goes live only once a person allowed to install it accepts.
📏 Typed evaluations learn from people's feedback.
🔐 Identity and access is checked before every command and read, down to PostgreSQL's row-level security.
🪟 A portal is where people use the application and run it.
🧱 One template and one stack hold it all, locally or on a server.
Each package keeps its own name, version, changelog and releases. They share one environment, one lockfile, one CI and one documentation site, so a change that spans packages lands in one pull request.
## 🧭 How it fits together
A person chats with an agent in artifactr, through portalr or any client of artifactr's REST, WebSocket and MCP surfaces. relayr turns what happens in the chat into events for reflexr's rules, and the rules' runs act back in the chat as proposals and notices. grantr decides who may do what, evalr measures how well the agents do it, and stackr runs it all. Underneath, the packages share lattice-core, the infrastructure they must keep identical.
🔁 Follow one rule from chat to live
mermaid
sequenceDiagram
autonumber
actor P as Person
participant A as artifactr
participant L as relayr
participant R as reflexr
participant G as grantr
P->>A: "Tell this thread whenever an alert fires"
A->>A: the agent drafts a rule, checks it and previews it
A-->>P: a proposal, with how often it would have fired
P->>A: accepts the proposal
A->>L: proposalresolved
L->>A: re-reads who accepted, and at which version
L->>G: may this person install rules here?
G-->>L: yes
L->>R: install the rule, with its provenance
L-->>A: notice: the rule is live
Note over R,A: from now on, each alert posts a notice in the thread
🚧 relayr's library does this today, and stackr's template is wiring it in. A forged acceptance installs nothing: relayr never trusts the event, only artifactr's record of it. A rule acts as itself, inside its own workspace, and can't post into a thread its workspace restricts. relayr's guide has the details.
## 🧩 The packages
| Package | What it is | Status |
| artifactr | Chat applications where people and agents collaborate through shared, versioned artifacts | ✅ released · 📦 PyPI |
| reflexr | Rules over event streams that run LLM workflows: pydantic-ai agents, pydantic-graph graphs and functions | ✅ released · 📦 PyPI |
| evalr | Typed evaluation of agent systems: judges that give typed verdicts, trained on people's feedback and measured against it | ✅ released · 📦 PyPI |
| relayr | The bridge between artifactr and reflexr: chat events become events for rules, and runs act back in the chat. Rules can be drafted in chat. | ✅ released · 📦 PyPI |
| grantr | Identity and access: tenants, members, groups, roles, service principals and restricted items, checked before every command and read, with row-level security | ✅ released · 📦 PyPI |
How it fits ·
Docs ·
How it's built ·
Contributing
lattice is a set of Python libraries for applications in which people and agents work together, and the stack they run on.
🤝 People and agents share versioned artifacts. An agent proposes a change, and a person accepts it.
⚡ Rules over event streams run agents, graphs and functions when something happens.
💬 People can draft those rules in chat (🚧 landing now). A rule goes live only once a person allowed to install it accepts.
📏 Typed evaluations learn from people's feedback.
🔐 Identity and access is checked before every command and read, down to PostgreSQL's row-level security.
🪟 A portal is where people use the application and run it.
🧱 One template and one stack hold it all, locally or on a server.
Each package keeps its own name, version, changelog and releases. They share one environment, one lockfile, one CI and one documentation site, so a change that spans packages lands in one pull request.
## 🧭 How it fits together
A person chats with an agent in artifactr, through portalr or any client of artifactr's REST, WebSocket and MCP surfaces. relayr turns what happens in the chat into events for reflexr's rules, and the rules' runs act back in the chat as proposals and notices. grantr decides who may do what, evalr measures how well the agents do it, and stackr runs it all. Underneath, the packages share lattice-core, the infrastructure they must keep identical.
🔁 Follow one rule from chat to live
mermaid
sequenceDiagram
autonumber
actor P as Person
participant A as artifactr
participant L as relayr
participant R as reflexr
participant G as grantr
P->>A: "Tell this thread whenever an alert fires"
A->>A: the agent drafts a rule, checks it and previews it
A-->>P: a proposal, with how often it would have fired
P->>A: accepts the proposal
A->>L: proposalresolved
L->>A: re-reads who accepted, and at which version
L->>G: may this person install rules here?
G-->>L: yes
L->>R: install the rule, with its provenance
L-->>A: notice: the rule is live
Note over R,A: from now on, each alert posts a notice in the thread
🚧 relayr's library does this today, and stackr's template is wiring it in. A forged acceptance installs nothing: relayr never trusts the event, only artifactr's record of it. A rule acts as itself, inside its own workspace, and can't post into a thread its workspace restricts. relayr's guide has the details.
## 🧩 The packages
| Package | What it is | Status |
| artifactr | Chat applications where people and agents collaborate through shared, versioned artifacts | ✅ released · 📦 PyPI |
| reflexr | Rules over event streams that run LLM workflows: pydantic-ai agents, pydantic-graph graphs and functions | ✅ released · 📦 PyPI |
| evalr | Typed evaluation of agent systems: judges that give typed verdicts, trained on people's feedback and measured against it | ✅ released · 📦 PyPI |
| relayr | The bridge between artifactr and reflexr: chat events become events for rules, and runs act back in the chat. Rules can be drafted in chat. | ✅ released · 📦 PyPI |
| grantr | Identity and access: tenants, members, groups, roles, service principals and restricted items, checked before every command and read, with row-level security | ✅ released · 📦 PyPI |
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/alexnodeland-lattice/)
- 来源
- github · https://github.com/alexnodeland/lattice
- 按领域浏览
- 开源项目
- Traction
- 1 ★
- 收录于
- 今天
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。