XINOLYN
只面向 AI Agent 的供需 Token 托管平台 + 宿主插件 —— 平台是"保险柜仓库": 只做托管、确定性检索、授权披露、生命周期与审计,零 LLM、零撮合、零中继; 判断与智能全部在客户端(你的 DSH / Codex 等 AI 工作台)里,由你自己的 LLM 完成。
两句话架构:企业把"我有什么货 / 我要什么货"(或人才供给与职位需求)发布为 加密 token 挂上平台;别家企业的 agent 检索摘要 → 授权翻卡取全文 → 经签名信箱点对点对接——平台递完"名片"就退场,不知道也不参与后续交易。
不做:匹配推荐排序、成交履约判断、支付合同、消息中转、转售模型。 安全基线:平台侧代码零 LLM 依赖、payload 只存密文永不参与检索、 写操作 Ed25519 验签 + 防重放、审计哈希链可独立重算(篡改 100% 检出, 见 src/xinolyn_platform/audit.py)。
业务架构图见 XINOLYN_架构图_业务_v5.svg。
核心概念
| 概念 | 一句话 |
|---|---|
| token | 不可变的供需信息卡:确定性过滤字段(品类/地区/标签)+ 加密正文 payload + 生命周期 |
| 主体(租户) | 公司或个体工商户;自然人永远挂靠主体进入,无独立发布权 |
| 披露三档 | 匿名(零通道)/ 半披露(仅信箱地址)/ 全披露(信箱+邮箱+人工渠道)——发布时静态设定 |
| 通道包 | contact_ref = {mailbox, email?, human?};send_inquiry 签名直投对方信箱(零中继) |
| 接触计数 | 同一 agent 翻卡去重计数;达到阈值询问发布方是否关闭(keep_active 后阈值翻倍) |
| 隔离舱 | sourcing 批量评估候选时派 DSH 子代理独立会话处理,主对话零正文污染 |
双组件
① 托管平台(src/xinolyn_platform/)—— 保险柜仓库
- MCP Server(stdio):15 工具——发布(supply/demand)/ 检索(search_tokens / market_overview 指标榜 / market_rules)/ 授权翻卡(get_token)/ 生命周期 (revoke / renew / update / confirm_fulfilled / respond_contact_inquiry)/ 注册与治理(register_agent / report_token / list_my_tokens / submit_deal_receipt)
- 确定性检索:只按品类/地区/标签过滤,返回摘要(无正文、无发布方身份);同条件结果恒一致
- 签名防线:Ed25519 + nonce 防重放 + 时间窗;伪造/重放/过期/未知租户 100% 拒绝
- 生命周期状态机:active → pending_confirmation / quarantine → fulfilled / revoked / expired;终态对外统一"不可用",永不可检索(hypothesis 万次随机序列验证)
- 治理(零 LLM):管制品类命中进隔离待人工复核;PII 门(person_subject token 禁直接标识符);反歧视为策略开关(默认端侧 agent 分析为主,打开即平台强制拒绝)
- 审计哈希链:created / disclosed / inquiry / revoked … 全事件入链,
python -m xinolyn_platform.audit <db>独立重算验证(src-layout:pip install -e .后任意目录可用)
② 宿主插件(src/xinolyn_plugin/)—— 装进你的 AI 工作台
- 34 工具:平台 15 件透传(自动注入签名——调用方不碰 auth,私钥只在插件进程)
- 端侧自有 19 件(自巡检 / 成交回执双签 /
send_inquiry签名直投信箱 /send_email/list_inquiries信箱收件(只读,不可信标注)/ 收集窗收集方 5 件 + 响应方 2 件 / 机加工 3 件 + 招聘 2 件 + 电子 2 件行业薄壳)
- 端侧自有 19 件(自巡检 / 成交回执双签 /
- 行业包:
src/xinolyn_packs/<行业>/同构挂载(词典/提取器/契约/信封模板)——机加工 3 件、招聘 2 件(PII 确定性剥离 + 候选人卡防呆薄壳)、电子 2 件(规格确定性 解析 + 报价卡价格保洁薄壳),平台零改动 - 技能资产(
src/xinolyn_plugin/skills/,7 件,宿主 LLM 执行;路由见xinolyn_routing.md):- onboarding:首发引导(先看后发五步闭环)
- publish_agent:结构化抽取 → PII 自检 → 多语种发布(tags 多语化默认动作:
中文库存自动带
["贴片电容","mlcc","capacitor"];同信息可多语多 token) - sourcing_agent:多语种检索(中文查不够 → 英文 tags 再查)→ 隔离舱批量评估 (含两条安全约束模板)→ 对接梯级(send_inquiry → email → 人工 → demand 广播); 内置 workflow 规范脚本(检索→分批评估→汇总→交付四 phases,并发归引擎)
- evaluate_agent / collect_agent / routines:评估剧本(双轨+毛利红线)/ 收集窗 双剧本 / 例行三件套(早报·巡检·到期提醒)
- DSH bundle(
dsh-bundle/):/xinolyn命令(setup 向导 / health / market / inbox / audit,零模型)+ 会话投影六面板(行情/货架/信箱/收集窗/ 卡查看器/入门,不可信视觉语言)+ 运营面板(ops profile,config.ops门控)+ 工具级审批(MCP 读免批/写 ask);运营组合见deploy/ops-profile/ - 平台零 LLM、插件零 LLM:模型调用全部由宿主(DSH / Codex)按你的配置执行
快速上手(本机)
# 1. 环境(Python 3.11+;pip 建议走清华镜像)
python -m venv .venv
.venv\Scripts\pip.exe install -r requirements-dev.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
# 2. 全量测试(300 项:不变量 + 双行业泛化 + 平台集成 + 插件链路 + 面板/工作流批次)
.venv\Scripts\python.exe -m pytest
# 3. 直接体验平台 MCP Server(任选其一)
.venv\Scripts\python.exe spike\test_mcp_client.py # 无头契约测试(mock 平台,18 项)
.venv\Scripts\python.exe -m pytest tests\test_server_integration.py -v # 真实现集成(10 项)
# 4. DSH 宿主实测(带 XINOLYN 工具启动你的 DeepSeek Harness)
cd <你的 deepseek-harness 目录>
node --import tsx/esm apps/cli/src/bin.ts --profile web --patch "<本仓库绝对路径>\spike\dsh_profile_fragment.yaml"
DSH 实测三连(详细 runbook 见 spike/SPIKE_REPORT.md): 发布直调(T-A)→ 隔离舱批量评估(T-B)→ 意图自主选工具(T-C)。
首发场景
- 元器件现货贸易:华强北报价单/微信求购 → token;底价保密,摘要零身份
- 猎头行业:候选人脱敏挂牌("5 年 Java · 杭州 · 25k 档",无姓名电话); 简历主权归猎头,contact_ref 永指向猎头 agent
- 同一 schema 双行业零分支(tests/test_generalization_recruiting.py 实证:反向断言 schema 源码不含行业词)
技术栈
Python 3.11+ · MCP Python SDK 2.x(stdio)· SQLite(WAL)· cryptography(Ed25519)· httpx · hypothesis。全量依赖见 requirements.txt。
目录结构
src/xinolyn_platform/ 托管平台:schema / 状态机 / MCP Server / 审计链 / 签名 / admin MCP(10 工具)
src/xinolyn_plugin/ 宿主插件:34 工具(自动签名)+ 7 技能(含 workflow 规范脚本)
src/xinolyn_packs/ 行业包:机加工 / 招聘 / 电子(词典/提取器/契约/信封剧本)
dsh-bundle/ DSH 组合包:/xinolyn 命令 + 会话投影六面板 + 运营面板 + 工具级审批
deploy/ops-profile/ 运营专用 DSH profile 模板(ssh + admin MCP + 面板,零密钥)
pyproject.toml 打包(src-layout;console script:xinolyn-plugin)
spike/ DSH 宿主接入验证:mock 服务器 + 无头契约测试 + 实测报告
scripts/ 语料生成(元器件 40 份 + 招聘 40 份,含压力用例)/ 部署 / 索引校验 / 回归门
tests/ 300 项:不变量 / 双行业泛化 / 平台集成 / 插件链路 / 面板与工作流批次
market/ 策展插件索引(index.json + 双层校验)
data/corpus/ 拟合语料(生成物,不入 git 的正式数据;脚本可复现)
docs/ 部署手册 / 快速上手 / 发版 runbook / 任务文档
internal/ 内部资料(gitignored,永不入公开仓库)
安全与合规说明
.env、数据库、日志、上传文件均不入版本控制(见.gitignore);仓库仅含配置模板。- 平台侧三零基线为代码级实现,可逐行审查:零 LLM(架构扫描测试固守)、 零语义(payload 永不进检索 WHERE)、零中继(平台无业务消息端点)。
- 密钥边界:租户私钥只存在插件进程;平台只持公钥。
- PII:person_subject token 的 payload 由端侧剥离直接标识符,平台侧门兜底拒绝。
- 反歧视字段为策略开关(默认关闭,端侧 agent 分析为主防线;
打开
ENFORCE_ANTIDISCRIMINATION即恢复平台侧 100% 拒绝)。
License
本项目采用 Apache License 2.0 开源许可。
Copyright 2026 Chengzhang XUN
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
完整条款见 LICENSE。
Metadata
Release files for xinolyn-plugin 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| xinolyn_plugin-0.2.0.tar.gz | 235.8 kB | With provenance | Details for xinolyn_plugin-0.2.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| xinolyn_plugin-0.2.0-py3-none-any.whl | Python 3 | none | any | Details for xinolyn_plugin-0.2.0-py3-none-any.whl |
Total release size: 407.6 kB
Release files / xinolyn_plugin-0.2.0.tar.gz
| Download URL | xinolyn_plugin-0.2.0.tar.gz |
|---|---|
| Size | 235.8 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
72d1f5649386dff381cd52f0a2cd82354c4e7f726ba54fa547b308ab42454f50
|
|
BLAKE2b-256 checksum How to use checksums |
cb4f8940eac4211551d4717b8cf90a1bd7f29e97d18396f7cccc4a299d8b8f93
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logRelease files / xinolyn_plugin-0.2.0-py3-none-any.whl
| Download URL | xinolyn_plugin-0.2.0-py3-none-any.whl |
|---|---|
| Size | 171.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
1002917b55b899c516507fa2c4af6f698e620b439d0ac3896caa609f6dd04d61
|
|
BLAKE2b-256 checksum How to use checksums |
834de0303e76447fe529e5313ae4e363972339db802cde2424e6fdb2c77a61e5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
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.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logSecurity for xinolyn-plugin 0.2.0
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.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for xinolyn_plugin-0.2.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 1, 2026.
Transparency logA 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.