Training Guardian Agent · 训练守护智能体

一行命令，训练脚本零行改动，获得完整守护能力。

One command. Zero changes to your training script. Full guardian capabilities.

guarftrain init && guarftrain watch -- python train.py --epochs 20

---

What's New in v0.3.0

Feature	Description
架构分析 (Arch Analysis)	D3 treemap + backbone 可视化，FLOPs/参数量/瓶颈层检测，参考 archify 设计
远程通信 (Remote Server)	算力服务器端 FastAPI 服务，PC Dashboard 远程连接，鉴权 token
Sub-agent 自主决策	--autonomy supervised/auto/full，自主调整参数/干预训练
DSH Web GUI Plugin	DeepSeek Harness 侧栏面板，实时 metrics/GPU/anomalies/decisions/architecture/history（插件文档）
CPU 模式兼容	无 GPU 时自动降级，训练曲线正常显示，GPU 面板提示不可用
PyTorch >= 1.13 支持	resource_estimator 回退兼容 PyTorch 1.x
MCP 工具扩展	+1 analyze_architecture 工具（共 36 个）
Dashboard 架构分析标签	独立「架构分析」标签页，treemap/backbone 双视图

---

What's New in v0.2.0

Feature	Description
guarftrain CLI	pip install 后全局可用，替换旧 python run.py
guarftrain init	自动扫描训练脚本，生成 contract.yaml
guarftrain check	环境自检：Python/GPU/依赖/项目结构
Dashboard 远程配置	外部 Agent 通过 MCP 控制 Dashboard 图表/面板，用户操作受 dirty flag 保护
Agent 图表推荐	chart_selection 决策点：Agent 分析训练状态，推荐应关注的指标组
MCP 委托模式	外部 Claude Code 连接时内置 Agent 进入 provisional 模式，决策可被覆盖
增量图表更新	Dashboard 实时推送图表数据，不再全量重建
依赖瘦身	核心安装 ~2MB，torch/anthropic 按需安装

---

What does it do? · 它做什么？

Phase · 阶段	Capability · 能力	How · 方式
训练前 Pre-flight	GPU 显存预估 + batch 推荐	guarftrain preflight
训练中 During	GPU+Loss 监控告警 / 崩溃自动恢复 / LLM 决策 / Sub-agent 自主干预	guardian watch
训练后 Post	摘要+AI 解读 / Checkpoint 分析 / 模型可视化 / 架构分析	guarftrain summarize
跨实验 Cross	自然语言查询 / 实验对比 / 数据导入	guarftrain query "best lr?"
外部接入 External	MCP 36 工具 + Dashboard 远程配置 + Agent 图表推荐 + 远程通信	guarftrain start

Quick Start · 快速开始

Install · 安装

# 方式 1: pip 安装（推荐，轻量核心 ~2MB，torch 已有不重装）
pip install guarftrain

# 方式 2: 从源码安装
git clone https://github.com/Washington5533/Guid-traince.git
cd Guid-traince
pip install .

# 按需安装可选组件
pip install guarftrain[agent]       # AI 决策层 (anthropic)
pip install guarftrain[mcp]         # MCP 外部 Agent 接入
pip install guarftrain[dashboard]   # Web 控制面板
pip install guarftrain[full]        # 全部安装

Three steps to guard · 三步守护

# 1. 初始化项目（自动扫描训练脚本，生成配置）
cd /path/to/your-project
guarftrain init

# 2. 守护训练（纯规则，零外部依赖）
guarftrain watch -- python train.py --epochs 20

# 3. 或启用 AI + Dashboard + MCP
guarftrain watch --agent --with-dashboard --with-mcp -- python train.py --epochs 20

What does the training script need? · 训练脚本要满足什么？

Four contracts (script interface agreements). Each one gates a capability — missing one disables only that feature, training still runs normally.

1. --resume / --ckpt flags for checkpoint resumption → enables crash recovery + restart-based interventions
2. cp_{epoch}/model.pth with epoch/model_state_dict/optimizer_state_dict → enables checkpoint analysis + post-training tools
3. Structured logging: epoch {n} loss {v} val_acc {v} lr {v} → enables loss anomaly detection + progress monitoring
4. Importable entry: train:build_model / train:get_dataloaders → enables preflight resource estimation + model visualization + inference

Missing any one? Only the corresponding capability is disabled — training still runs.

四项契约（训练脚本的接口约定），每一项控制一个能力——缺任一项只关闭对应能力，不阻断训练。guarftrain init 会自动扫描你的脚本，逐项报告开启/降级状态。

Architecture · 架构

┌─ Guardian Process (sidecar) ────────────────────────────────────┐
│                                                                  │
│  CLI (guarftrain) ──→ 18 subcommands                              │
│  ├─ watch ──→ Watchdog: Popen + crash recovery + CLI rewrite    │
│  │             └─ Monitor: log tail + GPU poll + anomaly detect  │
│  │                  └─ AgentAdvisor: LLM decide → intervene       │
│  │                  └─ Sub-agent: --autonomy (supervised/auto/full) │
│  ├─ remote ──→ FastAPI 远程通信服务（算力服务器端）                │
│  ├─ serve ──→ MCP Server: 36 tools (25 read + 11 write)          │
│  ├─ start ──→ Dashboard + MCP one-click                        │
│  └─ experiments / query / compare ──→ Cross-experiment analysis  │
│                                                                  │
│  Decision Layers · 决策分层:                                      │
│  ┌─ Contract (hard boundary, human-defined)                     │
│  ├─ Agent (LLM, optional, within action space)                  │
│  ├─ Sub-agent (autonomous, --autonomy supervised/auto/full)     │
│  ├─ Rules (deterministic, always-on fallback)                   │
│  ├─ MCP (external agent access, dual-mode delegation)           │
│  └─ Dashboard (remote config, dirty-flag user protection)       │
│                                                                  │
│  Architecture Analysis · 架构分析:                                │
│  └─ ArchAnalyzer: forward hooks → FLOPs → tree → D3 render     │
│                                                                  │
│  Training Process: python train.py (0 changes required)          │
└──────────────────────────────────────────────────────────────────┘

CLI Commands · 命令速查

Command	Description
init	Auto-detect project + generate contract.yaml
check	Environment readiness check (deps, GPU, config)
watch	Guard any training command
start	Dashboard + MCP one-click launch
serve	Standalone MCP server
remote	Start remote communication server (compute server side)
contract check	Validate training script contract
preflight	GPU memory estimate + batch recommendation
analyze	Scan existing checkpoints
analyze_architecture	Analyze model architecture (D3 treemap/backbone)
experiments	List all historical experiments
query	Natural language query ("best lr?")
compare	Compare two experiments
visualize	Model structure visualization (D3.js HTML)
infer	Run inference with checkpoint
gallery	Image filtering + selection
dashboard	Web control panel (standalone)
project	Project context management (init/show/scan/fill)

MCP Tools · MCP 工具

25 read-only (always available, no auth):

get_training_status · get_metrics_history · list_checkpoints · compare_checkpoints · get_anomaly_history · get_recovery_history · get_summary · get_agent_decision_log · get_contract_status · list_contract_proposals · list_experiments · query_experiment · compare_experiments · get_model_structure · analyze_architecture · get_guardian_mode · get_gallery_config · get_import_format · inspect_source · get_training_log · get_post_training_checklist · get_pending_decisions · get_dashboard_config · recommend_charts · list_dashboard_templates

11 write (token auth + training-phase gating):

trigger_recovery · restart_with_params · stop_training · approve_contract_proposal · reject_contract_proposal · run_visualization · set_gallery_config · run_inference · submit_import · resolve_decision · set_dashboard_config

→ Full API reference: docs/MCP_API_REFERENCE.md

Configuration · 配置

Three layers, zero secrets in YAML:

DEFAULTS  <  guardian.yaml  <  GUARDIAN_* env vars  <  CLI flags

# configs/guardian.yaml — only override what you need
watchdog:
  max_retries: 3
monitor:
  poll_interval: 5
mcp:
  enable_write_tools: true

# Env override: GUARDIAN_ + section + __ + key
export GUARDIAN_WATCHDOG__MAX_RETRIES=5
export GUARDIAN_MCP_TOKEN=your-secret   # write tool auth

DSH Web GUI Plugin · DSH 插件

配套 DSH Web GUI 插件 @rrrelink/dsh-client-ui-training-guardian，在 DSH 侧栏提供六标签页的 Training Guardian 面板（概览/设备/异常/决策/架构/历史），通过 SSE + REST 消费 guarftrain remote 服务。

# 安装插件（profile 目录 ~/.dsh/profiles/web）
dsh plugin add @rrrelink/dsh-client-ui-training-guardian --profile web

# 训练机侧启动数据源
guarftrain remote --port 8765
guarftrain watch -- python train.py --epochs 50

• 源码：dsh-plugin/dsh-client-ui-training-guardian
• 完整使用说明书：README.zh.md / README.md
• 插件镜像仓库：https://github.com/Washington5533/Guid-traince

Project Status · 项目状态

Metric	Value
Version	0.3.0
Modules	21 (cp_1 ~ cp_21)
Production code	~12,500 lines
Tests	266 (CI on push)
MCP tools	36 (25 read + 11 write)
CLI commands	18
Test coverage	~13% (core paths: 100%)
Python	3.10+

Docs · 文档索引

Document	Content
docs/ARCHITECTURE.md	Architecture & workflow (ZH)
docs/DEPLOYMENT.md	User manual (ZH)
docs/MCP.md	MCP integration guide (ZH)
docs/MCP_API_REFERENCE.md	36-tool API reference (ZH)
docs/MCP_QUICKSTART.md	5-minute MCP onboarding (ZH)
docs/IMPLEMENTATION_REPORT.md	Per-module completion report (ZH)
dsh-plugin/…/README.zh.md	DSH plugin user manual (ZH/EN)

License

MIT