能力标签

⚙️

Agent工作流

M31A智能工作流

基于 Go · 无代码搭建完整 AI 自动化流程

英文名：M31A

⭐ 8 Stars 💻 Go 📄 MIT 🏷 AI 8.0分

8.0AI 综合评分

aiai-assistantgo

⬇ 下载源码 ZIP ⚙️ 配置说明 📺 TG 频道

✦ AI Skill Hub 推荐

经 AI Skill Hub 精选评估，M31A智能工作流获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色，AI 评分 8.0 分，适合有一定技术背景的用户使用。

📚 深度解析

M31A智能工作流是一套完整的 AI Agent 自动化工作流方案。随着 AI 能力的不断提升，基于 Agent 的自动化工作流正在成为提升个人和团队效率的核心方式。区别于传统的 RPA 自动化（模拟鼠标键盘操作），AI Agent 工作流通过理解任务意图、动态规划执行路径，能够处理更复杂的非结构化任务。

M31A智能工作流工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。

在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 8.0 分，是同类 Agent 工作流中的精选推荐。

📋 工具概览

M31A智能工作流是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。

GitHub Stars

⭐ 8

开发语言

支持平台

Windows / macOS / Linux（跨平台）

维护状态

轻量级项目，按需更新

开源协议

MIT

AI 综合评分

8.0 分

工具类型

Agent工作流

Forks

—

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

📌 核心特色

可视化 Agent 工作流编排，无需编写复杂代码
支持多步骤自动化任务链，实现全流程无人值守
与外部 API、数据库和第三方服务无缝集成
内置错误处理与自动重试机制，保障稳定运行
提供可复用的自动化模板，快速在同类场景部署

🎯 主要使用场景

自动化日常重复性工作，将精力集中于创造性任务
构建数据采集 → 处理 → 输出的完整自动化管线
实现跨平台、跨系统的数据流转和业务协同

以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。

安装命令

# 方式一：go install（推荐）
go install github.com/eshanized/M31A@latest

# 方式二：从源码编译
git clone https://github.com/eshanized/M31A
cd M31A
go build -o m31a .

# 方式三：下载预编译二进制
# 访问 Releases 页面下载对应平台二进制文件
# https://github.com/eshanized/M31A/releases

📋 安装步骤说明

访问 GitHub 仓库获取工作流文件
在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
上传工作流文件
按照提示配置必要的环境变量和 API Key
运行测试确认流程正常后投入使用

以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。

常用命令 / 代码示例

# 查看帮助
m31a --help

# 基本运行
m31a [options] <input>

# 详细使用说明请查阅文档
# https://github.com/eshanized/M31A

以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。

配置示例

# m31a 配置说明
# 查看配置选项
m31a --config-example > config.yml

# 常见配置项
# output_dir: ./output
# log_level: info
# workers: 4

# 环境变量（覆盖配置文件）
export M31A_CONFIG="/path/to/config.yml"

📑 README 深度解析真实文档完整度 57/100 含工作流图查看 GitHub 原文 →

以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

简介

Features

Install

Pick your weapon.

```bash

Screenshots

Configuration

~/.m31a/config.toml (override with M31A_CONFIG):

```toml [provider] default = "openrouter" auto_fallback = true

Workflow & Intelligence

Seven-phase workflow — Initialize → Discuss → Plan → Execute → Verify → Runtime → Ship with four modes (auto, full, fast, direct) to match task complexity. Every run ends with a verified git commit and a ledger entry.
Workflow quality gates — Plan checker with revision loops, coverage gates, security heuristics, and gap analysis. Discuss-phase quality scoring and completeness checks. Execute-phase pre-flight validation and tool-call loop detection.
Code intelligence — Parses Go (via tree-sitter), TypeScript, Python, and Rust. Builds import dependency graphs, indexes symbols, scores file relevance. The LLM gets context about which files matter for the current task.
Code complexity analysis — CodeComplexity tool classifies your codebase as simple (<10K lines), moderate (10K–50K), or complex (50K+) across all 4 languages, informing model selection.
18 built-in tools — Bash, FileRead, FileWrite, Edit, Glob, Grep, WebFetch, WebSearch, CodeMap, CodeComplexity, FileDelete, FileMove, FileList, TodoWrite, TodoRead, DevServer, HTTPCheck, AskUserQuestion. All gated by a permission system with rate limiting and concurrency control.
Parallel subagents — The LLM can spawn child agents (up to depth 2) that run in isolated git worktrees, each with their own dispatcher and permissions.
Task runner — Kahn's algorithm for topological sort, bounded parallelism (4 concurrent), per-task timeouts, retry support.

Workflow quality enhancements (all opt-in):

plan_research = true # pre-plan research step plan_check = true # plan checker + revision loop plan_security_gate = true # security heuristic gate plan_coverage_gate = true # requirements coverage gate plan_chunked = true # chunked plan generation discuss_quality_check = true # question quality checker discuss_completeness = true # answer completeness check execute_preflight = true # pre-execution validation execute_loop_detect = true # tool call loop detection verify_report = true # verification report generation ship_preflight = true # pre-ship checklist ship_changelog = true # changelog generation init_deep_analysis = true # deep project analysis

[tools] max_glob_results = 50 max_grep_results = 50 bash_kill_grace_secs = 5 websearch_enabled = true

[git] commit_prefix = "feat" fix_prefix = "fix" ship_prefix = "ship"

[verify] build_command = "" # empty = auto-detect test_command = "" # empty = auto-detect

[agents] default = "" plan = "" # cheap model for planning execute = "" # powerful model for execution verify = "" ship = "" discuss = "" ```

Full reference: docs/CONFIG.md

Package Summary

Package	Description
`cmd/m31a/`	Binary entry point with CLI flags
`internal/tui/`	Bubble Tea TUI app — 33 screens, 11 themes, responsive layout, command palette
`internal/workflow/`	Seven-phase orchestration engine with quality gates, chunked plans, and pre-flight checks
`internal/provider/`	OpenRouter, Zen, and Nvidia clients with auto-fallback, model cache, and SSE streaming
`internal/tools/`	18 tools with permission system, token-bucket rate limiting, and concurrency control
`internal/tools/subagent/`	Parallel subagent manager with git worktree isolation (max depth 2)
`internal/codeintel/`	4-language parser (Go, TypeScript, Python, Rust) with import graph and relevance scoring
`internal/config/`	TOML loader with project context detection, hot-reload, and workflow enhancement flags
`internal/git/`	Commit, rollback, diff, stash, and branch operations
`internal/tokens/`	tiktoken-based estimation with EMA self-calibration
`internal/types/`	Shared types, constants, workflow modes, risk levels, and plan structures
`internal/fileutil/`	Atomic file write operations
`internal/errors/`	Sentinel errors with user-friendly messages
`internal/log/`	Structured logging with daily rotation
`pkg/session/`	Session lifecycle and persistence (JSON in `~/.m31a/sessions/`)
`pkg/ledger/`	Cross-session learning store (markdown-backed)
`pkg/rollback/`	Commit-chain manager (soft/hard/safe reset)
`pkg/bisect/`	Git-bisect wrapper for model comparison
`pkg/taskrunner/`	Parallel task executor with Kahn's algorithm and bounded parallelism
`pkg/keychain/`	OS keychain abstraction (Linux dbus, macOS Keychain, Windows Credential Manager)
`pkg/autodream/`	Context consolidation with reentrancy guard
`pkg/arbitrage/`	Model-cost optimizer with task classification
`pkg/history/`	Frecent prompt history with scoring