Skip to main content

A powerful framework for managing long-running AI Agent tasks

Project description

LRA - AI Agent Task Manager v3.2

通用 AI Agent 任务管理框架

核心特性

  • 通用任务模型:支持软件开发、小说写作、数据处理等多种场景
  • Jinja2 模板:强大的模板引擎,支持条件/循环语法
  • 任务依赖:DAG 依赖关系,自动解锁完成的任务
  • 优先级调度:P0-P3 优先级,Agent 自评
  • 多 Agent 协作:层级锁机制,支持大模型拆分任务、小模型并行开发
  • 输出限制感知:根据模型输出能力推荐/拆分任务

安装

# 基础安装
pip install long-run-agent

# 完整安装(包含 Jinja2 模板引擎)
pip install long-run-agent[full]

快速开始

# 初始化项目
cd /your/project
lra init --name "My Project"

# Agent 获取上下文
lra context --output-limit 8k

命令参考

核心命令

命令 用途
lra context [--output-limit Xk] 获取项目状态 + 可领取任务
lra list [--status X] [--template X] 列出任务
lra create <desc> --template <name> 创建任务
lra show <id> 任务详情
lra set <id> <status> 更新状态(受模板约束)
lra split <id> --plan '<json>' 拆分任务(模型提供方案)

锁命令

命令 用途
lra claim <id> 领取任务(锁定自己+子任务)
lra publish <id> 发布子任务(释放子任务锁)
lra pause <id> 暂停并保存快照
lra resume <id> 查看快照
lra heartbeat <id> 心跳保活(每5分钟)

模板命令

命令 用途
lra template list 列出模板
lra template show <name> 查看模板详情
lra template create <name> 创建模板

依赖命令

命令 用途
lra deps <id> 查看任务依赖
lra deps <id> --dependents 查看依赖此任务的其他任务
lra check-blocked 检查并解锁 blocked 任务

优先级命令

命令 用途
lra set-priority <id> <P0|P1|P2|P3> 设置任务优先级

内置模板

模板 用途 状态
task 通用任务 pending → in_progress → completed
novel-chapter 小说章节 drafting → revising → finalized
code-module 代码模块 pending → in_progress → pending_test → completed
data-pipeline 数据流程 pending → running → success

多 Agent 协作流程

1. 大模型 claim task_001(整个模块)
2. 大模型编写架构/接口契约
3. 大模型 split task_001 --plan '[...]'
4. 大模型 publish task_001(释放子任务锁)
5. 小模型 context --output-limit 8k(获取可领取任务)
6. 小模型 claim task_001_01(领取子任务)
7. 小模型按契约开发
8. 大模型验收/集成

输出限制适配

模型 输出限制 使用
GPT-4o-mini 4K --output-limit 4k
Claude 3.5 8K --output-limit 8k
Claude 3.5 Sonnet 16K --output-limit 16k
Claude 3.5 Sonnet Max 128K --output-limit 128k

数据结构

.long-run-agent/
├── config.json          # 项目配置
├── task_list.json       # 任务列表
├── locks.json           # 任务锁
├── templates/           # 模板(可自定义)
│   ├── task.yaml
│   ├── novel-chapter.yaml
│   ├── code-module.yaml
│   └── data-pipeline.yaml
├── tasks/               # 任务文件
│   └── task_001.md
└── records/             # 变更记录
    └── task_001.json

自定义模板

# .long-run-agent/templates/my-template.yaml
name: my-template
description: 我的自定义模板
version: "2.0"
template_engine: jinja2  # 使用 Jinja2 引擎

structure: |
  # {{ id }}
  
  ## 描述
  {{ description }}
  
  {% if tech_stack %}
  ## 技术栈
  {{ tech_stack }}
  {% endif %}
  
  ## 交付物
  <!-- 请列出交付物 -->

states:
  - pending
  - working
  - done

transitions:
  pending: [working]
  working: [done]
  done: []

acceptance:
  - 验收标准 1
  - 验收标准 2

创建任务示例

# 基础创建
lra create "实现用户登录"

# 带优先级
lra create "紧急 Bug 修复" --priority P0

# 带依赖
lra create "集成测试" --dependencies task_001,task_002 --dependency-type all

# 带截止时间
lra create "发布版本" --deadline "2026-02-28T23:59:59"

# 带模板变量
lra create "API 开发" --template code-module \
  --variables '{"tech_stack": "FastAPI", "input_params": "user_id"}'

状态说明

状态 说明
blocked 依赖未完成,不可领取
pending 初始状态,可领取
in_progress 进行中
completed 完成(终态)

blocked 状态自动解锁:当依赖的任务完成后,blocked 任务会自动变为 pending 状态。

环境要求

  • Python ≥ 3.8
  • Git ≥ 2.0(可选)

链接

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

long_run_agent-3.2.0.tar.gz (23.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

long_run_agent-3.2.0-py3-none-any.whl (24.3 kB view details)

Uploaded Python 3

File details

Details for the file long_run_agent-3.2.0.tar.gz.

File metadata

  • Download URL: long_run_agent-3.2.0.tar.gz
  • Upload date:
  • Size: 23.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for long_run_agent-3.2.0.tar.gz
Algorithm Hash digest
SHA256 d0523e98abb4310df1388ec6673e31e0908733f1e5ef2a0db0c3fe479dd06830
MD5 e2ae7c2e35b6efaed59299d03949ce2e
BLAKE2b-256 25fe87fc5a273d65da18d8cc3a1d663f1ef9b10173dba1ac395689ad86b74168

See more details on using hashes here.

Provenance

The following attestation bundles were made for long_run_agent-3.2.0.tar.gz:

Publisher: publish.yml on hotjp/long-run-agent

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file long_run_agent-3.2.0-py3-none-any.whl.

File metadata

  • Download URL: long_run_agent-3.2.0-py3-none-any.whl
  • Upload date:
  • Size: 24.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for long_run_agent-3.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 5ad9c697de950338829e07f8c720314f421ce3644803471a8d7d2ab94f946b7e
MD5 63f2f921529c0b87fdde74e9945a40b0
BLAKE2b-256 be120933c0b0c686962febc45993ee6bf5c89778f8ebe06bd4119cfd13fda1d0

See more details on using hashes here.

Provenance

The following attestation bundles were made for long_run_agent-3.2.0-py3-none-any.whl:

Publisher: publish.yml on hotjp/long-run-agent

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page