Verified & Tested

Verified Examples

Every example is built using only the TLL OS Public Contract. No internal imports. No hidden dependencies. Clone, run, and learn.

🧪

hello-tll-agent — Minimal PoC

The first proof that an AI Agent can build a TLL OS application using only the public contract. Single module, complete test-fix-retest loop.

3/3 Tests 1 Module 1 API 1 Tool 1 Agent
run
npx tsx examples/hello-tll-agent/agent.ts
→ 测试结果: 2/3 通过, 1 失败 (intentional bug)
→ 修复后重新运行: 3/3 通过, 0 失败
→ ✅ 全部测试通过！修复成功！
→ 结论: TLL OS 的 AI-Native 核心设计成立。

What it proves: An Agent can create an application, write tests, observe a failure, diagnose the bug, fix the code, and re-run until all tests pass — all through the public contract.

🤖

autonomous-task-manager — Real Agent Development Experiment

The critical validation. An Agent that has never seen TLL OS source code independently builds a multi-module application with cross-module dependencies, 11 REST APIs, a unified Tool, and an intent-parsing Agent. 12 tests, all passing.

12/12 Tests 2 Modules 11 APIs 1 Tool (6 ops) 1 Agent Cross-module deps
run
npx tsx examples/autonomous-task-manager/agent.ts
→ 结果: 12/12 通过, 0 失败, 0 错误 (6ms)
→ ✅ PASS: project.create_and_get
→ ✅ PASS: project.list_filter_update_delete
→ ✅ PASS: task.create_requires_project
→ ✅ PASS: task.crud_lifecycle
→ ✅ PASS: task.list_filters
→ ✅ PASS: api.project_crud
→ ✅ PASS: api.task_validates_project
→ ✅ PASS: api.tasks_by_project
→ ✅ PASS: tool.manage_task_crud
→ ✅ PASS: tool.validates_params
→ ✅ PASS: agent.lists_tasks
→ ✅ PASS: agent.creates_task

What it proves:

  • Public Contract is sufficient for real multi-module application development
  • Agent can autonomously design module architecture and cross-module dependencies
  • Agent can implement complete CRUD + cross-module validation
  • Agent can design a unified Tool with multiple operations
  • Agent can design an Agent with intent parsing
  • Application Graph correctly reflects dual-module, API, Tool, Agent, and dependencies
  • Real Agent experiments discover core bugs that unit tests miss — this experiment found and fixed the path parameter parsing bug

Application Graph of task-manager

What the Agent built, visualized as a Graph.

graph.json (simplified)
{
  "nodes": [
    { "type": "application", "name": "task-manager-app" },
    { "type": "module", "name": "project" },
    { "type": "module", "name": "task" },
    { "type": "api", "name": "GET /api/projects" },
    { "type": "api", "name": "GET /api/projects/:id" },
    { "type": "api", "name": "POST /api/projects" },
    { "type": "api", "name": "PUT /api/projects/:id" },
    { "type": "api", "name": "DELETE /api/projects/:id" },
    { "type": "api", "name": "GET /api/tasks" },
    { "type": "api", "name": "GET /api/tasks/:id" },
    { "type": "api", "name": "POST /api/tasks" },
    { "type": "api", "name": "PUT /api/tasks/:id" },
    { "type": "api", "name": "DELETE /api/tasks/:id" },
    { "type": "api", "name": "GET /api/projects/:id/tasks" },
    { "type": "tool", "name": "manage_task" },
    { "type": "agent", "name": "task_manager_agent" }
  ],
  "edges": [
    { "from": "module:project", "type": "belongs_to", "to": "application:task-manager-app" },
    { "from": "module:task", "type": "belongs_to", "to": "application:task-manager-app" },
    { "from": "module:task", "type": "depends_on", "to": "module:project" },
    { "from": "api:GET /api/projects", "type": "belongs_to", "to": "module:project" },
    { "from": "tool:manage_task", "type": "belongs_to", "to": "module:task" },
    { "from": "agent:task_manager_agent", "type": "calls", "to": "tool:manage_task" }
  ]
}

Build Your Own Example

Clone the repo, study the examples, then build something amazing.

Clone from GitHub Read the Guide