Views
No views yet
1# 安装
2pip install -e ".[dev]"
3
4# 运行内置 Demo
5python -m su_memory_agent_poc.cli demo market_research
6
7# 一键运行全部 Demo
8bash scripts/demo.sh
9
10# 运行基准测试
11python -m su_memory_agent_poc.benchmark.cli run gaia
12bash scripts/benchmark.sh superclue| 场景 | 命令 | 演示能力 |
|---|---|---|
| market_research | demo market_research | WebSearch + 报告生成 |
| product_compare | demo product_compare | WebSearch + 对比分析 |
| code_verify | demo code_verify | CodeExecutor + 验证 |
| news_summary | demo news_summary | WebSearch + FileSystem |
| medical_diagnosis | demo medical_diagnosis | WebBrowser + 诊断推理 |
1# GAIA 基准测试
2python -m su_memory_agent_poc.benchmark.cli run gaia --verbose
3
4# SuperCLUE 基准测试
5python -m su_memory_agent_poc.benchmark.cli run superclue --verbosesu-memory-poc --help
su-memory-poc run "任务描述" --model auto --verbose
su-memory-poc demo <scenario>1docker build -t su-memory-poc .
2docker run --rm su-memory-poc demo market_research
3docker-compose run demo market_research
4docker-compose run benchmark --suite gaia_samplesrc/su_memory_agent_poc/
├── cli.py # CLI 入口
├── loop.py # AgentLoop 核心
├── tools/ # 工具集 (echo/search/code/file/mcp)
├── verify/ # 验证层 (ConfidenceGate/SafetyGate/Rollback)
├── demos/ # 5 个内置 Demo 场景
├── benchmark/ # 基准测试运行器 + GAIA/SuperCLUE 适配器
└── types.py # 共享类型定义1pip install -e ".[dev]"
2ruff check src/ tests/
3pytest -v