每日开源速递 - 2026-04-16

1
2
3
4
5
6
7
8
9

# 方式一：通过 npx 安装
npx claude-mem install

# 方式二：通过插件市场安装
/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

# OpenClaw 网关安装
curl -fsSL https://install.cmem.ai/openclaw.sh | bash

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22

# 克隆并安装依赖
git clone https://github.com/lsdefine/GenericAgent.git
cd GenericAgent
pip install streamlit pywebview

# 配置 API Key
cp mykey_template.py mykey.py
# 编辑 mykey.py 填入 LLM API Key

# 启动（支持多种前端）
python launch.pyw           # 默认 Streamlit UI
python frontends/qtapp.py   # Qt 桌面应用
streamlit run frontends/stapp2.py  # 替代 Streamlit UI

# 微信机器人前端
pip install pycryptodome qrcode requests
python frontends/wechatapp.py

# QQ 机器人前端
pip install qq-botpy
# 配置 qq_app_id、qq_app_secret 后
python frontends/qqapp.py

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# macOS（Apple Silicon）
# 下载地址: https://voicebox.sh/download/mac-arm

# Windows
# 下载地址: https://voicebox.sh/download/windows

# Docker 部署
docker compose up

# REST API 调用示例
curl -X POST http://localhost:17493/generate \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello world", "profile_id": "abc123", "language": "en"}'

# 列出语音配置
curl http://localhost:17493/profiles

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

# 本地开发
bun install
cp apps/web/.env.example apps/web/.env
# 填写必要环境变量
bun run web

# Vercel 部署步骤
# 1. Fork 此仓库
# 2. 创建 PostgreSQL 数据库
# 3. 生成密钥
openssl rand -base64 32 | tr '+/' '-_' | tr -d '=\n'  # JWE_SECRET
openssl rand -hex 32  # ENCRYPTION_KEY
# 4. 在 Vercel 中导入仓库并配置环境变量

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

# 多种安装方式
pipx install magika
brew install magika  # macOS/Linux
curl -LsSf https://securityresearch.google/magika/install.sh | sh
powershell -ExecutionPolicy Bypass -c "irm https://securityresearch.google/magika/install.ps1 | iex"
cargo install --locked magika-cli

# Python API
from magika import Magika
m = Magika()
res = m.identify_bytes(b'function log(msg) {console.log(msg);}')
print(res.output.label)  # javascript

res = m.identify_path('./tests_data/basic/ini/doc.ini')
print(res.output.label)  # ini

# CLI 使用
magika -r *  # 递归扫描目录
magika --json file.py  # JSON 输出格式
magika --mime-type file  # MIME 类型输出

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

import cognee
import asyncio

async def main():
    # 永久存储到知识图谱
    await cognee.remember("Cognee turns documents into AI memory.")
    
    # 会话内存（快速缓存，后台同步到图谱）
    await cognee.remember("User prefers detailed explanations.", session_id="chat_1")
    
    # 自动路由查询
    results = await cognee.recall("What does Cognee do?")
    
    # 指定会话查询
    results = await cognee.recall("What does the user prefer?", session_id="chat_1")
    
    # 删除数据
    await cognee.forget(dataset="main_dataset")

asyncio.run(main())

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22

# 安装
uv pip install cognee

# 配置 LLM
export LLM_API_KEY="YOUR_OPENAI_API_KEY"

# CLI 使用
cognee-cli remember "Cognee turns documents into AI memory."
cognee-cli recall "What does Cognee do?"
cognee-cli -ui  # 打开本地 UI

# Claude Code 插件
git clone https://github.com/topoteretes/cognee-integrations.git
claude --plugin-dir ./cognee-integrations/integrations/claude-code

# Hermes Agent 集成
# ~/.hermes/config.yaml
memory:
  provider: cognee

# 云端托管
await cognee.serve(url="https://your-instance.cognee.ai", api_key="ck_...")

1
2
3
4
5

# 项目地址
https://github.com/Lordog/dive-into-llms

# 昇腾社区教程
# https://www.hiascend.com/edu/growth/lm-development

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

# 基础 Agent 使用
from agents import Agent, Runner

agent = Agent(
    name="Assistant",
    instructions="You are a helpful assistant."
)

result = Runner.run_sync(agent, "Write a Python function to calculate fibonacci.")
print(result.final_output)

# 沙箱 Agent（可执行文件操作）
from agents import Runner, RunConfig
from agents.sandbox import SandboxAgent, SandboxRunConfig, Manifest, GitRepo
from agents.sandbox.sandboxes import UnixLocalSandboxClient

sandbox_agent = SandboxAgent(
    name="Workspace Assistant",
    instructions="Inspect the sandbox workspace before answering.",
    default_manifest=Manifest(
        entries={"repo": GitRepo(repo="openai/openai-agents-python", ref="main")}
    ),
)

result = Runner.run_sync(
    sandbox_agent,
    "Inspect the repo README and summarize what this project does.",
    run_config=RunConfig(sandbox=SandboxRunConfig(client=UnixLocalSandboxClient())),
)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

# 安装（Python 3.10+）
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install openai-agents

# 或使用 uv
uv init
uv add openai-agents

# 语音支持
pip install 'openai-agents[voice]'

# Redis 会话支持
pip install 'openai-agents[redis]'

1
2
3
4
5
6
7
8
9

# 启动方式
node index.js                    # 单次演化
node index.js --review           # 人工确认模式
node index.js --loop             # 持续后台运行

# 策略控制
EVOLVE_STRATEGY=innovate node index.js --loop   # 最大化新功能
EVOLVE_STRATEGY=harden node index.js --loop     # 专注稳定性
EVOLVE_STRATEGY=repair-only node index.js --loop # 紧急修复模式

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19

# 安装
git clone https://github.com/EvoMap/evolver.git
cd evolver
npm install

# 可选：连接 EvoMap Hub
# 注册 https://evomap.ai 获取 Node ID
# 创建 .env 文件
A2A_HUB_URL=https://evomap.ai
A2A_NODE_ID=your_node_id_here

# 下载共享技能
node index.js fetch --skill <skill_id>

# 生命周期管理
node src/ops/lifecycle.js start   # 后台启动
node src/ops/lifecycle.js stop    # 优雅停止
node src/ops/lifecycle.js status  # 运行状态
node src/ops/lifecycle.js check   # 健康检查

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# 通过插件市场安装
/plugin marketplace add SimoneAvogadro/android-reverse-engineering-skill
/plugin install android-reverse-engineering@android-reverse-engineering-skill

# 或从本地路径安装
git clone https://github.com/SimoneAvogadro/android-reverse-engineering-skill.git
/plugin marketplace add /path/to/android-reverse-engineering-skill
/plugin install android-reverse-engineering@android-reverse-engineering-skill

# 使用
/decompile path/to/app.apk

# 独立脚本使用
bash plugins/android-reverse-engineering/skills/android-reverse-engineering/scripts/check-deps.sh
bash plugins/android-reverse-engineering/skills/android-reverse-engineering/scripts/decompile.sh app.apk
bash plugins/android-reverse-engineering/skills/android-reverse-engineering/scripts/find-api-calls.sh output/sources/ --retrofit

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# 最简方式（连接云端后端）
git clone https://github.com/BasedHardware/omi.git && cd omi/desktop && ./run.sh --yolo

# 本地完整开发
xcode-select --install
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
git clone https://github.com/BasedHardware/omi.git
cd omi/desktop
cp Backend-Rust/.env.example Backend-Rust/.env
./run.sh

# 移动端开发
cd app && bash setup.sh ios  # 或 android

# Python SDK
pip install omi-sdk

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

# 安装（支持多种后端）
uv pip install -e ".[transformers]"   # Transformers
uv pip install -e ".[sglang]"         # SGLang
uv pip install -e ".[mlx]"            # MLX (Apple Silicon)
uv pip install -U vllm --torch-backend=auto --extra-index-url https://wheels.vllm.ai/nightly  # vLLM

# vLLM 服务
vllm serve Qwen/Qwen3.5-27B \
  --speculative-config '{"method": "dflash", "model": "z-lab/Qwen3.5-27B-DFlash", "num_speculative_tokens": 15}' \
  --attention-backend flash_attn \
  --max-num-batched-tokens 32768

# SGLang 服务
python -m sglang.launch_server \
  --model-path Qwen/Qwen3.5-35B-A3B \
  --speculative-algorithm DFLASH \
  --speculative-draft-model-path z-lab/Qwen3.5-35B-A3B-DFlash \
  --speculative-num-draft-tokens 16

# Transformers API（Qwen3 和 Llama）
from transformers import AutoModel, AutoModelForCausalLM, AutoTokenizer

draft = AutoModel.from_pretrained("z-lab/Qwen3-8B-DFlash-b16", trust_remote_code=True, dtype="auto", device_map="cuda:0").eval()
target = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B", dtype="auto", device_map="cuda:0").eval()
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-8B")

output = draft.spec_generate(input_ids=input_ids, max_new_tokens=2048, temperature=0.0, target=target, stop_token_ids=[tokenizer.eos_token_id])

1
2
3
4
5
6
7
8
9

# vLLM 基准
python -m dflash.benchmark --backend vllm \
  --base-url http://127.0.0.1:8000 --model Qwen/Qwen3.5-27B \
  --dataset gsm8k --num-prompts 128 --concurrency 1 --enable-thinking

# MLX 基准（Apple Silicon）
python -m dflash.benchmark --backend mlx \
  --model Qwen/Qwen3.5-4B --draft-model z-lab/Qwen3.5-4B-DFlash \
  --dataset gsm8k --max-samples 128 --enable-thinking