安装与快速开始

系统要求

最低系统要求

操作系统：Linux (Ubuntu 20.04+), macOS (10.15+), Windows 10+
Python 版本：Python 3.8 或更高版本
内存：最少 4GB RAM（推荐 8GB+）
存储空间：至少 10GB 可用空间
网络：稳定的互联网连接

安装方法

方法 1：通过 pip 安装（推荐）

# 安装 LLMESH 核心包
pip install llmesh-network

# 验证安装
llmesh --version

方法 2：从源码安装

# 克隆仓库
git clone https://github.com/llmesh-cor/llmesh.git
cd llmesh

# 创建虚拟环境
python -m venv llmesh-env
source llmesh-env/bin/activate  # Linux/macOS
# 或在 Windows 上
# llmesh-env\Scripts\activate

# 安装依赖
pip install -r requirements.txt

# 安装 LLMESH
pip install -e .

方法 3：使用 Docker

# 拉取官方镜像
docker pull llmesh/llmesh-node:latest

# 运行容器
docker run -d \
  --name llmesh-node \
  -p 8080:8080 \
  -p 9000:9000 \
  -v ~/.llmesh:/root/.llmesh \
  llmesh/llmesh-node:latest

配置环境

初始化配置

# 初始化 LLMESH 配置
llmesh init

# 这将创建默认配置文件在 ~/.llmesh/config.yaml

配置文件示例

# ~/.llmesh/config.yaml
node:
  id: "node_${RANDOM_ID}"
  name: "My LLMESH Node"
  description: "Personal LLMESH node for AI services"
  
network:
  listen_address: "0.0.0.0"
  listen_port: 9000
  bootstrap_nodes:
    - "bootstrap1.llmesh.network:9000"
    - "bootstrap2.llmesh.network:9000"
  max_peers: 50
  
models:
  storage_path: "~/.llmesh/models"
  cache_size: "5GB"
  supported_formats: ["onnx", "pytorch", "tensorflow"]
  
tokens:
  mesh_token_address: "0x1234567890123456789012345678901234567890"
  wallet_private_key: "${MESH_PRIVATE_KEY}"
  
logging:
  level: "INFO"
  file: "~/.llmesh/logs/llmesh.log"

快速开始

1. 启动第一个节点

# 启动节点（需要质押代币）
llmesh-node start --stake 1000

# 或者以测试模式启动（无需质押）
llmesh-node start --testnet

启动成功后，你将看到类似输出：

🚀 LLMESH Node starting...
📡 Node ID: node_a1b2c3d4e5f6
🌐 Listening on: 0.0.0.0:9000
💰 Staked tokens: 1000 MESH
🔗 Connected to 12 peers
✅ Node is ready!

2. 部署你的第一个 LLM 模型

#!/usr/bin/env python3
"""部署模型示例"""

from llmesh import ModelDeployer, ModelConfig

def deploy_model():
    # 创建模型配置
    config = ModelConfig(
        name="my-gpt-model",
        model_path="./models/gpt-model.onnx",
        model_type="text-generation",
        description="Personal GPT model for text generation",
        fee_per_request=0.1,  # 每次请求收费 0.1 MESH
        max_context_length=2048,
        temperature=0.7
    )
    
    # 创建部署器
    deployer = ModelDeployer()
    
    # 部署模型
    try:
        deployment = deployer.deploy(config)
        print(f"✅ Model deployed successfully!")
        print(f"📝 Model ID: {deployment.model_id}")
        print(f"🔗 Access URL: {deployment.access_url}")
        print(f"💰 Fee per request: {config.fee_per_request} MESH")
        
        return deployment
        
    except Exception as e:
        print(f"❌ Deployment failed: {e}")
        return None

if __name__ == "__main__":
    deploy_model()

运行部署脚本：

python deploy_model.py

3. 使用命令行工具部署模型

# 部署 ONNX 模型
llmesh-deploy ./models/my-model.onnx \
  --name "my-llm" \
  --type "text-generation" \
  --fee 0.1 \
  --description "My personal LLM model"

# 查看已部署的模型
llmesh models list

# 输出示例：
# Model ID        Name       Type              Fee    Status
# abc123def456    my-llm     text-generation   0.1    Active
# def789ghi012    chat-bot   conversational    0.05   Active

4. 调用网络中的模型

#!/usr/bin/env python3
"""调用模型示例"""

from llmesh import LLMeshClient

async def main():
    # 创建客户端
    client = LLMeshClient()
    
    # 搜索可用模型
    models = await client.search_models(
        model_type="text-generation",
        max_fee=0.2
    )
    
    print(f"Found {len(models)} available models:")
    for model in models:
        print(f"  - {model.name} (Fee: {model.fee} MESH)")
    
    if models:
        # 选择第一个模型进行推理
        model = models[0]
        
        # 发送推理请求
        response = await client.generate_text(
            model_id=model.id,
            prompt="What is the future of artificial intelligence?",
            max_tokens=200,
            temperature=0.7
        )
        
        print(f"\n🤖 Model Response:")
        print(f"📝 Text: {response.text}")
        print(f"💰 Cost: {response.cost} MESH")
        print(f"⏱️ Response time: {response.response_time}ms")

if __name__ == "__main__":
    import asyncio
    asyncio.run(main())

5. Web 界面使用

启动节点后，你可以通过浏览器访问 Web 界面：

# 启动带 Web 界面的节点
llmesh-node start --web-ui --port 8080

然后在浏览器中打开 http://localhost:8080，你将看到：

仪表板：节点状态、网络统计、收益概览
模型管理：部署、配置、监控你的模型
市场：浏览和使用网络中的其他模型
钱包：管理 MESH 代币和交易记录

高级配置

GPU 加速配置

# config.yaml
compute:
  use_gpu: true
  gpu_devices: [0, 1]  # 使用 GPU 0 和 1
  gpu_memory_limit: "8GB"
  batch_size: 8
  
models:
  inference_backend: "cuda"  # 或 "cpu", "opencl"
  mixed_precision: true
  optimization_level: "O2"

集群部署配置

#!/usr/bin/env python3
"""集群部署脚本"""

from llmesh import ClusterManager, NodeConfig

def setup_cluster():
    cluster = ClusterManager()
    
    # 配置主节点
    master_config = NodeConfig(
        role="master",
        compute_resources={"cpu": 16, "memory": "32GB", "gpu": 2},
        services=["routing", "model_hosting", "load_balancing"]
    )
    
    # 配置工作节点
    worker_configs = []
    for i in range(3):
        worker_config = NodeConfig(
            role="worker",
            compute_resources={"cpu": 8, "memory": "16GB", "gpu": 1},
            services=["model_hosting"]
        )
        worker_configs.append(worker_config)
    
    # 部署集群
    cluster.deploy_master(master_config)
    for worker_config in worker_configs:
        cluster.add_worker(worker_config)
    
    print("✅ Cluster deployed successfully!")
    print(f"📊 Master node: {cluster.master_node.id}")
    print(f"👥 Worker nodes: {len(cluster.worker_nodes)}")

if __name__ == "__main__":
    setup_cluster()

常见问题解决

连接问题

# 检查节点状态
llmesh-node status

# 检查网络连接
llmesh-node ping-peers

# 重置网络连接
llmesh-node reset-network

模型加载问题

# 检查模型格式
from llmesh.utils import ModelValidator

validator = ModelValidator()
is_valid = validator.validate_model("./my-model.onnx")

if not is_valid:
    print("❌ Model format is not supported")
    print("Supported formats:", validator.supported_formats)

性能调优

# 性能优化配置
performance:
  worker_threads: 8
  async_io: true
  connection_pool_size: 100
  request_timeout: 30
  batch_processing: true
  cache_enabled: true
  cache_ttl: 3600

通过以上步骤，你就可以成功安装并开始使用 LLMESH 网络了。接下来可以探索更多高级功能和配置选项。

Previous核心架构与特性 Next网络状态与性能

Last updated 1 day ago

# 克隆仓库 git clone https://github.com/llmesh-cor/llmesh.git cd llmesh # 创建虚拟环境 python -m venv llmesh-env source llmesh-env/bin/activate # Linux/macOS # 或在 Windows 上 # llmesh-env\Scripts\activate # 安装依赖 pip install -r requirements.txt # 安装 LLMESH pip install -e .

# 拉取官方镜像 docker pull llmesh/llmesh-node:latest # 运行容器 docker run -d \ --name llmesh-node \ -p 8080:8080 \ -p 9000:9000 \ -v ~/.llmesh:/root/.llmesh \ llmesh/llmesh-node:latest

# ~/.llmesh/config.yaml node: id: "node_${RANDOM_ID}" name: "My LLMESH Node" description: "Personal LLMESH node for AI services" network: listen_address: "0.0.0.0" listen_port: 9000 bootstrap_nodes: - "bootstrap1.llmesh.network:9000" - "bootstrap2.llmesh.network:9000" max_peers: 50 models: storage_path: "~/.llmesh/models" cache_size: "5GB" supported_formats: ["onnx", "pytorch", "tensorflow"] tokens: mesh_token_address: "0x1234567890123456789012345678901234567890" wallet_private_key: "${MESH_PRIVATE_KEY}" logging: level: "INFO" file: "~/.llmesh/logs/llmesh.log"

#!/usr/bin/env python3 """部署模型示例""" from llmesh import ModelDeployer, ModelConfig def deploy_model(): # 创建模型配置 config = ModelConfig( name="my-gpt-model", model_path="./models/gpt-model.onnx", model_type="text-generation", description="Personal GPT model for text generation", fee_per_request=0.1, # 每次请求收费 0.1 MESH max_context_length=2048, temperature=0.7 ) # 创建部署器 deployer = ModelDeployer() # 部署模型 try: deployment = deployer.deploy(config) print(f"✅ Model deployed successfully!") print(f"📝 Model ID: {deployment.model_id}") print(f"🔗 Access URL: {deployment.access_url}") print(f"💰 Fee per request: {config.fee_per_request} MESH") return deployment except Exception as e: print(f"❌ Deployment failed: {e}") return None if __name__ == "__main__": deploy_model()

# 部署 ONNX 模型 llmesh-deploy ./models/my-model.onnx \ --name "my-llm" \ --type "text-generation" \ --fee 0.1 \ --description "My personal LLM model" # 查看已部署的模型 llmesh models list # 输出示例： # Model ID Name Type Fee Status # abc123def456 my-llm text-generation 0.1 Active # def789ghi012 chat-bot conversational 0.05 Active

#!/usr/bin/env python3 """调用模型示例""" from llmesh import LLMeshClient async def main(): # 创建客户端 client = LLMeshClient() # 搜索可用模型 models = await client.search_models( model_type="text-generation", max_fee=0.2 ) print(f"Found {len(models)} available models:") for model in models: print(f" - {model.name} (Fee: {model.fee} MESH)") if models: # 选择第一个模型进行推理 model = models[0] # 发送推理请求 response = await client.generate_text( model_id=model.id, prompt="What is the future of artificial intelligence?", max_tokens=200, temperature=0.7 ) print(f"\n🤖 Model Response:") print(f"📝 Text: {response.text}") print(f"💰 Cost: {response.cost} MESH") print(f"⏱️ Response time: {response.response_time}ms") if __name__ == "__main__": import asyncio asyncio.run(main())

# config.yaml compute: use_gpu: true gpu_devices: [0, 1] # 使用 GPU 0 和 1 gpu_memory_limit: "8GB" batch_size: 8 models: inference_backend: "cuda" # 或 "cpu", "opencl" mixed_precision: true optimization_level: "O2"

#!/usr/bin/env python3 """集群部署脚本""" from llmesh import ClusterManager, NodeConfig def setup_cluster(): cluster = ClusterManager() # 配置主节点 master_config = NodeConfig( role="master", compute_resources={"cpu": 16, "memory": "32GB", "gpu": 2}, services=["routing", "model_hosting", "load_balancing"] ) # 配置工作节点 worker_configs = [] for i in range(3): worker_config = NodeConfig( role="worker", compute_resources={"cpu": 8, "memory": "16GB", "gpu": 1}, services=["model_hosting"] ) worker_configs.append(worker_config) # 部署集群 cluster.deploy_master(master_config) for worker_config in worker_configs: cluster.add_worker(worker_config) print("✅ Cluster deployed successfully!") print(f"📊 Master node: {cluster.master_node.id}") print(f"👥 Worker nodes: {len(cluster.worker_nodes)}") if __name__ == "__main__": setup_cluster()

# 检查模型格式 from llmesh.utils import ModelValidator validator = ModelValidator() is_valid = validator.validate_model("./my-model.onnx") if not is_valid: print("❌ Model format is not supported") print("Supported formats:", validator.supported_formats)

系统要求

最低系统要求

推荐配置

安装方法

方法 1：通过 pip 安装（推荐）

方法 2：从源码安装

方法 3：使用 Docker

配置环境

初始化配置

配置文件示例

快速开始

1. 启动第一个节点

2. 部署你的第一个 LLM 模型

3. 使用命令行工具部署模型

4. 调用网络中的模型

5. Web 界面使用

高级配置

GPU 加速配置

集群部署配置

常见问题解决

连接问题

模型加载问题

性能调优

系统要求

最低系统要求

推荐配置

安装方法

方法 1：通过 pip 安装（推荐）

方法 2：从源码安装

方法 3：使用 Docker

配置环境

初始化配置

配置文件示例

快速开始

1. 启动第一个节点

2. 部署你的第一个 LLM 模型

3. 使用命令行工具部署模型

4. 调用网络中的模型

5. Web 界面使用

高级配置

GPU 加速配置

集群部署配置

常见问题解决

连接问题

模型加载问题

性能调优