# 环境变量配置
export ZAI_API_KEY="zai-..."

# OpenAI 兼容客户端（Cline、OpenCode、Roo Code 等）
export OPENAI_BASE_URL="https://api.z.ai/api/coding/paas/v4"
export OPENAI_API_KEY="$ZAI_API_KEY"
export OPENAI_MODEL="glm-5.2[1m]"  # 启用 1M 上下文

# Claude Code 切换到 GLM 5.2
export ANTHROPIC_BASE_URL="https://api.z.ai/api/anthropic"
export ANTHROPIC_AUTH_TOKEN="$ZAI_API_KEY"
export ANTHROPIC_MODEL="glm-5.2[1m]"
export API_TIMEOUT_MS="3000000"

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.z.ai/api/coding/paas/v4",
    api_key=os.environ["ZAI_API_KEY"],
)
resp = client.chat.completions.create(
    model="glm-5.2[1m]",
    messages=[{"role": "user", "content": "用 Python 写一个反转链表的函数"}],
    temperature=0.7,
    max_tokens=1024,
)
print(resp.choices[0].message.content)

# 安装依赖
pip install huggingface-hub

# 下载 FP8 版本（推荐生产用）
huggingface-cli download zai-org/GLM-5.2-FP8 \
  --local-dir /data/models/glm52-fp8 \
  --local-dir-use-symlinks False

# 或使用国内镜像加速
HF_ENDPOINT=https://hf-mirror.com huggingface-cli download zai-org/GLM-5.2-FP8 \
  --local-dir /data/models/glm52-fp8

# 安装 vLLM
pip install vllm==0.23.0

# 或使用 Docker
docker pull vllm/vllm-openai:nightly

vllm serve "zai-org/GLM-5.2-FP8" \
  --tensor-parallel-size 8 \
  --max-model-len 262144 \
  --kv-cache-dtype fp8_e5m2 \
  --enable-prefix-caching \
  --enable-chunked-prefill \
  --enable-expert-parallel \
  --gpu-memory-utilization 0.92 \
  --port 8000

python -m vllm.entrypoints.openai.api_server \
  --model /data/models/glm52-fp8 \
  --served-model-name glm52-fp8 \
  --tensor-parallel-size 8 \
  --quantization fp8 \
  --enable-expert-parallel \
  --max-model-len 1048576 \
  --kv-cache-dtype fp8_e5m2 \
  --gpu-memory-utilization 0.92 \
  --enable-chunked-prefill \
  --max-num-chunked-tokens 32768 \
  --max-num-seqs 1 \
  --port 8000

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="not-needed"
)

response = client.chat.completions.create(
    model="zai-org/GLM-5.2-FP8",
    messages=[
        {"role": "user", "content": "用 Python 写一个快速排序算法"}
    ],
    max_tokens=512,
    temperature=0.7
)
print(response.choices[0].message.content)

# 安装
pip install 'sglang[all]>=0.5.10'

# 或 Docker
docker pull lmsysorg/sglang:latest

python -m sglang.launch_server \
  --model-path /data/models/glm52-fp8 \
  --tp 8 \
  --quantization fp8 \
  --enable-moe-ep \
  --context-length 1048576 \
  --mem-fraction-static 0.88 \
  --kv-cache-dtype fp8_e5m2 \
  --port 30000

docker run --gpus all \
  --shm-size 32g \
  -p 30000:30000 \
  -v /data/models/glm52-fp8:/models/glm52-fp8 \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  --env "HF_TOKEN=<your_token>" \
  --ipc=host \
  lmsysorg/sglang:latest \
  python3 -m sglang.launch_server \
    --model-path /models/glm52-fp8 \
    --host 0.0.0.0 \
    --port 30000 \
    --tp 8 \
    --context-length 262144 \
    --kv-cache-dtype fp8_e4m3

# 一行命令搞定
ollama pull glm5
ollama run glm5

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "zai-org/GLM-5.2-FP8"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True
)

messages = [{"role": "user", "content": "什么是 MoE 架构？请详细解释。"}]
inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt"
).to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    temperature=0.7,
    do_sample=True
)

response = tokenizer.decode(
    outputs[0][inputs["input_ids"].shape[-1]:],
    skip_special_tokens=True
)
print(response)