LLM

2025-07-09

API

mix

名称 网址
OpenCode https://opencode.ai/console
Cursor https://cursor.com/dashboard/usage
Command Code https://commandcode.ai/usage
OpenRouter https://openrouter.ai/settings/credits
AiHubMix https://console.aihubmix.com/statistics
Groq https://console.groq.com/dashboard/usage
OpenDesign https://open-design.ai/
Kilo https://kilo.ai/
Vercel AI Gateway https://vercel.com/ai-gateway
Cloudflare Workers AI https://dash.cloudflare.com/workers-ai/usage
硅基流动 https://cloud.siliconflow.cn/me/bills
七牛 https://portal.qiniu.com/ai-inference/bill
火山引擎 https://console.volcengine.com/
阿里云百炼 https://bailian.console.aliyun.com/
非线智能 https://nonelinear.com/static/signin.html
华为云 MaaS https://console.huaweicloud.com/modelarts/
Portkey https://app.portkey.ai/dashboard
Helicone https://us.helicone.ai/dashboard
Together AI https://api.together.ai/settings/billing
Fireworks AI https://fireworks.ai/account/billing
DeepInfra https://deepinfra.com/dash/billing
Hugging Face https://huggingface.co/settings/inference-providers
Replicate https://replicate.com/account/billing
Not Diamond https://api.notdiamond.ai/

manufacturer

自研模型且提供开放 API 的平台

国内

名称 网址
DeepSeek https://platform.deepseek.com/usage
智谱 https://bigmodel.cn/finance-center/bill/expensebill/list
阶跃星辰 https://platform.stepfun.com/account-overview
月之暗面 https://platform.moonshot.cn/console/fees
通义千问 https://dashscope.console.aliyun.com/billing
文心一言 https://console.bce.baidu.com/qianfan/ais/console/onlineService
腾讯混元 https://console.cloud.tencent.com/hunyuan/overview
豆包 https://console.volcengine.com/ark/region:ark+cn-beijing/billing
MiniMax https://platform.minimaxi.com/user-center/payment/balance
科大讯飞星火 https://console.xfyun.cn/services/bm4
商汤日日新 https://platform.sensenova.cn/user/bill
百川智能 https://platform.baichuan-ai.com/settings/billing
零一万物 https://platform.lingyiwanwu.com/billing
华为云盘古 https://console.huaweicloud.com/modelarts/?region=cn-north-4#/billing
美团龙猫 https://longcat.chat/platform
小米 MiMo https://aistudio.xiaomimimo.com/

国外

名称 网址
OpenAI https://platform.openai.com/usage
Anthropic https://console.anthropic.com/settings/plans
Google Gemini https://aistudio.google.com/apikey
xAI https://console.x.ai/team/default/billing
Mistral AI https://admin.mistral.ai/billing
Cohere https://dashboard.cohere.com/billing
AI21 Labs https://studio.ai21.com/account/billing
AWS Bedrock https://console.aws.amazon.com/bedrock/home
Azure AI https://ai.azure.com/managementCenter/billing
IBM watsonx https://dataplatform.cloud.ibm.com/iam/usage
NVIDIA https://build.nvidia.com/settings/billing
Aleph Alpha https://docs.aleph-alpha.com/docs/account-management
Inception Labs https://platform.inceptionlabs.ai/dashboard
Perplexity https://www.perplexity.ai/settings/api

Chat

网站名称 网址 开发公司
DeepSeek https://chat.deepseek.com 深度求索
Kimi https://www.kimi.com 月之暗面 (Moonshot AI)
MiMo https://aistudio.xiaomimimo.com/#/c 小米
豆包 https://www.doubao.com 字节跳动
通义千问 https://www.qianwen.com 阿里巴巴
海螺 AI https://agent.minimaxi.com MiniMax
腾讯元宝 https://yuanbao.tencent.com 腾讯
龙猫 AI https://longcat.chat 美团
Z.AI https://chat.z.ai 智谱 AI
文心一言 https://yiyan.baidu.com 百度
商汤 https://chat.sensetime.com 商汤科技
百川 https://ying.baichuan-ai.com 百川智能
星火 https://xinghuo.xfyun.cn/desk 科大讯飞

chat px

网站名称 网址 开发公司
ChatGPT https://chatgpt.com OpenAI
Gemini https://gemini.google.com Google
Grok https://grok.com xAI
Copilot https://copilot.microsoft.com Microsoft
Mistral AI https://chat.mistral.ai Mistral AI
Inception https://chat.inceptionlabs.ai Inception Labs
Perplexity https://www.perplexity.ai Perplexity AI
Pi https://pi.ai Pi
Notdiamond https://chat.notdiamond.ai Notdiamond

chat mix

网站名称 网址 开发公司
Poe https://poe.com Quora
Hugging Chat https://huggingface.co/chat Hugging Face
Together AI https://chat.together.ai Together AI
Groq https://console.groq.com/playground Groq
AskManyAI https://askmany.cn 浮点奇迹

IDE

名称 网址
VS Code https://code.visualstudio.com
Cursor https://cursor.com
Antigravity https://antigravity.google
Trae https://www.trae.ai
Verdent https://www.verdent.ai
Qoder https://qoder.com
CodeBuddy https://www.codebuddy.ai
Devin Desktop(原 Windsurf) https://devin.ai
Kiro https://kiro.dev
Void https://voideditor.com

Desktop Agent

名称 网址
Qoder https://qoder.com / https://qoder.cn
CodeBuddy https://www.codebuddy.ai / https://www.codebuddy.cn
WorkBuddy https://www.workbuddy.ai / https://www.workbuddy.cn/
TRAE / TRAE Work https://www.trae.ai / https://www.trae.cn
Cursor https://cursor.com
Devin Desktop(原 Windsurf) https://devin.ai
Claude Desktop / Cowork https://claude.com/product/cowork
Claude Code(Desktop) https://code.claude.com
QClaw https://qclaw.qq.com
Marvis https://marvis.qq.com
Kimi Work https://www.kimi.com/products/kimi-work
AutoClaw https://autoclaw.z.ai
LobsterAI / 有道龙虾 https://lobsterai.youdao.com
OpenClaw https://openclaw.ai
Manus https://manus.im
DuMate(百度搭子) https://www.dumate.cn
Google Antigravity https://antigravity.google
GitHub Copilot app https://github.com/features/copilot
华为云码道(CodeArts) https://codearts.huaweicloud.com
Kilo Code https://kilo.ai
Kiro https://kiro.dev
豆包 https://www.doubao.com
Qwen Work(通义) https://qwenwork.cn

Code CLI Tools

工具名称 网址
Claude Code https://docs.anthropic.com/zh-CN/docs/claude-code/settings
Codex https://openai.com/codex
Cursor CLI https://cursor.com/cli
Gemini CLI https://github.com/google-gemini/gemini-cli
Pi https://pi.dev/
phi https://pulseaiclub.github.io/
opencode https://opencode.ai/
Cline CLI https://docs.cline.bot/cline-cli/getting-started
AMP https://ampcode.com/manual
iFlow CLI https://cli.iflow.cn/
Qwen Code https://github.com/QwenLM/qwen-code
Qoder CLI https://docs.qoder.com/zh/cli/quick-start
Kode https://github.com/shareAI-lab/Kode
Kilo CLI https://kilo.ai/cli

pi 插件

允许 AI 代理观察和控制 macOS、Windows 和 Linux 应用程序。

https://pi.dev/packages/@injaneity/pi-computer-use

网络搜索、URL 获取、GitHub 仓库克隆、PDF 提取、YouTube 视频理解和本地视频分析 Pi 编码代理。支持 OpenAI、勇敢、并行、TinyFish、Search1API、Search Infinity、Querit、Tavily、Firecrawl、Crawl4AI、Jina、SE

https://pi.dev/packages/pi-web-access

在微信里继续 pi 对话:双向同步、图片与文件、会话交接。

https://pi.dev/packages/pi-wechat-assistant

接入cursor cli模型

https://pi.dev/packages/@jiah-liu/pi-cursor-provider

支持 tool_calls 本地模型

模型 实测 tool_calls pi 可用性
qwen3.5:9b ✅ ✅
qwen3.8:latest ✅ ✅
gpt-oss:20b ✅ ✅
mistral-small:24b ✅ ✅
minicpm5-2b ✅ ✅

cursor 连通测试接口

https://api2.cursor.sh/

https://agentn.global.api5.cursor.sh/

Claude Code 配置

切换到litellm

~/.claude/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://127.0.0.1:4000",
    "ANTHROPIC_AUTH_TOKEN": "sk-litellm-static-key",
    "CLAUDE_CODE_MAX_OUTPUT_TOKENS": "16384"
  },
  "model": "claude-3-5-haiku-20241022"
}

Windows下使用Claude Code

1.0.51 (Claude Code) 以上版本支持Windows了

需在添加用户环境变量:

CLAUDE_CODE_GIT_BASH_PATH=C:\Program Files\Git\bin\bash.exe

Gemini CLI 配置

# 认证
# Get your key from https://aistudio.google.com/apikey
export GEMINI_API_KEY="YOUR_API_KEY"
gemini

# Proxy
# set the proxy
$env:http_proxy = "http://xxx.xxx.xxx.xxx:8080"
$env:https_proxy = "http://xxx.xxx.xxx.xxx:8080"

# unset the proxy
$env:http_proxy = $null
$env:https_proxy = $null

MCP

chrome-devtools-mcp

https://github.com/ChromeDevTools/chrome-devtools-mcp

直接控制和检查真实Chrome浏览器(Chrome 145+)

先开启浏览器的远程调试:chrome://inspect/#remote-debugging

检查端口是否开启(默认9222):netstat -na -p TCP|grep LISTENING

内网搭建LLM服务架构

+-----------------------------+
|        Frontend UI          | ←→ 用户
|        (Open WebUI)         |
+-----------------------------+
               │
               ▼
+-----------------------------+
|         API Gateway         | ←→ 鉴权 / 限流 / 多模型路由
|         (LiteLLM)           |
+-----------------------------+
               │
               ▼
+-----------------------------+
|      LLM Inference Layer    | ←→ 多实例,支持横向扩容
| (vLLM / TGI / Ollama 等)    |
+-----------------------------+
               │
               ▼
+-----------------------------+
|         Model Store         | ←→ 存模型文件、本地磁盘或对象存储
+-----------------------------+

UI:

https://github.com/open-webui/open-webui

GW:

https://www.litellm.ai/

LLM:

https://github.com/vllm-project/vllm

https://ollama.com/download

https://lmstudio.ai/

vllm

# 查看模型
hf cache scan

# 安装vllm (注意使用gpt-oss模型务必使用新的虚拟环境,旧环境存在包冲突)
uv venv --python 3.12 --seed
source .venv/bin/activate
uv pip install --pre vllm==0.10.1+gptoss \
    --extra-index-url https://wheels.vllm.ai/gpt-oss/ \
    --extra-index-url https://download.pytorch.org/whl/nightly/cu128 \
    --index-strategy unsafe-best-match


# 加载 Hugging Face 在线模型(自动下载)
vllm serve openai/gpt-oss-20b

# 下载模型
hf download openai/gpt-oss-20b


HF环境变量:

# 设置 Hugging Face 缓存的本地目录位置,所有模型、tokenizer、dataset 等将保存在这里,而不是默认的 ~/.cache/huggingface
export HF_HOME=/mnt/jqsd4_dataware/LLM/model_llm/.cache/huggingface

# 启用 Hugging Face Hub 的加速下载功能(hf_transfer),可提升大文件下载速度(如模型权重)
export HF_HUB_ENABLE_HF_TRANSFER=1

# 启用离线模式,不会尝试从 Hugging Face 服务器下载内容,只使用本地已有缓存
export HF_HUB_OFFLINE=1

# 默认使用 CPU(防止意外使用 GPU)
export TRANSFORMERS_NO_CUDA=1

# 设置 Hugging Face 的 API 镜像地址(如清华镜像或其他加速服务),用于替代默认的 https://huggingface.co
export HF_ENDPOINT=https://hf-mirror.com

# 禁用所有远程镜像访问(用于完全离线环境)
export HF_HUB_DISABLE_TELEMETRY=1

ollama

https://ollama.com/library/gpt-oss

用systemctl启动

# vim /etc/systemd/system/ollama.service
[Unit]
Description=Ollama Service
After=network-online.target

[Service]
ExecStart=/usr/local/bin/ollama serve
User=ollama
Group=ollama
Restart=always
RestartSec=3
Environment="PATH=/root/.nvm/versions/node/v22.17.0/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin:/data/tools/nvim/nvim-linux-x86_64/bin"
Environment="OLLAMA_HOST=0.0.0.0"
Environment="OLLAMA_MODELS=\"/opt/cw_test/ollama/models\""

[Install]
WantedBy=default.target


# 重新加载配置
systemctl restart ollama

# 重启服务
systemctl restart ollama

用脚本启动

#!/bin/sh
export OLLAMA_MODELS="/data/ollama/models"
export OLLAMA_HOST=0.0.0.0
export OLLAMA_ORIGINS=*
export http_proxy=http://127.0.0.1:10808
export https_proxy=http://127.0.0.1:10808
# 启动ollama服务 注意在启动前配置代理
ollama serve


# 使用cli下载使用 gpt-oss-20b (注意不能在这里配置代理, 下载请求是发往服务端的)
ollama pull gpt-oss:20b
ollama run gpt-oss:20b


litellm

启动

litellm –config config.yaml –port 5000

config.yaml 配置:

# 这是一个示例的 litellm 配置文件 config.yaml
litellm_settings:
  drop_params: true  # 是否丢弃未使用的参数
  enable_stream: true  # 是否启用流式输出

model_list:
  - model_name: GLM-4-Flash-250414
    litellm_params:
      api_base: https://open.bigmodel.cn/api/paas/v4/
      api_key: your_api_key_here
      model: openai/GLM-4-Flash-250414
      temperature: 0.7  # 模型温度参数

  - model_name: glm-4v-flash
    litellm_params:
      api_base: https://open.bigmodel.cn/api/paas/v4/
      api_key: your_api_key_here
      model: openai/glm-4v-flash
      max_tokens: 150  # 最大输出token数

  - model_name: claude-3-5-haiku-20241022
    litellm_params:
      api_base: https://api.siliconflow.cn/v1
      api_key: your_api_key_here
      model: openai/deepseek-ai/DeepSeek-R1-Distill-Qwen-7B

  - model_name: DeepSeek-R1-0528-Qwen3-8B
    litellm_params:
      api_base: https://api.siliconflow.cn/v1
      api_key: your_api_key_here
      model: openai/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B

  # - model_name: claude-3-5-haiku-20241022
  #   给claude-code使用这里可以用这个模型名称

  - model_name: deepseek-v3-250324
    litellm_params:
      model: volcengine/deepseek-v3-250324
      api_key: your_api_key_here

  - model_name: deepseek-r1-250528
    litellm_params:
      model: volcengine/deepseek-r1-250528
      api_key: your_api_key_here

  - model_name: gemma-3-4b-it
    litellm_params:
      model: hosted_vllm/google/gemma-3-4b-it
      api_base: http://172.16.1.166:8000/v1
      provider: google

  - model_name: gemini-2.5-flash
    litellm_params:
      model: gemini/gemini-2.5-flash-preview-04-17
      api_key: your_api_key_here

  - model_name: kimi-dev-72b
    litellm_params:
      model: openrouter/moonshotai/kimi-dev-72b:free
      api_key: your_api_key_here

  - model_name: deepseek-r1t2-chimera
    litellm_params:
      model: openrouter/tngtech/deepseek-r1t2-chimera:free
      api_key: your_api_key_here

  - model_name: deepseek-r1
    litellm_params:
      model: openrouter/deepseek/deepseek-r1:free
      api_key: your_api_key_here

  - model_name: mistral-small-3.2-24b-instruct-2506:free
    litellm_params:
      model: openrouter/mistralai/mistral-small-3.2-24b-instruct-2506:free
      api_key: your_api_key_here
      api_base: https://openrouter.ai/api/v1

FIM

continue、copilot、fitten code、Amazon Q

continue 插件

~/.continue/config.yaml

name: Local Assistant
version: 1.0.0
schema: v1
models:
  - name: GLM-4-Flash-250414
    provider: openai
    model: GLM-4-Flash-250414
    apiKey: your_api_key_here
    apiBase: https://open.bigmodel.cn/api/paas/v4/

  - name: a100/GLM-4-Flash-250414
    provider: openai
    model: GLM-4-Flash-250414
    apiKey: your_api_key_here
    apiBase: http://172.16.1.166:4000/v1

  - name: a100/gemma-3-4b-it
    provider: openai
    model: google/gemma-3-4b-it
    apiKey: your_api_key_here
    apiBase: http://172.16.1.166:8000/v1
    roles:
      - autocomplete
      - chat
      - embed
    max-tokens: 3000

  - name: a100/火山引擎/deepseek-v3-250324
    provider: deepseek
    model: deepseek-v3-250324
    apiKey: your_api_key_here
    apiBase: http://172.16.1.166:4000/v1

  - name: a100/火山引擎/DeepSeek-R1-0528-Qwen3-8B
    provider: openai
    model: DeepSeek-R1-0528-Qwen3-8B
    apiKey: your_api_key_here
    apiBase: http://172.16.1.166:4000/v1

  - name: 硅基流动/DeepSeek-R1-Distill-Qwen-7B
    provider: siliconflow
    model: deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
    apiKey: your_api_key_here
    apiBase: https://api.siliconflow.cn/v1

  - name: FIM/硅基流动/Qwen2.5-Coder-7B-Instruct
    provider: siliconflow
    model: Qwen/Qwen2.5-Coder-7B-Instruct
    apiKey: your_api_key_here
    apiBase: https://api.siliconflow.cn/v1
    roles:
      - autocomplete

  - name: Gemini 2.5 Pro Experimental
    provider: gemini
    model: gemini-2.5-pro-exp-03-25
    apiKey: your_api_key_here

  - name: a100-vllm/gemma-3-4b-it
    provider: openai
    model: google/gemma-3-4b-it
    apiKey: your_api_key_here
    apiBase: http://172.16.1.166:8000/v1
    chatOptions:
      baseSystemMessage: >-
        <important_rules>
          You are in chat mode.

          If the user asks to make changes to files offer that they can use the Apply Button on the code block, or switch to Agent Mode to make the suggested updates automatically.
          If needed consisely explain to the user they can switch to agent mode using the Mode Selector dropdown and provide no other details.

          Always include the language and file name in the info string when you write code blocks.
          If you are editing "src/main.py" for example, your code block should start with '```python src/main.py'

          When addressing code modification requests, present a concise code snippet that
          emphasizes only the necessary changes and uses abbreviated placeholders for
          unmodified sections. For example:

          ```language /path/to/file
          // ... existing code ...

          

          // ... existing code ...

          

          // ... rest of code ...
          ```

          In existing files, you should always restate the function or class that the snippet belongs to:

          ```language /path/to/file
          // ... existing code ...

          function exampleFunction() {
            // ... existing code ...

            

            // ... rest of function ...
          }

          // ... rest of code ...
          ```

          Since users have access to their complete file, they prefer reading only the
          relevant modifications. It's perfectly acceptable to omit unmodified portions
          at the beginning, middle, or end of files using these "lazy" comments. Only
          provide the complete file when explicitly requested. Include a concise explanation
          of changes unless the user specifically asks for code only.

        </important_rules>

        You are an expert software developer. You give helpful and concise
        responses.
        用中文回答问题
    roles:
      - autocomplete
      - chat
      - embed

  - name: openrouter/deepseek-r1t2-chimera:free
    provider: openai
    model: tngtech/deepseek-r1t2-chimera:free
    apiBase: https://openrouter.ai/api/v1
    apiKey: your_api_key_here

  - name: openrouter/mistral-small-3.2-24b-instruct-2506:free
    provider: openai
    model: mistralai/mistral-small-3.2-24b-instruct-2506:free
    apiKey: your_api_key_here
    apiBase: https://openrouter.ai/api/v1
    roles:
      - chat
      - embed
    max-tokens: 3000

context:
  - provider: code
  - provider: docs
  - provider: diff
  - provider: terminal
  - provider: problems
  - provider: folder
  - provider: codebase

安装

Neo4j Desktop 下载地址

https://neo4j.com/download/

AI报告流程

1.下载年日志,存成txt文本

2.用notebooklm分析日志,并建立演示文稿:

https://notebooklm.google.com

3.PDF转PPT

https://pdftoword.55.la/pdf-to-ppt/

4.用Grok生成图标:

https://grok.com

提示词:根据这个 架构优化、硬件集成、现场交付 3个关键词生成一些图标,要求无背景色,不要出现文字

5.去除图片背景:

https://bgsub.com/webapp/