鱼皮AI导航
鱼皮AI导航
  • 主页
  • 社区
  • AI工具
  • AI知识库
  • AI提示词
  • MCP
  • Skills
  • AI学习资源
  • AI资讯
  • GitHub开源
登录/注册

dsh-chatgpt-web

基于免费 ChatGPT 网页端(GPT 5.6 Luna)的 DeepSeek Harness 零成本对话模型提供者,支持流式 Markdown。

模型接入
访问 GitHub
文档README

dsh-chatgpt-web

English | 中文

Zero-cost conversational AI model provider for DeepSeek Harness powered by free ChatGPT Web (GPT 5.6 Luna).

dsh-chatgpt-web Interactive Demo


Overview

dsh-chatgpt-web turns your local browser session of ChatGPT into a seamless, $0.00 API-free conversational model provider directly inside DeepSeek Harness (DSH).

It connects via headless or visible Chrome browser automation to chatgpt.com, streaming real-time Markdown responses, code solutions, explanations, and reasoning back to your DSH chats without consuming API credits.

Why Pure Chat?

Small conversational web models (like GPT 5.6 Luna) excel at explanations, dialogue, Q&A, brainstorming, code snippet generation, and side-assistant tasks. By running in Pure Chat Mode, the bridge eliminates prompt overhead, tool hallucinations, and syntax errors of autonomous multi-tool execution loops, providing a fast, rock-solid, zero-cost LLM provider.

Key Features

  • 100% Free ($0.00 Cost): Uses your existing free ChatGPT Web session. No OpenAI API keys or credit cards needed.
  • Cordis Plugin-First Lifecycle: Seamlessly managed by DSH via ctx.effect. DeepSeek Harness starts the background sidecar automatically on launch and shuts it down on exit.
  • Dynamic Model Auto-Detection: Automatically detects your ChatGPT tier:
    • Free Accounts: Defaults to chatgpt-web/luna (gpt-5-6-luna).
    • Plus / Team Accounts: Automatically detects and exposes available paid models (such as gpt-4o, o1).
  • Full Streaming Markdown & Code Blocks: Delivers tokens in real time directly to the DSH Web UI or CLI.

Quick Start

1. Installation

Install @wlv-zedd/dsh-chatgpt-web in your DeepSeek Harness environment or profile:

▼
bash
复制代码
pnpm add @wlv-zedd/dsh-chatgpt-web

2. One-Time Browser Sign-In

Authenticate your ChatGPT account once:

▼
bash
复制代码
npx @wlv-zedd/dsh-chatgpt-web login

A dedicated Chrome window will open. Log into your OpenAI / ChatGPT account. Once the ChatGPT composer is visible, the browser session is safely and securely saved locally to ~/.dsh/storages/chatgpt-free/.

3. Enable in DeepSeek Harness

Add the provider to your ~/.dsh/settings.yaml:

▼
yaml
复制代码
providers: chatgpt-web: displayName: "ChatGPT Web (Free)" api: openai-responses baseURL: http://127.0.0.1:17841/v1 headers: Authorization: "Bearer chatgpt-web-free" streamIdleTimeoutMs: 300000 models: - id: chatgpt-web/luna name: "ChatGPT Web — Luna (Free)" contextWindow: 1050000 maxTokens: 32768 input: - text - image agent-default-model: provider: chatgpt-web model: chatgpt-web/luna

And add the sidecar plugin to your profile's cordis.patch.yml (or cordis.yml):

▼
yaml
复制代码
- insert: - id: dsh-chatgpt-web name: '@wlv-zedd/dsh-chatgpt-web'

Now start DeepSeek Harness:

▼
bash
复制代码
pnpm dsh web

DSH will automatically start the background sidecar process, connect to your authenticated ChatGPT session, and accept prompts!


Diagnostics & Health Check

Verify your setup at any time with the built-in diagnostic doctor:

▼
bash
复制代码
npx @wlv-zedd/dsh-chatgpt-web doctor

Example healthy output:

▼
text
复制代码
✓ Configuration is valid (~/.dsh/storages/chatgpt-free/config.json) ✓ Chrome executable found: C:\Program Files\Google\Chrome\Application\chrome.exe ✓ ChatGPT login state has authenticated browser evidence ✓ Responses proxy is healthy on 127.0.0.1:17841 Doctor result: ready

Notes & Limitations

  1. Unofficial Bridge: Operates via local Playwright browser automation on chatgpt.com. Not affiliated with or endorsed by OpenAI.
  2. Single-Session Concurrency: Runs within a single browser tab. Sequential queries and normal DSH agent chats work seamlessly; avoid launching parallel multi-subagent swarms against the same tab simultaneously.
  3. Pure Chat Only: This provider generates pure conversational responses, explanations, reasoning, and code blocks. It does not execute local filesystem, terminal, or autonomous tool loops.
  4. Standard Free Tier Rate Limits: Subject to standard OpenAI free-tier hourly usage limits.

Support & Community Perks

If you find this plugin helpful, consider sponsoring.

Claim up to $200 free API credits on AgentRouter via GitHub.

Claim $50 free API credits on Vyce AI for fast LLM inference.


License

MIT License © 2026 WLV-ZEDD. DeepSeek Harness is an open-source project by DeepSeek AI.

模型接入

快捷安装

复制到终端去安装插件

dshplugin--profilewebaddgithub:WLV-ZEDD/dsh-chatgpt-web

温馨提示

装插件等于在本机跑第三方代码,权限与你相同,能读文件、用凭据、访问网络,工具审批管不到。请只装可信来源,并尽量锁定 commit。

站长:程序员鱼皮 老鱼简历 商务合作 增值电信业务经营许可证
2026 鱼皮AI导航 | 沪ICP备19026706号-6
站长:程序员鱼皮 老鱼简历 商务合作 增值电信业务经营许可证
2026 鱼皮AI导航 | 沪ICP备19026706号-6
🔥热门工具
朱雀AI检测朱雀AI检测
朱雀AI检测助手朱雀AI检测助手
LovartLovart
老鱼简历老鱼简历
WritePass官网WritePass官网
面多多面多多
CursorCursor
星月写作星月写作
零感AI零感AI
墨得问题墨得问题
dsh 插件推荐
dsh-subscriptions
通过 OAuth 将 ChatGPT Codex、Claude、Grok 与 Antigravity 订阅注册为 DeepSeek Harness 的 LLM 提供方,无需为其配置 API Key。
dsh-zhipu-toolkit
智谱 BigModel GLM 双端点模型目录(Coding Plan 与普通 API),支持实时模型发现与实测校准的思考档位映射,提供凭据、端点与默认推理档的可视化设置卡片。
dsh-qwen38-compaction-fix
针对 NInfer 托管的 qwen3.8-27b 网关的压缩修复:在 dsh 上下文压缩与会话标题调用中关闭思考,避免 xhigh 推理耗尽全部输出预算;同一思路适用于其他启动方式,本包仅面向 NInfer。
dsh-model-sync
一键同步 / 自动同步供应商模型列表,与 /v1/models 保持一致。
dsh-openrouter-provider-advisor
按量化、速度、缓存重度价格、上下文和可用率为当前 DSH 模型推荐 OpenRouter 供应商,并支持一键切换以帮助降低 token 成本。