Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Run DeepSeek Harness Web with a LAN-hosted Ollama model

AI summary

A hands-on guide explains how to install DeepSeek Harness Web, start it with dsh web, and connect it to Ollama 0.40.16 running on another LAN host. The author documents isolated web and desktop profiles, provider configuration, port and model-tag troubleshooting, and a custom proxy that disables reasoning output. The author also reports that both Ollama’s native port and the proxy port were reachable from the test machine, contrary to a commonly repeated claim.

In-depth insight

The tested configuration and troubleshooting steps provide a practical reference for running DeepSeek Harness with a private local model without relying on cloud API calls.

本文基于我自己的局域网环境做了实测验证(Ollama 0.40.16、DeepSeek Harness 桌面版 / Web 版双 profile)。主线只有一条:Web 版怎么装、怎么起;我这套环境(Harness 在 172.17.16.46、Ollama 在 172.17.16.52)是附带案例,用来把接本地模型时最容易混淆的端口选择、模型 tag、配置隔离讲透,并纠正一条很多教程里被实测推翻的说法(见第七节)。

一、环境准备

先过一遍清单,缺哪项后面都会卡:

项要求 / 实测
Node.js^22.19.0 或 >=24.0.0;实测本机 v24.14.1,过关
npm随 Node 自带;实测 11.11.0
Ollama(可选)只有要接本地模型才需要,实测 0.40.16
桌面版安装包内置运行时,双击即用,不需要 Node
Web 版需要 Node + 一个常开的终端窗口
默认端口3080,地址 http://127.0.0.1:3080

#DeepSeek Harness#Ollama#qwen-api:latest#qwen3.5:4b-q4_K_M#dsh-local-model

Original · 11 min readRead the original
Group chat

No comments yet. Start the conversation.