dsh-webfetch
In a curated listMaintenance: Activetyeclipse/dsh-webfetch
Web page reader for DeepSeek Harness: clean markdown/text extraction, link inventory, RSS/Atom feeds, HTTP header probes, HTML tables, page metadata (canonical/OG/hreflang/JSON-LD) and site URL inventories from robots.txt + sitemap.xml — zero runtime dependencies
$ dsh plugin add dsh-webfetchInstall
dsh has no central install command — add this plugin’s entry (documented in its README) to your profile or patch config, then restart.
How installs work3
stars
0
forks
TypeScript
Language
MIT
License
2026-08-13
Created
2026-11-16
Last push
README
dsh-webfetch
为 DeepSeek Harness 智能体装上「阅读器」:给定 URL,抓取网页并提取干净的 Markdown / 纯文本正文,附带链接清单、RSS/Atom 订阅源解析、HTTP 头部探测、网页表格结构化提取与页面元数据(canonical / Open Graph / hreflang / JSON-LD)读取。零运行时依赖,只读,不发送任何凭证。
English | 中文简介
为 DeepSeek Harness 智能体装上「阅读器」:给定 URL,抓取网页并提取干净的 Markdown / 纯文本正文,附带链接清单、RSS/Atom 订阅源解析、HTTP 头部探测、网页表格结构化提取与页面元数据(canonical / Open Graph / hreflang / JSON-LD)读取。零运行时依赖,只读,不发送任何凭证。 English | 中文简介
A web page reader plugin for DeepSeek Harness (dsh).
dsh agents can search, but until now they could not read the page behind a URL.
dsh-webfetch closes that gap with eight read-only tools and zero runtime dependencies (Node built-ins + global fetch only).
Tools
web_fetch
Fetch a web page and extract its readable content.
| Parameter | Type | Default | Description |
|---|---|---|---|
url |
string (required) | — | Full http/https URL of the page to fetch. |
format |
'markdown' | 'text' |
markdown |
Markdown keeps headings, links, lists and code fences; text is plain prose. |
extractLinks |
boolean | false |
Also return every link found on the page (resolved, absolute). |
maxChars |
number | 50000 |
Cap on extracted content length (1000–200000). |
Returns { url, finalUrl, status, title, content, length, truncated, links? }.
user: what does the dsh README say about the architecture?
agent: web_fetch("https://github.com/deepseek-ai/deepseek-harness")
→ HTTP 200 — title: deepseek-harness
## DeepSeek Harness
..."everything is a plugin"...
web_links
Collect every link on a page with its visible label, resolved to absolute
URLs, deduplicated, capped at limit (1–200, default 50). Useful for mapping
what a page points to or crawling site structure.
web_feed
Read an RSS 2.0 or Atom feed and return its entries as a clean,
LLM-friendly listing. Feed URLs are XML documents that web_fetch cannot
extract from — web_feed parses them into structured entries instead.
| Parameter | Type | Default | Description |
|---|---|---|---|
url |
string (required) | — | Full http/https URL of the RSS or Atom feed. |
maxItems |
number | 10 |
Max entries to return (1–50). |
includeContent |
boolean | false |
Also return each entry's full content (else summaries only). |
Returns { url, finalUrl, status, feedTitle, entryCount, truncated, entries },
where each entry is { title, url, published?, author?, summary?, content? }
with CDATA unwrapped, HTML entities decoded, markup stripped and relative
links resolved against the feed URL.
user: what did the example blog post this week?
agent: web_feed("https://blog.example.com/feed.xml", maxItems: 5)
→ feed: Example Blog
5 entries from https://blog.example.com/feed.xml
1. First & foremost post — https://blog.example.com/posts/first
published: Mon, 01 Jan 2024 10:00:00 GMT
author: Alice
Hello world — café & tea.
web_headers
Inspect the HTTP status, response headers and redirect chain of a URL
without downloading the page body — the diagnostic companion to
web_fetch: check status codes, content types, redirects, caching or
security headers before (or instead of) pulling the content. Uses HEAD
by default and falls back to GET automatically when the server answers
405/501; unlike web_fetch, any status is reported (404/500/… included)
rather than thrown.
| Parameter | Type | Default | Description |
|---|---|---|---|
url |
string (required) | — | Full http/https URL to inspect. |
method |
'HEAD' | 'GET' |
HEAD |
HEAD downloads no body; GET always works but transfers the body. |
followRedirects |
boolean | true |
Follow redirects and report every hop of the chain. |
Returns { url, finalUrl, status, statusText, method, headers, redirects }
where headers is the full response-header map (lower-cased keys) and
redirects lists each hop as { url, status, location }.
user: is that download link still alive, and does it redirect?
agent: web_headers("https://example.com/downloads/latest")
→ HTTP 200 OK — HEAD https://example.com/downloads/latest
redirect chain:
1. 301 https://example.com/downloads/latest → https://cdn.example.com/releases/latest.zip
final URL: https://cdn.example.com/releases/latest.zip
12 header(s):
content-type: application/zip
content-length: 4821337
...
web_table
Extract the HTML tables of a page as structured rows — the tables
counterpart of web_fetch: spec sheets, pricing pages, comparison charts
and any data locked in <table> markup come back as arrays of cell
strings instead of mashed-together prose. The first row is reported as the
header when the page marks it (<thead> or an all-<th> first row);
colspan / rowspan are expanded into a rectangular grid with the
spanning text repeated in every covered slot (rows stay self-contained);
nested tables are flattened into the containing cell; cells are
entity-decoded, whitespace-normalized and capped at 200 characters.
More in Web & Browser
anysearch-dsh
by anysearch-team
AnySearch web search provider and advanced search tools for DeepSeek Harness (DSH)
dsh-find-plugin
by awesome-dsh-plugin
In-agent plugin discovery: live GitHub repository search scoped to the official `dsh-plugin` topic, star-ranked; every result comes with a ready-to-run `dsh plugin add` command.
fnos-dsh
by fnosp
Multi-source web search and fetch failover for DeepSeek Harness sessions: takes over the session searchProvider and fetchProvider seams and tries configured sources in a chosen order — TinyFish, Tavily and the official search as a fallback that always participates and can be repositioned but not removed. Multiple API keys per platform are spread round-robin, usage snapshots drive pre-flight exclusion of exhausted accounts, and per-request timeouts plus same-platform retry control failover granularity. Web fetching runs local HTTP first, then falls back to TinyFish Fetch server-side rendering and Tavily Extract, so pages still load when local egress is blocked by proxy or DNS policies. A per-platform panel shows key usage and plan limits. Install from npm with `dsh plugin --profile web add @tnnevol/dsh-failover-search`.
modsearch
by liustack
Web/X 搜索与抓取桥:为纯文本 coding agent 提供结构化 JSON 证据(search/fetch/citations),引擎可选 Antigravity CLI(免 key)/带 key 引擎/X(Grok Build 登录态)。
