jaco-tech/dsh-web-fetch-crw0

@jaco-tech/dsh-web-fetch-crw

crw (Firecrawl-compatible) fetch provider for the DeepSeek Harness web capability seam (ctx.web). No auth, no API key.

包名
@jaco-tech/dsh-web-fetch-crw
版本
0.1.0
许可证
MIT
最近更新
2026年8月19日

安装

此插件尚未提供可验证的 bundle,或兼容性检查未通过。请先阅读仓库说明。 阅读完整 README ↗

dsh-web-fetch-crw

dsh-plugin CI/CD

crw (Firecrawl-compatible) fetch provider plugin for DeepSeek Harness.

Registers a WebFetchProvider into ctx.web so that the existing web_fetch model tool retrieves page content through your crw instance — which converts HTML to clean markdown. No auth, no API key.

crw is a self-hosted web scraping and crawling service that implements the Firecrawl API. It provides a drop-in replacement for the Firecrawl /v1/scrape endpoint.

Status

⚠️ Not yet published to npm. The CI/CD workflow builds on every push and publishes automatically when a v* tag is pushed. Once the first release is cut:

npm install @jaco-tech/dsh-web-fetch-crw

Configure

Add to your DSH profile's cordis.patch.yml:

- insert:
    - id: web-fetch-crw
      name: '@jaco-tech/dsh-web-fetch-crw'
      config:
        baseURL: 'http://your-crw-instance:3000'

Optionally pin ctx.web to use this provider explicitly:

- replace:
    - id: web
      config:
        fetchProvider: crw

How it works

This is a Cordis plugin that implements the WebFetchProvider interface from @deepseek-ai/dsh-web:

export const name = "web-fetch-crw"
export const inject = ["web"]

export function apply(ctx, config) {
  ctx.web.registerFetchProvider({
    id: "crw",
    available() { /* URL.canParse(config.baseURL) */ },
    async fetch(request, signal) {
      // POST /v1/scrape { url, formats: ["markdown"] }
      // maps response to WebFetchResult (body as markdown text)
    }
  })
}

The existing @deepseek-ai/dsh-tool-web consumer calls ctx.web.fetch() — your crw instance serves the content automatically.

Provider interface

MethodBehavior
available()Returns true when baseURL is a valid URL
fetch(request, signal?)POST /v1/scrape with { url, formats: ["markdown"] }. Returns the scraped markdown as kind: "text" body (no double-conversion through turndown).

Upstream

This plugin communicates with crw via its Firecrawl-compatible API (POST /v1/scrape). crw is a self-hosted web scraping and crawling service that renders pages (via Lightpanda or Chromium) and produces clean markdown. You need a running crw instance to use this plugin.

Release process

# Tag and push — the CI workflow builds and publishes to npm automatically
git tag v0.1.0
git push origin v0.1.0

Requires the NPM_TOKEN secret to be set in the repository.

Development

npm install
npm run build

License

MIT