NuGet v3 endpoints

SharpDocs implements the minimum slice of the NuGet v3 protocol needed for dotnet restore and the Visual Studio package manager to consume a folder of nupkgs.

All routes are under /v3 and served by NuGetController.

GET /v3/index.json

Service index. Advertises the resources below. The @id values are built from the incoming request, so the feed works behind reverse proxies without configuration.

Resources advertised:

• PackageBaseAddress/3.0.0 — flatcontainer
• RegistrationsBaseUrl/3.6.0 and /Versioned
• SearchQueryService (+ 3.0.0-beta, 3.0.0-rc)
• SearchAutocompleteService (+ 3.0.0-rc)

`GET /v3/flatcontainer/

List of versions for a package id. Returns 404 when the id is unknown.

{ "versions": ["1.0.0", "1.1.0", "2.0.0"] }

`GET /v3/flatcontainer/

Downloads the nupkg. The file segment must match {id}.{version} exactly (case-insensitive).

`GET /v3/flatcontainer/

Extracts and returns the nuspec from the nupkg.

`GET /v3/registration/

Registration index. Returns a single page with one leaf per version. Each leaf includes:

• catalogEntry with id, version, description, authors, tags, license expression, project URL, dependency groups
• packageContent URL pointing at the flatcontainer nupkg

GET /v3/search

Query parameters:

• q — optional, matched (case-insensitive) against id, description, and tags
• skip — default 0
• take — default 20
• prerelease — accepted but currently ignored (all packages are listed)

Returns the latest version per id, paged.

GET /v3/autocomplete

Query parameters:

• q — optional substring filter on id
• take — default 20

Returns a list of package ids.

Composite mode

If one or more upstream feeds are configured via AddUpstream(...), every endpoint above merges local data with results from the upstreams. Responses are protocol-compatible:

• Version lists and search results include upstream packages alongside local ones.
• Registration packageContent URLs and search @id/registration URLs are rewritten to point back at this feed, so clients always resolve through you.
• .nupkg and .nuspec downloads are proxy-streamed from the upstream when the package isn't local. No on-disk cache.
• Local packages win on id+version collisions.

See NuGet feed for configuration.

GET /api/manifest

Sibling endpoint (not under /v3) that returns a JSON summary of the site for tooling and integrations. The shape varies by mode.

Single-project:

{
  "title": "My library",
  "description": "...",
  "primaryPackageId": "MyLibrary",
  "pageCount": 12
}

Multi-project:

{
  "title": "Codezerg Libraries",
  "description": "...",
  "projects": [
    {
      "id": "alpha",
      "title": "Alpha",
      "description": "...",
      "primaryPackageId": "Codezerg.Alpha",
      "pageCount": 9,
      "error": null
    },
    {
      "id": "beta",
      "title": "beta",
      "description": null,
      "primaryPackageId": null,
      "pageCount": 0,
      "error": "Per-project sharpdocs.json not found: ..."
    }
  ]
}

primaryPackageId is the lone package in the project's artifacts folder if there's exactly one, or — if there are several — the one whose id matches the project id (case-insensitive). Otherwise null. error is non-null on degraded projects.

What isn't implemented

• Pushing packages (no write endpoints — drop files into ArtifactsRoot and restart)
• Catalog, symbols server, report abuse, deprecated listing
• Multi-page registration responses (everything fits in one page; huge version histories would need sharding)
• API key authentication (the feed is read-only — put it behind a reverse proxy if you need access control)