Currently docs-mcp-server isn't ecosystem aware.
This can be a bit of a problem when registering a library such as pandoc. Is that the original haskell release? The Python wrapper SDK called pandoc? The nodejs wrapper SDK also called pandoc?
It should be possible to set a library as a specific language/ecosystem or just leave empty/generic.
Otherwise I'll need to adopt the practice of <ecosystem>-<actual name> where I know there's clashing names.
Going further, for ecosystems with standard public documentation like Go with https://pkg.go.dev/ (related #149) or Rust with https://docs.rs/ it could have automatic identification of the docs location.
Either the user selects Rust then types something like hashbrown or maybe a link to the crates.io page. In the case of Go they either put the full package name (pretty much a link without https:// or maybe they could put the repo, it identifies the package name from go.mod and then finds the pkg.go.dev page.
It could also be done for ecosystems like NPM or PyPi if it was able to identify the README of the package -> Mention of docs -> Identified high quality documentation website/location.
When it has a URL identified the user can then either continue or adjust settings as they typically would like depth, ignores, etc.
(This second part could be split into it's own issue.)
Currently docs-mcp-server isn't ecosystem aware.
This can be a bit of a problem when registering a library such as
pandoc. Is that the original haskell release? The Python wrapper SDK calledpandoc? The nodejs wrapper SDK also calledpandoc?It should be possible to set a library as a specific language/ecosystem or just leave empty/generic.
Otherwise I'll need to adopt the practice of
<ecosystem>-<actual name>where I know there's clashing names.Going further, for ecosystems with standard public documentation like Go with https://pkg.go.dev/ (related #149) or Rust with https://docs.rs/ it could have automatic identification of the docs location.
Either the user selects
Rustthen types something likehashbrownor maybe a link to the crates.io page. In the case of Go they either put the full package name (pretty much a link withouthttps://or maybe they could put the repo, it identifies the package name fromgo.modand then finds the pkg.go.dev page.It could also be done for ecosystems like NPM or PyPi if it was able to identify the README of the package -> Mention of docs -> Identified high quality documentation website/location.
When it has a URL identified the user can then either continue or adjust settings as they typically would like depth, ignores, etc.
(This second part could be split into it's own issue.)