microlinkhq / microlink / 34700801283

80%

push

github

feat(mcp): declare output schemas for every tool (#47)

* feat(mcp): declare output schemas for every tool

Mirror the TypeScript definitions shipped by the library
(packages/core/src/index.d.ts, packages/search/src/index.d.ts) as zod
schemas so clients can rely on structuredContent.data shapes:

- Asset for screenshot/pdf, Asset | null for logo/video/audio (the
  primary media fields are null when no asset is detected).
- Metadata, Embed, FunctionResult<unknown>, string content, string
  collections, records for extract/lighthouse, unknown[] technologies.
- Search pages as their JSON projection (lazy html()/markdown()/next()
  helpers never reach the wire); one tolerant shape per vertical.
- Index signatures map to catchall(unknown) so new API fields always
  validate.

Error results are exempt: the SDK skips output validation when isError
is true.

Signed-off-by: Kiko Beats <2096101+Kikobeats@users.noreply.github.com>

* fix(mcp): content output schemas accept null

markdown/html/text return null when the data rule matches nothing
(verified against the live API: data.markdown and data.text are null
on a non-matching selector). Collections are safe: the library
defaults them to [].

Signed-off-by: Kiko Beats <2096101+Kikobeats@users.noreply.github.com>

* fix(mcp): accept null embed and string lighthouse output

embed() returns data.iframe verbatim, so a successful oEmbed miss is
null. Lighthouse output html/csv is a string, not an object.

Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Signed-off-by: Kiko Beats <2096101+Kikobeats@users.noreply.github.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

652 of 844 branches covered (77.25%)

Branch coverage included in aggregate %.

117 of 117 new or added lines in 2 files covered. (100.0%)

3998 of 5052 relevant lines covered (79.14%)

25.21 hits per line