fix(embed): Go router vs the real Monad API [PRO-456] - #18
Open
Credgate wants to merge 6 commits into
Open
Conversation
…[PRO-455] Review of the Go port (#15) surfaced defects the TypeScript reference has too. Validated against the API source rather than the published spec, which is stale. - resolve a connector's pipeline via `GET /v1/{org}/{kind}s/{id}`'s `component_of` instead of scanning every pipeline. The scan sent no `limit`, so past the tenant's tenth pipeline it reported "not connected" for pipelines that exist. - `setEnabled` now PATCHes `{enabled}` alone. PATCH became a true partial update in api#2163; the read-modify-write it replaces silently dropped node `config_overrides` and edge `schema_detection_spec` on every toggle. - drain the connector list instead of one `limit=1000` request - send `type` on output create; `output_type` is deprecated (api#2156) - drop the speculative `data` / `config` response-shape fallbacks — Monad returns neither - map `kind` to its path segment explicitly rather than appending "s"
…of [PRO-455]
The mock returned every row regardless of `limit` and had no
`GET /v1/{org}/{kind}s/{id}` route at all, so it could not catch two bugs the
routers actually have: they never send `limit` (Monad defaults it to 10) and
they walk the whole pipeline list to answer a question `component_of` answers
in one call.
- page every list at Monad's default limit of 10, with no maximum
- serve the connector detail route, including `component_of` — pinned to the
same narrow projection the datastore emits (no `status`, no `nodes`)
- accept `type` on output create, `output_type` staying a deprecated alias
- track connectors statefully so a delete is observable
- add the >10-pipelines scenario; it fails against the pre-fix routers
Addresses the review on #15. Each item was validated against the API source rather than the published spec, which is ~15 months stale. - resolve a connector's pipeline via `GET /v1/{org}/{kind}s/{id}`'s `component_of` instead of scanning every pipeline. The scan sent no `limit`, and Monad defaults it to 10 — past a tenant's tenth pipeline the router reported "not connected" for pipelines that exist. Also turns an O(n) request fan-out into two calls. - `setEnabled` now PATCHes `{enabled}` alone. PATCH became a true partial update in api#2163; the read-modify-write it replaces rebuilt the whole graph from a partial field list, silently dropping node `config_overrides` and edge `schema_detection_spec` on every toggle. - drain the connector list rather than one `limit=1000` request - send `type` on output create; `output_type` is deprecated (api#2156) - drop the speculative `data` / `config` response-shape fallbacks — the list returns only `{pipelines, pagination}` and the detail is flat - `componentKind` type with an explicit path mapping, replacing `kind + "s"` - inline `url.PathEscape` instead of the one-letter `esc` alias - README: document `GetCustomerOrgID` as the tenant-isolation boundary
Credgate
requested review from
alastaird-monad,
andrewc-monad,
brenth-monad,
mattj-monad,
mon-aneesh and
saikapuluru99
as code owners
August 4, 2026 16:39
…[PRO-455] Review of the Go port (#15) surfaced defects the TypeScript reference has too. Validated against the API source rather than the published spec, which is stale. - resolve a connector's pipeline via `GET /v1/{org}/{kind}s/{id}`'s `component_of` instead of scanning every pipeline. The scan sent no `limit`, so past the tenant's tenth pipeline it reported "not connected" for pipelines that exist. - `setEnabled` now PATCHes `{enabled}` alone. PATCH became a true partial update in api#2163; the read-modify-write it replaces silently dropped node `config_overrides` and edge `schema_detection_spec` on every toggle. - drain the connector list instead of one `limit=1000` request - send `type` on output create; `output_type` is deprecated (api#2156) - drop the speculative `data` / `config` response-shape fallbacks — Monad returns neither - map `kind` to its path segment explicitly rather than appending "s"
# Conflicts: # routers/typescript/src/monad.ts
Credgate
force-pushed
the
feature/pro-456
branch
from
August 4, 2026 16:52
af89b91 to
f663124
Compare
Credgate
marked this pull request as draft
August 4, 2026 17:10
Credgate
marked this pull request as ready for review
August 4, 2026 18:40
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Addresses the CHANGES_REQUESTED review on #15. Eight of the ten comments were accurate, and three are live bugs rather than style nits. Every item was validated against the API source in
monad-inc/api, not the published spec —sdk/openapi.jsonis ~15 months stale and still marksUpdatePipelineRequestfields required.Second of three stacked fix PRs. Stacked on #17 (PRO-455) — it carries the mock changes this PR's conformance run depends on, so review/merge bottom-up. That's also why #17's commits appear in this diff; they drop out when it merges.
Linear: PRO-456
The three real bugs
Pipeline pagination.
GET /v2/{org}/pipelines/defaults tolimit=10(pipelines.go:727) and the router sent none. Past a tenant's tenth pipeline,GET /embed/pipelinesreportedhasPipeline: falsefor pipelines that exist.setEnabledwiped graph fields. PATCH became a true partial update in api#2163 (your commit). The read-modify-write it replaces rebuilt the whole graph from a partial field list and sent it back, so every enable/disable silently dropped nodeconfig_overridesand edgeschema_detection_spec. Now a one-linePATCH {"enabled": …}.component_of. You were right that this already exists —GET /v1/{org}/{kind}s/{id}returns it unconditionally, no expand flag (organization_inputs.go:79,109-124). It replaces the whole N+1 scan and fixes the pagination bug at the same time: N+1 requests → 2.The rest
typeinstead of the deprecatedoutput_type(api#2156); exhaustive paging on the connector list; the speculativedata/configfallbacks deleted (the list returns only{pipelines, pagination}, the detail is flat); acomponentKindtype with an explicit path mapping instead ofkind + "s";url.PathEscapeinline instead of theescalias — the escaping stays, it's a real boundary sinceconnectorIdcomes from the browser.On the unmarshal comment (
client.go:147): that one was accurate againste287e687, but the later force-push had already rewritten it tomap[string]json.RawMessage. No change needed — flagging so it doesn't read as ignored.On the
Next()pagination helper: kept internal rather than exposed. The/embedcontract has no pagination at all (zero matches forlimit|offset|cursor|pageinembed.openapi.yaml), so a public paginated type would need a contract amendment across all three routers and the browser client. Callers still get complete lists.Behavior change
An unknown connector id now returns
404 not_foundrather than200 {hasPipeline:false}, since resolution goes through a connector fetch.Verification
go vet,gofmt,go testclean (8 new tests coveringcomponent_ofresolution, the PATCH body,typeon create, and page draining). ConformanceROUTER=go35/35.Not yet run: live mode. Three things the mock cannot settle and I'd want confirmed against staging before merge — (a)
component_ofis populated on real responses, (b)PATCH {"enabled":false}leavesnodes/edges/nameintact on re-GET, (c) the real status code for an unknown connector id.responder.Errormasks to 500 unless the error implements bothErrorCodeandPublicError(responder.go:145-166), so the 404 mapping is inference, not verified fact.