UN-4628 [FEAT] Global search: split adapters into LLM, vector DB, embedding and text extractor types

backend/global_search/README.md

@@ -66,7 +66,7 @@ changes.
 - `context` is a single display-only line, or `null`. Do not parse it.
 - `route.key` is a route-registry name, not a URL. `route.params` values are
   strings. It is one of the keys the table below lists for the result's
-  type; adapters have two (see below the table). A client that does not
+  type; the adapter types have two (see below the table). A client that does not
   know a key should render the result without a link, never guess a
   screen.
 - `type` is an open string. A client should render an unknown type with its
@@ -117,9 +117,26 @@ empty 200 (see the errors above).
 | 40 | `api_deployment` | API deployments | `api_deployment.detail` | `id` | parent workflow name |
 | 50 | `etl_pipeline` | ETL pipelines | `etl_pipeline.detail` | `id` | parent workflow name |
 | 60 | `task_pipeline` | Task pipelines | `task_pipeline.detail` | `id` | parent workflow name |
-| 80 | `adapter` | Adapters | `adapter.detail` / `adapter.info` | `id`, `adapter_type` | adapter type · provider |
+| 80 | `llm` | LLMs | `adapter.detail` / `adapter.info` | `id`, `adapter_type` | adapter type · provider |
+| 81 | `vector_db` | Vector DBs | `adapter.detail` / `adapter.info` | `id`, `adapter_type` | adapter type · provider |
+| 82 | `embedding` | Embedding | `adapter.detail` / `adapter.info` | `id`, `adapter_type` | adapter type · provider |
+| 83 | `text_extractor` | Text Extractor | `adapter.detail` / `adapter.info` | `id`, `adapter_type` | adapter type · provider |
 | 90 | `connector` | Connectors | `connector.detail` | `id` | provider |
 
+Adapters are split by kind into four types, one per rail destination
+(`adapter_type` `LLM`, `VECTOR_DB`, `EMBEDDING` and `X2TEXT`), each with its
+own `limit` and `has_more`. OCR adapters are not searchable. All four kinds
+share the `adapter.*` route keys (the notifications inbox sends `adapter.info`
+with the same params); `adapter_type` in the params says which adapter page to
+open.
+
+**Changed (UN-4628):** these four types replace the single `adapter` type
+(order 80, label "Adapters") that the endpoint returned from v0.194.0. A
+request for `types=adapter` now returns no adapter groups, since unknown keys
+are ignored, and OCR adapters, which `adapter` returned, are no longer
+returned at all. No released client used the `adapter` key when it was
+removed: the frontend panel (P2-T10) was not built yet.
+
 An adapter opens on `adapter.detail` only for a caller who may edit it
 (owner or co-owner, org admin, service account), the same rule as the
 adapter `retrieve` endpoint. Everyone else gets `adapter.info`, with the

backend/global_search/constants.py

@@ -41,5 +41,10 @@ class OSSSearchType:
     API_DEPLOYMENT = "api_deployment"
     ETL_PIPELINE = "etl_pipeline"
     TASK_PIPELINE = "task_pipeline"
-    ADAPTER = "adapter"
+    # Adapters split by kind, one type per rail destination. OCR adapters
+    # are not searchable: v1 drops OCR (decided 5 Nov 2026).
+    LLM = "llm"
+    VECTOR_DB = "vector_db"
+    EMBEDDING = "embedding"
+    TEXT_EXTRACTOR = "text_extractor"
     CONNECTOR = "connector"

backend/global_search/oss_providers.py

@@ -2,15 +2,16 @@
 
 Each provider starts from the same ``Model.objects.for_user(user)`` its list
 view uses, so search never returns a row the caller could not open from the
-list. Search may narrow further (pipeline type, No Op and frictionless
-adapters), never widen.
+list. Search may narrow further (pipeline type; adapter kind, which leaves
+OCR adapters out; No Op and frictionless adapters), never widen.
 """
 
 from __future__ import annotations
 
 import dataclasses
 from typing import Any
 
+from adapter_processor_v2.constants import AdapterKeys
 from adapter_processor_v2.models import AdapterInstance
 from api_v2.models import APIDeployment
 from connector_v2.models import ConnectorInstance
@@ -147,24 +148,39 @@ class TaskPipelineProvider(_PipelineProvider):
 
 
 class AdapterProvider(ModelSearchProvider):
-    type_key = OSSSearchType.ADAPTER
-    label = "Adapters"
-    order = 80
+    """Adapters of one kind (``adapter_type``), one search type per rail
+    destination. Not registered itself; the kinds below are.
+
+    All kinds share the ``adapter.detail`` / ``adapter.info`` route keys
+    (the notifications inbox sends ``adapter.info`` with the same params);
+    ``adapter_type`` in the route params tells the client which adapter page
+    to open.
+    """
+
+    adapter_type: str
     name_fields = ("adapter_name",)
     description_field = "description"
 
+    @property
+    def route_key(self) -> str:
+        return "adapter.detail"
+
+    @property
+    def info_route_key(self) -> str:
+        return "adapter.info"
+
     def visible_queryset(self, user: Any) -> QuerySet[Any]:
         # Frictionless adapters stay in the adapter list, where any member
         # may delete them, but nobody can open their detail view: retrieve
         # refuses them for everyone, admins included (IsFrictionLessAdapter).
-        return AdapterInstance.objects.for_user(user).exclude(
-            Q(adapter_id__startswith=NO_OP_ADAPTER_PREFIX) | Q(is_friction_less=True)
+        return (
+            AdapterInstance.objects.for_user(user)
+            .filter(adapter_type=self.adapter_type)
+            .exclude(
+                Q(adapter_id__startswith=NO_OP_ADAPTER_PREFIX) | Q(is_friction_less=True)
+            )
         )
 
-    @property
-    def info_route_key(self) -> str:
-        return f"{self.type_key}.info"
-
     def annotate(self, queryset: QuerySet[Any], user: Any) -> QuerySet[Any]:
         # Same rule as IsFrictionLessAdapter, which gates retrieve: service
         # accounts, org admins and owners (creator or co-owner). Anyone else
@@ -197,6 +213,34 @@ def route_params(self, obj: Model) -> dict[str, str]:
         return {"id": str(obj.pk), "adapter_type": obj.adapter_type}
 
 
+class LLMAdapterProvider(AdapterProvider):
+    type_key = OSSSearchType.LLM
+    label = "LLMs"
+    order = 80
+    adapter_type = AdapterKeys.LLM
+
+
+class VectorDBAdapterProvider(AdapterProvider):
+    type_key = OSSSearchType.VECTOR_DB
+    label = "Vector DBs"
+    order = 81
+    adapter_type = AdapterKeys.VECTOR_DB
+
+
+class EmbeddingAdapterProvider(AdapterProvider):
+    type_key = OSSSearchType.EMBEDDING
+    label = "Embedding"
+    order = 82
+    adapter_type = AdapterKeys.EMBEDDING
+
+
+class TextExtractorAdapterProvider(AdapterProvider):
+    type_key = OSSSearchType.TEXT_EXTRACTOR
+    label = "Text Extractor"
+    order = 83
+    adapter_type = AdapterKeys.X2TEXT
+
+
 class ConnectorProvider(ModelSearchProvider):
     type_key = OSSSearchType.CONNECTOR
     label = "Connectors"
@@ -216,6 +260,9 @@ def context(self, obj: Model) -> str | None:
     APIDeploymentProvider,
     ETLPipelineProvider,
     TaskPipelineProvider,
-    AdapterProvider,
+    LLMAdapterProvider,
+    VectorDBAdapterProvider,
+    EmbeddingAdapterProvider,
+    TextExtractorAdapterProvider,
     ConnectorProvider,
 )

backend/global_search/tests/base.py

@@ -64,6 +64,7 @@ def build(org, user, name, description=""):
 
 
 def build_adapter(org, user, name, description="", **overrides):
+    """An LLM adapter unless ``adapter_type`` is overridden."""
     fields = {
         "adapter_name": name,
         "adapter_id": "openai|test",
@@ -77,6 +78,13 @@ def build_adapter(org, user, name, description="", **overrides):
     return AdapterInstance.objects.create(**fields)
 
 
+def _adapter(adapter_type: str) -> Callable[..., AdapterInstance]:
+    def build(org, user, name, description=""):
+        return build_adapter(org, user, name, description, adapter_type=adapter_type)
+
+    return build
+
+
 def build_connector(org, user, name, description=""):
     return ConnectorInstance.objects.create(
         connector_name=name,
@@ -93,7 +101,10 @@ def build_connector(org, user, name, description=""):
     OSSSearchType.API_DEPLOYMENT: build_api_deployment,
     OSSSearchType.ETL_PIPELINE: _pipeline(Pipeline.PipelineType.ETL),
     OSSSearchType.TASK_PIPELINE: _pipeline(Pipeline.PipelineType.TASK),
-    OSSSearchType.ADAPTER: build_adapter,
+    OSSSearchType.LLM: _adapter("LLM"),
+    OSSSearchType.VECTOR_DB: _adapter("VECTOR_DB"),
+    OSSSearchType.EMBEDDING: _adapter("EMBEDDING"),
+    OSSSearchType.TEXT_EXTRACTOR: _adapter("X2TEXT"),
     OSSSearchType.CONNECTOR: build_connector,
 }
 
@@ -102,7 +113,18 @@ def build_connector(org, user, name, description=""):
     OSSSearchType.PROMPT_STUDIO_PROJECT,
     OSSSearchType.WORKFLOW,
     OSSSearchType.API_DEPLOYMENT,
-    OSSSearchType.ADAPTER,
+    OSSSearchType.LLM,
+    OSSSearchType.VECTOR_DB,
+    OSSSearchType.EMBEDDING,
+    OSSSearchType.TEXT_EXTRACTOR,
+}
+
+# The adapter kinds, keyed by type key.
+ADAPTER_TYPES = {
+    OSSSearchType.LLM: "LLM",
+    OSSSearchType.VECTOR_DB: "VECTOR_DB",
+    OSSSearchType.EMBEDDING: "EMBEDDING",
+    OSSSearchType.TEXT_EXTRACTOR: "X2TEXT",
 }
 
 

backend/global_search/tests/test_budget.py

@@ -39,7 +39,7 @@
 from workflow_manager.workflow_v2.models.workflow import Workflow
 
 from global_search.oss_providers import OSS_PROVIDERS
-from global_search.tests.base import search
+from global_search.tests.base import ADAPTER_TYPES, search
 
 ROWS_PER_TYPE = 2_000
 USERS = 200
@@ -139,13 +139,15 @@ def setUpTestData(cls) -> None:
             ),
             *AdapterInstance.objects.bulk_create(
                 AdapterInstance(
-                    adapter_name=_name("ad", i),
+                    adapter_name=_name(f"ad{kind}", i),
                     adapter_id="openai|seed",
-                    adapter_type="LLM",
+                    adapter_type=kind,
                     adapter_metadata={},
                     description=_name("d", i),
                     **common,
                 )
+                # Each adapter kind is its own search type.
+                for kind in ADAPTER_TYPES.values()
                 for i in n
             ),
             *ConnectorInstance.objects.bulk_create(

backend/global_search/tests/test_contract.py

@@ -26,7 +26,7 @@
 from global_search.constants import OSSSearchType as T
 from global_search.oss_providers import OSS_PROVIDERS, AdapterProvider, WorkflowProvider
 from global_search.providers import SearchProvider
-from global_search.tests.base import OSS_BUILDERS, search
+from global_search.tests.base import ADAPTER_TYPES, OSS_BUILDERS, search
 from global_search.views import GlobalSearchView
 
 APP_DIR = Path(__file__).parents[1]
@@ -66,19 +66,21 @@ def parent(type_key: str) -> str:
             T.API_DEPLOYMENT: parent(T.API_DEPLOYMENT),
             T.ETL_PIPELINE: parent(T.ETL_PIPELINE),
             T.TASK_PIPELINE: parent(T.TASK_PIPELINE),
-            T.ADAPTER: "LLM · openai",
+            **{key: f"{kind} · openai" for key, kind in ADAPTER_TYPES.items()},
             T.CONNECTOR: "minio",
         }
         for type_key, item in items.items():
             with self.subTest(type=type_key):
                 result = results[type_key]
                 self.assertEqual(result["context"], expected_context[type_key])
                 expected_params = {"id": str(item.pk)}
-                if type_key == T.ADAPTER:
-                    expected_params["adapter_type"] = "LLM"
+                route_key = f"{type_key}.detail"
+                if type_key in ADAPTER_TYPES:
+                    expected_params["adapter_type"] = ADAPTER_TYPES[type_key]
+                    route_key = "adapter.detail"
                 self.assertEqual(
                     result["route"],
-                    {"key": f"{type_key}.detail", "params": expected_params},
+                    {"key": route_key, "params": expected_params},
                 )
 
     def test_owner_context_skips_service_accounts_and_takes_the_earliest(self) -> None:

backend/global_search/tests/test_visibility.py

@@ -20,6 +20,7 @@
 
 from global_search.constants import OSSSearchType
 from global_search.tests.base import (
+    ADAPTER_TYPES,
     OSS_BUILDERS,
     _pipeline,
     build_adapter,
@@ -100,15 +101,13 @@ def test_frictionless_adapters_are_never_returned(self) -> None:
         )
         adapter.memberships.create(user=self.owner, role=ResourceRole.OWNER)
         for user in (self.owner, self.admin, self.service_account, self.outsider):
-            self.assertFalse(
-                self._found(user, OSSSearchType.ADAPTER, adapter), user.email
-            )
+            self.assertFalse(self._found(user, OSSSearchType.LLM, adapter), user.email)
 
     def test_adapter_route_follows_who_may_edit_it(self) -> None:
         # adapter.detail is gated like adapter retrieve: owners, co-owners,
         # admins and service accounts. Anyone else who can list the adapter
         # opens the read-only adapter.info.
-        adapter = self._private(OSSSearchType.ADAPTER)
+        adapter = self._private(OSSSearchType.LLM)
         set_resource_share_groups(adapter, [self.group.id])
         adapter.shared_to_org = True
         adapter.save()
@@ -123,9 +122,7 @@ def test_adapter_route_follows_who_may_edit_it(self) -> None:
         }
         for user, route_key in expected.items():
             with self.subTest(user=user.email):
-                response = search(
-                    user, q=QUERY, types=OSSSearchType.ADAPTER, limit=20
-                )
+                response = search(user, q=QUERY, types=OSSSearchType.LLM, limit=20)
                 (result,) = response.data["groups"][0]["results"]
                 self.assertEqual(
                     result["route"],
@@ -158,7 +155,35 @@ def test_no_op_adapters_are_never_returned(self) -> None:
             adapter_id="noOpLlm|f673a5a2-90f9-40f5-94c0-9fbc663b7553",
             shared_to_org=True,
         )
-        self.assertFalse(self._found(self.admin, OSSSearchType.ADAPTER, adapter))
+        self.assertFalse(self._found(self.admin, OSSSearchType.LLM, adapter))
+
+    def test_each_adapter_kind_lands_only_in_its_own_type(self) -> None:
+        # One group per rail destination: an adapter is found under its own
+        # kind and no other, and OCR adapters under none (v1 drops OCR).
+        adapters = {
+            type_key: build_adapter(
+                self.org,
+                self.owner,
+                f"{QUERY}-{type_key}",
+                adapter_type=adapter_type,
+                shared_to_org=True,
+            )
+            for type_key, adapter_type in ADAPTER_TYPES.items()
+        }
+        ocr = build_adapter(
+            self.org, self.owner, f"{QUERY}-ocr", adapter_type="OCR", shared_to_org=True
+        )
+        response = search(self.admin, q=QUERY, types=",".join(ADAPTER_TYPES), limit=20)
+        found = {
+            group["type"]: {result["id"] for result in group["results"]}
+            for group in response.data["groups"]
+        }
+        self.assertEqual(
+            found,
+            {key: {str(adapter.pk)} for key, adapter in adapters.items()},
+        )
+        everything = search(self.admin, q=QUERY, limit=20)
+        self.assertNotIn(str(ocr.pk), result_ids(everything))
 
     def test_default_and_app_pipelines_are_never_returned(self) -> None:
         for pipeline_type in (Pipeline.PipelineType.DEFAULT, Pipeline.PipelineType.APP):