CLI auth migration, document embeddings core lifecycle (#913)

Migrate get_kg_core and put_kg_core CLI tools to use Api/SocketClient
with first-frame auth (fixes broken raw websocket path). Fix wire
format field names (root/vector). Remove ~600 lines of dead raw
websocket code from invoke_graph_rag.py.

Add document embeddings core lifecycle to the knowledge service:
list/get/put/delete/load operations across schema, translator,
Cassandra table store, knowledge manager, gateway registry, REST API,
socket client, and CLI (tg-get-de-core, tg-put-de-core).

Fix delete_kg_core to also clean up document embeddings rows.

trustgraph-base/trustgraph/api/knowledge.py

@@ -132,3 +132,34 @@ class Knowledge:
 
         self.request(request = input)
 
+    def list_de_cores(self):
+
+        input = {
+            "operation": "list-de-cores",
+            "workspace": self.api.workspace,
+        }
+
+        return self.request(request = input)["ids"]
+
+    def delete_de_core(self, id):
+
+        input = {
+            "operation": "delete-de-core",
+            "workspace": self.api.workspace,
+            "id": id,
+        }
+
+        self.request(request = input)
+
+    def load_de_core(self, id, flow="default", collection="default"):
+
+        input = {
+            "operation": "load-de-core",
+            "workspace": self.api.workspace,
+            "id": id,
+            "flow": flow,
+            "collection": collection,
+        }
+
+        self.request(request = input)
+

trustgraph-base/trustgraph/api/socket_client.py

@@ -491,6 +491,58 @@ class SocketClient:
             triples=raw_triples,
         )
 
+    def get_kg_core(self, id: str) -> Iterator[Dict[str, Any]]:
+        request = {
+            "operation": "get-kg-core",
+            "workspace": self.workspace,
+            "id": id,
+        }
+        for response in self._send_request_sync(
+            "knowledge", None, request, streaming_raw=True,
+        ):
+            if response.get("eos"):
+                break
+            yield response
+
+    def put_kg_core(
+        self, id: str, triples=None, graph_embeddings=None,
+    ) -> Dict[str, Any]:
+        request = {
+            "operation": "put-kg-core",
+            "workspace": self.workspace,
+            "id": id,
+        }
+        if triples is not None:
+            request["triples"] = triples
+        if graph_embeddings is not None:
+            request["graph-embeddings"] = graph_embeddings
+        return self._send_request_sync("knowledge", None, request)
+
+    def get_de_core(self, id: str) -> Iterator[Dict[str, Any]]:
+        request = {
+            "operation": "get-de-core",
+            "workspace": self.workspace,
+            "id": id,
+        }
+        for response in self._send_request_sync(
+            "knowledge", None, request, streaming_raw=True,
+        ):
+            if response.get("eos"):
+                break
+            yield response
+
+    def put_de_core(
+        self, id: str, document_embeddings=None,
+    ) -> Dict[str, Any]:
+        request = {
+            "operation": "put-de-core",
+            "workspace": self.workspace,
+            "id": id,
+        }
+        if document_embeddings is not None:
+            request["document-embeddings"] = document_embeddings
+        return self._send_request_sync("knowledge", None, request)
+
     def close(self) -> None:
         """Close the persistent WebSocket connection."""
         if self._loop and not self._loop.is_closed():

trustgraph-base/trustgraph/messaging/translators/knowledge.py

@@ -1,6 +1,7 @@
 from typing import Dict, Any, Tuple, Optional
 from ...schema import (
     KnowledgeRequest, KnowledgeResponse, Triples, GraphEmbeddings,
+    DocumentEmbeddings, ChunkEmbeddings,
     Metadata, EntityEmbeddings
 )
 from .base import MessageTranslator
@@ -43,6 +44,23 @@ class KnowledgeRequestTranslator(MessageTranslator):
                 ]
             )
 
+        document_embeddings = None
+        if "document-embeddings" in data:
+            document_embeddings = DocumentEmbeddings(
+                metadata=Metadata(
+                    id=data["document-embeddings"]["metadata"]["id"],
+                    root=data["document-embeddings"]["metadata"].get("root", ""),
+                    collection=data["document-embeddings"]["metadata"]["collection"]
+                ),
+                chunks=[
+                    ChunkEmbeddings(
+                        chunk_id=ch["chunk_id"],
+                        vector=ch["vector"],
+                    )
+                    for ch in data["document-embeddings"]["chunks"]
+                ]
+            )
+
         return KnowledgeRequest(
             operation=data.get("operation"),
             id=data.get("id"),
@@ -50,6 +68,7 @@ class KnowledgeRequestTranslator(MessageTranslator):
             collection=data.get("collection"),
             triples=triples,
             graph_embeddings=graph_embeddings,
+            document_embeddings=document_embeddings,
         )
 
     def encode(self, obj: KnowledgeRequest) -> Dict[str, Any]:
@@ -90,6 +109,22 @@ class KnowledgeRequestTranslator(MessageTranslator):
                 ],
             }
 
+        if obj.document_embeddings:
+            result["document-embeddings"] = {
+                "metadata": {
+                    "id": obj.document_embeddings.metadata.id,
+                    "root": obj.document_embeddings.metadata.root,
+                    "collection": obj.document_embeddings.metadata.collection,
+                },
+                "chunks": [
+                    {
+                        "chunk_id": ch.chunk_id,
+                        "vector": ch.vector,
+                    }
+                    for ch in obj.document_embeddings.chunks
+                ],
+            }
+
         return result
 
 
@@ -140,6 +175,25 @@ class KnowledgeResponseTranslator(MessageTranslator):
                 }
             }
 
+        # Streaming document embeddings response
+        if obj.document_embeddings:
+            return {
+                "document-embeddings": {
+                    "metadata": {
+                        "id": obj.document_embeddings.metadata.id,
+                        "root": obj.document_embeddings.metadata.root,
+                        "collection": obj.document_embeddings.metadata.collection,
+                    },
+                    "chunks": [
+                        {
+                            "chunk_id": ch.chunk_id,
+                            "vector": ch.vector,
+                        }
+                        for ch in obj.document_embeddings.chunks
+                    ],
+                }
+            }
+
         # End of stream marker
         if obj.eos is True:
             return {"eos": True}
@@ -155,7 +209,7 @@ class KnowledgeResponseTranslator(MessageTranslator):
         is_final = (
             obj.ids is not None or  # List response
             obj.eos is True or      # End of stream
-            (not obj.triples and not obj.graph_embeddings)  # Empty response
+            (not obj.triples and not obj.graph_embeddings and not obj.document_embeddings)  # Empty response
         )
         
         return response, is_final

trustgraph-base/trustgraph/schema/knowledge/knowledge.py

@@ -4,7 +4,7 @@ from ..core.topic import queue
 from ..core.metadata import Metadata
 from .document import Document, TextDocument
 from .graph import Triples
-from .embeddings import GraphEmbeddings
+from .embeddings import GraphEmbeddings, DocumentEmbeddings
 
 # get-kg-core
 #   -> (???)
@@ -41,6 +41,9 @@ class KnowledgeRequest:
     triples: Triples | None = None
     graph_embeddings: GraphEmbeddings | None = None
 
+    # put-de-core
+    document_embeddings: DocumentEmbeddings | None = None
+
 @dataclass
 class KnowledgeResponse:
     error: Error | None = None
@@ -48,6 +51,7 @@ class KnowledgeResponse:
     eos: bool = False     # Indicates end of knowledge core stream
     triples: Triples | None = None
     graph_embeddings: GraphEmbeddings | None = None
+    document_embeddings: DocumentEmbeddings | None = None
 
 knowledge_request_queue = queue('knowledge', cls='request')
 knowledge_response_queue = queue('knowledge', cls='response')