Fix/doc streaming proto (#673)

* Librarian streaming doc download

* Document stream download endpoint

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

@@ -174,4 +174,6 @@ class LibraryResponseTranslator(MessageTranslator):
     
     def from_response_with_completion(self, obj: LibrarianResponse) -> Tuple[Dict[str, Any], bool]:
         """Returns (response_dict, is_final)"""
-        return self.from_pulsar(obj), True
+        # For streaming responses, check end_of_stream to determine if this is the final message
+        is_final = getattr(obj, 'end_of_stream', True)
+        return self.from_pulsar(obj), is_final

trustgraph-base/trustgraph/schema/services/library.py

@@ -212,6 +212,9 @@ class LibrarianResponse:
     # list-uploads response
     upload_sessions: list[UploadSession] = field(default_factory=list)
 
+    # stream-document response - indicates final chunk in stream
+    end_of_stream: bool = False
+
 # FIXME: Is this right?  Using persistence on librarian so that
 # message chunking works
 

trustgraph-flow/trustgraph/gateway/dispatch/document_stream.py

@@ -0,0 +1,65 @@
+
+import asyncio
+import uuid
+import logging
+from . librarian import LibrarianRequestor
+
+# Module logger
+logger = logging.getLogger(__name__)
+
+class DocumentStreamExport:
+
+    def __init__(self, backend):
+        self.backend = backend
+
+    async def process(self, data, error, ok, request):
+
+        user = request.query.get("user")
+        document_id = request.query.get("document-id")
+        chunk_size = int(request.query.get("chunk-size", 1024 * 1024))
+
+        if not user or not document_id:
+            return await error("Missing required parameters: user, document-id")
+
+        response = await ok()
+
+        lr = LibrarianRequestor(
+            backend=self.backend,
+            consumer="api-gateway-doc-stream-" + str(uuid.uuid4()),
+            subscriber="api-gateway-doc-stream-" + str(uuid.uuid4()),
+        )
+
+        try:
+
+            await lr.start()
+
+            async def responder(resp, fin):
+                if "content" in resp:
+                    content = resp["content"]
+                    # Content is base64 encoded, write as-is for client to decode
+                    # Or decode here and write raw bytes
+                    import base64
+                    chunk_data = base64.b64decode(content)
+                    await response.write(chunk_data)
+
+            await lr.process(
+                {
+                    "operation": "stream-document",
+                    "user": user,
+                    "document-id": document_id,
+                    "chunk-size": chunk_size,
+                },
+                responder
+            )
+
+        except Exception as e:
+
+            logger.error(f"Document stream exception: {e}", exc_info=True)
+
+        finally:
+
+            await lr.stop()
+
+        await response.write_eof()
+
+        return response

trustgraph-flow/trustgraph/gateway/dispatch/manager.py

@@ -45,6 +45,7 @@ from . rows_import import RowsImport
 
 from . core_export import CoreExport
 from . core_import import CoreImport
+from . document_stream import DocumentStreamExport
 
 from . mux import Mux
 
@@ -135,6 +136,14 @@ class DispatcherManager:
     def dispatch_core_import(self):
         return DispatcherWrapper(self.process_core_import)
 
+    def dispatch_document_stream(self):
+        return DispatcherWrapper(self.process_document_stream)
+
+    async def process_document_stream(self, data, error, ok, request):
+
+        ds = DocumentStreamExport(self.backend)
+        return await ds.process(data, error, ok, request)
+
     async def process_core_import(self, data, error, ok, request):
 
         ci = CoreImport(self.backend)

trustgraph-flow/trustgraph/gateway/endpoint/manager.py

@@ -64,6 +64,12 @@ class EndpointManager:
                 method = "GET",
                 dispatcher = dispatcher_manager.dispatch_core_export(),
             ),
+            StreamEndpoint(
+                endpoint_path = "/api/v1/document-stream",
+                auth = auth,
+                method = "GET",
+                dispatcher = dispatcher_manager.dispatch_document_stream(),
+            ),
         ]
 
     def add_routes(self, app):

trustgraph-flow/trustgraph/librarian/librarian.py

@@ -654,14 +654,13 @@ class Librarian:
         """
         Stream document content in chunks.
 
-        This operation returns document content in smaller chunks, allowing
+        This is an async generator that yields document content in smaller chunks,
-        memory-efficient processing of large documents. The response includes
+        allowing memory-efficient processing of large documents. Each yielded
-        chunk information for reassembly.
+        response includes chunk information and an end_of_stream flag.
 
-        Note: This operation returns a single chunk at a time. Clients should
+        The final chunk will have end_of_stream=True.
-        call repeatedly with increasing chunk_index until all chunks are received.
         """
-        logger.debug(f"Streaming document {request.document_id}, chunk {request.chunk_index}")
+        logger.debug(f"Streaming document {request.document_id}")
 
         DEFAULT_CHUNK_SIZE = 1024 * 1024  # 1MB default
 
@@ -680,29 +679,29 @@ class Librarian:
         total_size = await self.blob_store.get_size(object_id)
         total_chunks = math.ceil(total_size / chunk_size)
 
-        if request.chunk_index >= total_chunks:
+        # Stream all chunks
-            raise RequestError(
+        for chunk_index in range(total_chunks):
-                f"Invalid chunk index {request.chunk_index}, "
-                f"document has {total_chunks} chunks"
-            )
-
             # Calculate byte range
-        offset = request.chunk_index * chunk_size
+            offset = chunk_index * chunk_size
             length = min(chunk_size, total_size - offset)
 
             # Fetch only the requested range
             chunk_content = await self.blob_store.get_range(object_id, offset, length)
 
-        logger.debug(f"Returning chunk {request.chunk_index}/{total_chunks}, "
+            is_last_chunk = (chunk_index == total_chunks - 1)
-                    f"bytes {offset}-{offset + length} of {total_size}")
 
-        return LibrarianResponse(
+            logger.debug(f"Streaming chunk {chunk_index}/{total_chunks}, "
+                        f"bytes {offset}-{offset + length} of {total_size}, "
+                        f"end_of_stream={is_last_chunk}")
+
+            yield LibrarianResponse(
                 error=None,
                 content=base64.b64encode(chunk_content),
-            chunk_index=request.chunk_index,
+                chunk_index=chunk_index,
-            chunks_received=1,  # Using as "current chunk" indicator
+                chunks_received=chunk_index + 1,
                 total_chunks=total_chunks,
                 bytes_received=offset + length,
                 total_bytes=total_size,
+                end_of_stream=is_last_chunk,
             )
 

trustgraph-flow/trustgraph/librarian/service.py

@@ -504,6 +504,15 @@ class Processor(AsyncProcessor):
 
         try:
 
+            # Handle streaming operations specially
+            if v.operation == "stream-document":
+                async for resp in self.librarian.stream_document(v):
+                    await self.librarian_response_producer.send(
+                        resp, properties={"id": id}
+                    )
+                return
+
+            # Non-streaming operations
             resp = await self.process_request(v)
 
             await self.librarian_response_producer.send(
@@ -517,7 +526,8 @@ class Processor(AsyncProcessor):
                 error = Error(
                     type = "request-error",
                     message = str(e),
-                )
+                ),
+                end_of_stream = True,
             )
 
             await self.librarian_response_producer.send(
@@ -530,7 +540,8 @@ class Processor(AsyncProcessor):
                 error = Error(
                     type = "unexpected-error",
                     message = str(e),
-                )
+                ),
+                end_of_stream = True,
             )
 
             await self.librarian_response_producer.send(