Template rejig (#48)

* document-rag / graph-rag refactor of templates

* Tweaking the docs and categories

* Clarify triple store vs RAG

* Tweak knowledge graph linkage

* Doc embedding for Qdrant

* Fix document RAG on Qdrant

* Fix templates

* Bump version

* Updated templates

trustgraph/query/doc_embeddings/qdrant/__main__.py

@@ -0,0 +1,7 @@
+#!/usr/bin/env python3
+
+from . hf import run
+
+if __name__ == '__main__':
+    run()
+

trustgraph/query/doc_embeddings/qdrant/service.py

@@ -0,0 +1,117 @@
+
+"""
+Document embeddings query service.  Input is vector, output is an array
+of chunks
+"""
+
+from qdrant_client import QdrantClient
+from qdrant_client.models import PointStruct
+from qdrant_client.models import Distance, VectorParams
+import uuid
+
+from .... schema import DocumentEmbeddingsRequest, DocumentEmbeddingsResponse
+from .... schema import Error, Value
+from .... schema import document_embeddings_request_queue
+from .... schema import document_embeddings_response_queue
+from .... base import ConsumerProducer
+
+module = ".".join(__name__.split(".")[1:-1])
+
+default_input_queue = document_embeddings_request_queue
+default_output_queue = document_embeddings_response_queue
+default_subscriber = module
+default_store_uri = 'http://localhost:6333'
+
+class Processor(ConsumerProducer):
+
+    def __init__(self, **params):
+
+        input_queue = params.get("input_queue", default_input_queue)
+        output_queue = params.get("output_queue", default_output_queue)
+        subscriber = params.get("subscriber", default_subscriber)
+        store_uri = params.get("store_uri", default_store_uri)
+
+        super(Processor, self).__init__(
+            **params | {
+                "input_queue": input_queue,
+                "output_queue": output_queue,
+                "subscriber": subscriber,
+                "input_schema": DocumentEmbeddingsRequest,
+                "output_schema": DocumentEmbeddingsResponse,
+                "store_uri": store_uri,
+            }
+        )
+
+        self.client = QdrantClient(url=store_uri)
+
+    def handle(self, msg):
+
+        try:
+
+            v = msg.value()
+
+            # Sender-produced ID
+            id = msg.properties()["id"]
+
+            print(f"Handling input {id}...", flush=True)
+
+            chunks = []
+
+            for vec in v.vectors:
+
+                dim = len(vec)
+                collection = "doc_" + str(dim)
+
+                search_result = self.client.query_points(
+                    collection_name=collection,
+                    query=vec,
+                    limit=v.limit,
+                    with_payload=True,
+                ).points
+
+                for r in search_result:
+                    ent = r.payload["doc"]
+                    chunks.append(ent)
+
+            print("Send response...", flush=True)
+            r = DocumentEmbeddingsResponse(documents=chunks, error=None)
+            self.producer.send(r, properties={"id": id})
+
+            print("Done.", flush=True)
+
+        except Exception as e:
+
+            print(f"Exception: {e}")
+
+            print("Send error response...", flush=True)
+
+            r = DocumentEmbeddingsResponse(
+                error=Error(
+                    type = "llm-error",
+                    message = str(e),
+                ),
+                documents=None,
+            )
+
+            self.producer.send(r, properties={"id": id})
+
+            self.consumer.acknowledge(msg)
+
+    @staticmethod
+    def add_args(parser):
+
+        ConsumerProducer.add_args(
+            parser, default_input_queue, default_subscriber,
+            default_output_queue,
+        )
+
+        parser.add_argument(
+            '-t', '--store-uri',
+            default=default_store_uri,
+            help=f'Milvus store URI (default: {default_store_uri})'
+        )
+
+def run():
+
+    Processor.start(module, __doc__)
+

trustgraph/retrieval/document_rag/rag.py

@@ -29,8 +29,6 @@ class Processor(ConsumerProducer):
         input_queue = params.get("input_queue", default_input_queue)
         output_queue = params.get("output_queue", default_output_queue)
         subscriber = params.get("subscriber", default_subscriber)
-        entity_limit = params.get("entity_limit", 50)
-        triple_limit = params.get("triple_limit", 30)
         pr_request_queue = params.get(
             "prompt_request_queue", prompt_request_queue
         )
@@ -105,7 +103,7 @@ class Processor(ConsumerProducer):
 
             print("Send error response...", flush=True)
 
-            r = GraphRagResponse(
+            r = DocumentRagResponse(
                 error=Error(
                     type = "llm-error",
                     message = str(e),
@@ -125,33 +123,6 @@ class Processor(ConsumerProducer):
             default_output_queue,
         )
 
-        parser.add_argument(
-            '-v', '--vector-store',
-            default='http://milvus:19530',
-            help=f'Vector host (default: http://milvus:19530)'
-        )
-
-        parser.add_argument(
-            '-e', '--entity-limit',
-            type=int,
-            default=50,
-            help=f'Entity vector fetch limit (default: 50)'
-        )
-
-        parser.add_argument(
-            '-t', '--triple-limit',
-            type=int,
-            default=30,
-            help=f'Triple query limit, per query (default: 30)'
-        )
-
-        parser.add_argument(
-            '-u', '--max-subgraph-size',
-            type=int,
-            default=3000,
-            help=f'Max subgraph size (default: 3000)'
-        )
-
         parser.add_argument(
             '--prompt-request-queue',
             default=prompt_request_queue,

trustgraph/retrieval/graph_rag/rag.py

@@ -142,18 +142,6 @@ class Processor(ConsumerProducer):
             default_output_queue,
         )
 
-        parser.add_argument(
-            '-g', '--graph-hosts',
-            default='cassandra',
-            help=f'Graph hosts, comma separated (default: cassandra)'
-        )
-
-        parser.add_argument(
-            '-v', '--vector-store',
-            default='http://milvus:19530',
-            help=f'Vector host (default: http://milvus:19530)'
-        )
-
         parser.add_argument(
             '-e', '--entity-limit',
             type=int,

trustgraph/storage/doc_embeddings/qdrant/__init__.py

@@ -0,0 +1,3 @@
+
+from . write import *
+

trustgraph/storage/doc_embeddings/qdrant/__main__.py

@@ -0,0 +1,7 @@
+#!/usr/bin/env python3
+
+from . write import run
+
+if __name__ == '__main__':
+    run()
+

trustgraph/storage/doc_embeddings/qdrant/write.py

@@ -0,0 +1,104 @@
+
+"""
+Accepts entity/vector pairs and writes them to a Qdrant store.
+"""
+
+from qdrant_client import QdrantClient
+from qdrant_client.models import PointStruct
+from qdrant_client.models import Distance, VectorParams
+import uuid
+
+from .... schema import ChunkEmbeddings
+from .... schema import chunk_embeddings_ingest_queue
+from .... log_level import LogLevel
+from .... base import Consumer
+
+module = ".".join(__name__.split(".")[1:-1])
+
+default_input_queue = chunk_embeddings_ingest_queue
+default_subscriber = module
+default_store_uri = 'http://localhost:6333'
+
+class Processor(Consumer):
+
+    def __init__(self, **params):
+
+        input_queue = params.get("input_queue", default_input_queue)
+        subscriber = params.get("subscriber", default_subscriber)
+        store_uri = params.get("store_uri", default_store_uri)
+
+        super(Processor, self).__init__(
+            **params | {
+                "input_queue": input_queue,
+                "subscriber": subscriber,
+                "input_schema": ChunkEmbeddings,
+                "store_uri": store_uri,
+            }
+        )
+
+        self.last_collection = None
+        self.last_dim = None
+
+        self.client = QdrantClient(url=store_uri)
+
+    def handle(self, msg):
+
+        v = msg.value()
+
+        chunk = v.chunk.decode("utf-8")
+
+        if chunk == "": return
+
+        for vec in v.vectors:
+
+            dim = len(vec)
+            collection = "doc_" + str(dim)
+
+            if dim != self.last_dim:
+
+                if not self.client.collection_exists(collection):
+
+                    try:
+                        self.client.create_collection(
+                            collection_name=collection,
+                            vectors_config=VectorParams(
+                                size=dim, distance=Distance.DOT
+                            ),
+                        )
+                    except Exception as e:
+                        print("Qdrant collection creation failed")
+                        raise e
+
+                self.last_collection = collection
+                self.last_dim = dim
+
+            self.client.upsert(
+                collection_name=collection,
+                points=[
+                    PointStruct(
+                        id=str(uuid.uuid4()),
+                        vector=vec,
+                        payload={
+                            "doc": chunk,
+                        }
+                    )
+                ]
+            )
+
+    @staticmethod
+    def add_args(parser):
+
+        Consumer.add_args(
+            parser, default_input_queue, default_subscriber,
+        )
+
+        parser.add_argument(
+            '-t', '--store-uri',
+            default=default_store_uri,
+            help=f'Qdrant store URI (default: {default_store_uri})'
+        )
+
+def run():
+
+    Processor.start(module, __doc__)
+