- Added graph_embeddings_client.py

- Added tests/test-graph-embeddings to demonstrate client works
- Fixed schema
This commit is contained in:
Cyber MacGeddon 2024-08-12 17:35:48 +01:00
parent e2896f6433
commit 6de72d7567
4 changed files with 134 additions and 8 deletions

23
tests/test-graph-embeddings Executable file
View file

@ -0,0 +1,23 @@
#!/usr/bin/env python3
import pulsar
from trustgraph.graph_embeddings_client import GraphEmbeddingsClient
from trustgraph.embeddings_client import EmbeddingsClient
ec = EmbeddingsClient(pulsar_host="pulsar://localhost:6650")
vectors = ec.request("What caused the space shuttle to explode?")
print(vectors)
llm = GraphEmbeddingsClient(pulsar_host="pulsar://localhost:6650")
limit=10
resp = llm.request(vectors, limit)
print("Response...")
for val in resp:
print(val.value)

View file

@ -0,0 +1,89 @@
#!/usr/bin/env python3
import pulsar
import _pulsar
from pulsar.schema import JsonSchema
import hashlib
import uuid
from . schema import GraphEmbeddingsRequest, GraphEmbeddingsResponse
from . schema import graph_embeddings_request_queue
from . schema import graph_embeddings_response_queue
# Ugly
ERROR=_pulsar.LoggerLevel.Error
WARN=_pulsar.LoggerLevel.Warn
INFO=_pulsar.LoggerLevel.Info
DEBUG=_pulsar.LoggerLevel.Debug
class GraphEmbeddingsClient:
def __init__(
self, log_level=ERROR,
subscriber=None,
input_queue=None,
output_queue=None,
pulsar_host="pulsar://pulsar:6650",
):
if input_queue == None:
input_queue = graph_embeddings_request_queue
if output_queue == None:
output_queue = graph_embeddings_response_queue
if subscriber == None:
subscriber = str(uuid.uuid4())
self.client = pulsar.Client(
pulsar_host,
logger=pulsar.ConsoleLogger(log_level),
)
self.producer = self.client.create_producer(
topic=input_queue,
schema=JsonSchema(GraphEmbeddingsRequest),
chunking_enabled=True,
)
self.consumer = self.client.subscribe(
output_queue, subscriber,
schema=JsonSchema(GraphEmbeddingsResponse),
)
def request(self, vectors, limit=10, timeout=500):
id = str(uuid.uuid4())
r = GraphEmbeddingsRequest(
vectors=vectors,
limit=limit,
)
self.producer.send(r, properties={ "id": id })
while True:
msg = self.consumer.receive(timeout_millis=timeout * 1000)
mid = msg.properties()["id"]
if mid == id:
resp = msg.value().entities
self.consumer.acknowledge(msg)
return resp
# Ignore messages with wrong ID
self.consumer.acknowledge(msg)
def __del__(self):
if hasattr(self, "consumer"):
self.consumer.close()
if hasattr(self, "producer"):
self.producer.flush()
self.producer.close()
self.client.close()

View file

@ -4,11 +4,11 @@ Graph embeddings query service. Input is vector, output is list of
e
"""
from ... direct.milvus import TripleVectors
from ... schema import GraphEmbeddingsRequest, GraphEmbeddingsResponse
from ... schema import graph_embeddings_request_queue
from ... schema import graph_embeddings_response_queue
from ... base import ConsumerProducer
from .... direct.milvus import TripleVectors
from .... schema import GraphEmbeddingsRequest, GraphEmbeddingsResponse, Value
from .... schema import graph_embeddings_request_queue
from .... schema import graph_embeddings_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
@ -55,7 +55,21 @@ class Processor(ConsumerProducer):
resp = self.vecstore.search(vec, limit=v.limit)
for r in resp:
entities.add(r)
ent = r["entity"]["entity"]
entities.add(ent)
# Convert set to list
entities = list(entities)
ents2 = []
for ent in entities:
if ent.startswith("http://") or ent.startswith("https://"):
ents2.append(Value(value=ent, is_uri=True))
else:
ents2.append(Value(value=ent, is_uri=False))
entities = ents2
print("Send response...", flush=True)
r = GraphEmbeddingsResponse(entities=entities)

View file

@ -73,11 +73,11 @@ graph_embeddings_store_queue = topic('graph-embeddings-store')
# Graph embeddings query
class GraphEmbeddingsQueryRequest(Record):
class GraphEmbeddingsRequest(Record):
vectors = Array(Array(Double()))
limit = Integer()
class GraphEmbeddingsQueryResponse(Record):
class GraphEmbeddingsResponse(Record):
entities = Array(Value())
graph_embeddings_request_queue = topic(