apunkt/trustgraph
1
0
Fork 0
mirror of https://github.com/trustgraph-ai/trustgraph.git synced 2026-05-03 20:32:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Feature/configure flows (#345)

Browse source 
- Keeps processing in different flows separate so that data can go to different stores / collections etc.
- Potentially supports different processing flows
- Tidies the processing API with common base-classes for e.g. LLMs, and automatic configuration of 'clients' to use the right queue names in a flow
This commit is contained in:
cybermaggedon 2025-04-22 20:21:38 +01:00 • committed by GitHub
parent a06a814a41
commit a9197d11ee
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
125 changed files with 3751 additions and 2628 deletions
Download patch file Download diff file Expand all files Collapse all files

2
trustgraph-flow/trustgraph/query/doc_embeddings/milvus/service.py
View file

@ -11,7 +11,7 @@ from .... schema import document_embeddings_request_queue
from .... schema import document_embeddings_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "de-query"
default_input_queue = document_embeddings_request_queue
default_output_queue = document_embeddings_response_queue

2
trustgraph-flow/trustgraph/query/doc_embeddings/pinecone/service.py
View file

@ -16,7 +16,7 @@ from .... schema import document_embeddings_request_queue
from .... schema import document_embeddings_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "de-query"
default_input_queue = document_embeddings_request_queue
default_output_queue = document_embeddings_response_queue

70
trustgraph-flow/trustgraph/query/doc_embeddings/qdrant/service.py
View file

@ -7,71 +7,51 @@ 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 DocumentEmbeddingsResponse
from .... schema import Error, Value
from .... schema import document_embeddings_request_queue
from .... schema import document_embeddings_response_queue
from .... base import ConsumerProducer
from .... base import DocumentEmbeddingsQueryService
module = ".".join(__name__.split(".")[1:-1])
default_ident = "de-query"
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):
class Processor(DocumentEmbeddingsQueryService):
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)
#optional api key
api_key = params.get("api_key", None)
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,
"api_key": api_key,
}
)
self.client = QdrantClient(url=store_uri, api_key=api_key)
self.qdrant = QdrantClient(url=store_uri, api_key=api_key)
async def handle(self, msg):
async def query_document_embeddings(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:
for vec in msg.vectors:
dim = len(vec)
collection = (
"d_" + v.user + "_" + v.collection + "_" +
"d_" + msg.user + "_" + msg.collection + "_" +
str(dim)
)
search_result = self.client.query_points(
search_result = self.qdrant.query_points(
collection_name=collection,
query=vec,
limit=v.limit,
limit=msg.limit,
with_payload=True,
).points
@ -79,37 +59,17 @@ class Processor(ConsumerProducer):
ent = r.payload["doc"]
chunks.append(ent)
print("Send response...", flush=True)
r = DocumentEmbeddingsResponse(documents=chunks, error=None)
await self.send(r, properties={"id": id})
print("Done.", flush=True)
return chunks
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,
)
await self.send(r, properties={"id": id})
self.consumer.acknowledge(msg)
raise e
@staticmethod
def add_args(parser):
ConsumerProducer.add_args(
parser, default_input_queue, default_subscriber,
default_output_queue,
)
DocumentEmbeddingsQueryService.add_args(parser)
parser.add_argument(
'-t', '--store-uri',
@ -125,5 +85,5 @@ class Processor(ConsumerProducer):
def run():
Processor.launch(module, __doc__)
Processor.launch(default_ident, __doc__)

2
trustgraph-flow/trustgraph/query/graph_embeddings/milvus/service.py
View file

@ -11,7 +11,7 @@ from .... schema import graph_embeddings_request_queue
from .... schema import graph_embeddings_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "ge-query"
default_input_queue = graph_embeddings_request_queue
default_output_queue = graph_embeddings_response_queue

2
trustgraph-flow/trustgraph/query/graph_embeddings/pinecone/service.py
View file

@ -16,7 +16,7 @@ from .... schema import graph_embeddings_request_queue
from .... schema import graph_embeddings_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "ge-query"
default_input_queue = graph_embeddings_request_queue
default_output_queue = graph_embeddings_response_queue

72
trustgraph-flow/trustgraph/query/graph_embeddings/qdrant/service.py
View file

@ -7,44 +7,32 @@ entities
from qdrant_client import QdrantClient
from qdrant_client.models import PointStruct
from qdrant_client.models import Distance, VectorParams
import uuid
from .... schema import GraphEmbeddingsRequest, GraphEmbeddingsResponse
from .... schema import GraphEmbeddingsResponse
from .... schema import Error, Value
from .... schema import graph_embeddings_request_queue
from .... schema import graph_embeddings_response_queue
from .... base import ConsumerProducer
from .... base import GraphEmbeddingsQueryService
module = ".".join(__name__.split(".")[1:-1])
default_ident = "ge-query"
default_input_queue = graph_embeddings_request_queue
default_output_queue = graph_embeddings_response_queue
default_subscriber = module
default_store_uri = 'http://localhost:6333'
class Processor(ConsumerProducer):
class Processor(GraphEmbeddingsQueryService):
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)
#optional api key
api_key = params.get("api_key", None)
super(Processor, self).__init__(
**params | {
"input_queue": input_queue,
"output_queue": output_queue,
"subscriber": subscriber,
"input_schema": GraphEmbeddingsRequest,
"output_schema": GraphEmbeddingsResponse,
"store_uri": store_uri,
"api_key": api_key,
}
)
self.client = QdrantClient(url=store_uri, api_key=api_key)
self.qdrant = QdrantClient(url=store_uri, api_key=api_key)
def create_value(self, ent):
if ent.startswith("http://") or ent.startswith("https://"):
@ -52,34 +40,27 @@ class Processor(ConsumerProducer):
else:
return Value(value=ent, is_uri=False)
async def handle(self, msg):
async def query_graph_embeddings(self, msg):
try:
v = msg.value()
# Sender-produced ID
id = msg.properties()["id"]
print(f"Handling input {id}...", flush=True)
entity_set = set()
entities = []
for vec in v.vectors:
for vec in msg.vectors:
dim = len(vec)
collection = (
"t_" + v.user + "_" + v.collection + "_" +
"t_" + msg.user + "_" + msg.collection + "_" +
str(dim)
)
# Heuristic hack, get (2*limit), so that we have more chance
# of getting (limit) entities
search_result = self.client.query_points(
search_result = self.qdrant.query_points(
collection_name=collection,
query=vec,
limit=v.limit * 2,
limit=msg.limit * 2,
with_payload=True,
).points
@ -92,10 +73,10 @@ class Processor(ConsumerProducer):
entities.append(ent)
# Keep adding entities until limit
if len(entity_set) >= v.limit: break
if len(entity_set) >= msg.limit: break
# Keep adding entities until limit
if len(entity_set) >= v.limit: break
if len(entity_set) >= msg.limit: break
ents2 = []
@ -105,36 +86,19 @@ class Processor(ConsumerProducer):
entities = ents2
print("Send response...", flush=True)
r = GraphEmbeddingsResponse(entities=entities, error=None)
await self.send(r, properties={"id": id})
return entities
print("Done.", flush=True)
except Exception as e:
print(f"Exception: {e}")
print("Send error response...", flush=True)
r = GraphEmbeddingsResponse(
error=Error(
type = "llm-error",
message = str(e),
),
entities=None,
)
await self.send(r, properties={"id": id})
self.consumer.acknowledge(msg)
raise e
@staticmethod
def add_args(parser):
ConsumerProducer.add_args(
parser, default_input_queue, default_subscriber,
default_output_queue,
)
GraphEmbeddingsQueryService.add_args(parser)
parser.add_argument(
'-t', '--store-uri',
@ -150,5 +114,5 @@ class Processor(ConsumerProducer):
def run():
Processor.launch(module, __doc__)
Processor.launch(default_ident, __doc__)

121
trustgraph-flow/trustgraph/query/triples/cassandra/service.py
View file

@ -7,38 +7,24 @@ null. Output is a list of triples.
from .... direct.cassandra import TrustGraph
from .... schema import TriplesQueryRequest, TriplesQueryResponse, Error
from .... schema import Value, Triple
from .... schema import triples_request_queue
from .... schema import triples_response_queue
from .... base import ConsumerProducer
from .... base import TriplesQueryService
module = ".".join(__name__.split(".")[1:-1])
default_ident = "triples-query"
default_input_queue = triples_request_queue
default_output_queue = triples_response_queue
default_subscriber = module
default_graph_host='localhost'
class Processor(ConsumerProducer):
class Processor(TriplesQueryService):
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)
graph_host = params.get("graph_host", default_graph_host)
graph_username = params.get("graph_username", None)
graph_password = params.get("graph_password", None)
super(Processor, self).__init__(
**params | {
"input_queue": input_queue,
"output_queue": output_queue,
"subscriber": subscriber,
"input_schema": TriplesQueryRequest,
"output_schema": TriplesQueryResponse,
"graph_host": graph_host,
"graph_username": graph_username,
"graph_password": graph_password,
}
)
@ -53,92 +39,85 @@ class Processor(ConsumerProducer):
else:
return Value(value=ent, is_uri=False)
async def handle(self, msg):
async def query_triples(self, query):
try:
v = msg.value()
table = (v.user, v.collection)
table = (query.user, query.collection)
if table != self.table:
if self.username and self.password:
self.tg = TrustGraph(
hosts=self.graph_host,
keyspace=v.user, table=v.collection,
keyspace=query.user, table=query.collection,
username=self.username, password=self.password
)
else:
self.tg = TrustGraph(
hosts=self.graph_host,
keyspace=v.user, table=v.collection,
keyspace=query.user, table=query.collection,
)
self.table = table
# Sender-produced ID
id = msg.properties()["id"]
print(f"Handling input {id}...", flush=True)
triples = []
if v.s is not None:
if v.p is not None:
if v.o is not None:
if query.s is not None:
if query.p is not None:
if query.o is not None:
resp = self.tg.get_spo(
v.s.value, v.p.value, v.o.value,
limit=v.limit
query.s.value, query.p.value, query.o.value,
limit=query.limit
)
triples.append((v.s.value, v.p.value, v.o.value))
triples.append((query.s.value, query.p.value, query.o.value))
else:
resp = self.tg.get_sp(
v.s.value, v.p.value,
limit=v.limit
query.s.value, query.p.value,
limit=query.limit
)
for t in resp:
triples.append((v.s.value, v.p.value, t.o))
triples.append((query.s.value, query.p.value, t.o))
else:
if v.o is not None:
if query.o is not None:
resp = self.tg.get_os(
v.o.value, v.s.value,
limit=v.limit
query.o.value, query.s.value,
limit=query.limit
)
for t in resp:
triples.append((v.s.value, t.p, v.o.value))
triples.append((query.s.value, t.p, query.o.value))
else:
resp = self.tg.get_s(
v.s.value,
limit=v.limit
query.s.value,
limit=query.limit
)
for t in resp:
triples.append((v.s.value, t.p, t.o))
triples.append((query.s.value, t.p, t.o))
else:
if v.p is not None:
if v.o is not None:
if query.p is not None:
if query.o is not None:
resp = self.tg.get_po(
v.p.value, v.o.value,
limit=v.limit
query.p.value, query.o.value,
limit=query.limit
)
for t in resp:
triples.append((t.s, v.p.value, v.o.value))
triples.append((t.s, query.p.value, query.o.value))
else:
resp = self.tg.get_p(
v.p.value,
limit=v.limit
query.p.value,
limit=query.limit
)
for t in resp:
triples.append((t.s, v.p.value, t.o))
triples.append((t.s, query.p.value, t.o))
else:
if v.o is not None:
if query.o is not None:
resp = self.tg.get_o(
v.o.value,
limit=v.limit
query.o.value,
limit=query.limit
)
for t in resp:
triples.append((t.s, t.p, v.o.value))
triples.append((t.s, t.p, query.o.value))
else:
resp = self.tg.get_all(
limit=v.limit
limit=query.limit
)
for t in resp:
triples.append((t.s, t.p, t.o))
@ -152,37 +131,17 @@ class Processor(ConsumerProducer):
for t in triples
]
print("Send response...", flush=True)
r = TriplesQueryResponse(triples=triples, error=None)
await self.send(r, properties={"id": id})
print("Done.", flush=True)
return triples
except Exception as e:
print(f"Exception: {e}")
print("Send error response...", flush=True)
r = TriplesQueryResponse(
error=Error(
type = "llm-error",
message = str(e),
),
response=None,
)
await self.send(r, properties={"id": id})
self.consumer.acknowledge(msg)
raise e
@staticmethod
def add_args(parser):
ConsumerProducer.add_args(
parser, default_input_queue, default_subscriber,
default_output_queue,
)
TriplesQueryService.add_args(parser)
parser.add_argument(
'-g', '--graph-host',
@ -205,5 +164,5 @@ class Processor(ConsumerProducer):
def run():
Processor.launch(module, __doc__)
Processor.launch(default_ident, __doc__)

2
trustgraph-flow/trustgraph/query/triples/falkordb/service.py
View file

@ -13,7 +13,7 @@ from .... schema import triples_request_queue
from .... schema import triples_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "triples-query"
default_input_queue = triples_request_queue
default_output_queue = triples_response_queue

2
trustgraph-flow/trustgraph/query/triples/memgraph/service.py
View file

@ -13,7 +13,7 @@ from .... schema import triples_request_queue
from .... schema import triples_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "triples-query"
default_input_queue = triples_request_queue
default_output_queue = triples_response_queue

2
trustgraph-flow/trustgraph/query/triples/neo4j/service.py
View file

@ -13,7 +13,7 @@ from .... schema import triples_request_queue
from .... schema import triples_response_queue
from .... base import ConsumerProducer
module = ".".join(__name__.split(".")[1:-1])
module = "triples-query"
default_input_queue = triples_request_queue
default_output_queue = triples_response_queue