mirror of
https://github.com/trustgraph-ai/trustgraph.git
synced 2026-07-02 22:41:01 +02:00
f2ae0e8623
- Put scores in all responses - Remove unused 'middle' vector layer. Vector of texts -> vector of (vector embedding)
113 lines
3 KiB
Python
Executable file
113 lines
3 KiB
Python
Executable file
|
|
"""
|
|
Graph embeddings, calls the embeddings service to get embeddings for a
|
|
set of entity contexts. Input is entity plus textual context.
|
|
Output is entity plus embedding.
|
|
"""
|
|
|
|
from ... schema import EntityContexts, EntityEmbeddings, GraphEmbeddings
|
|
from ... schema import EmbeddingsRequest, EmbeddingsResponse
|
|
|
|
from ... base import FlowProcessor, EmbeddingsClientSpec, ConsumerSpec
|
|
from ... base import ProducerSpec
|
|
|
|
import logging
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
default_ident = "graph-embeddings"
|
|
default_batch_size = 5
|
|
|
|
class Processor(FlowProcessor):
|
|
|
|
def __init__(self, **params):
|
|
|
|
id = params.get("id")
|
|
self.batch_size = params.get("batch_size", default_batch_size)
|
|
|
|
super(Processor, self).__init__(
|
|
**params | {
|
|
"id": id,
|
|
}
|
|
)
|
|
|
|
self.register_specification(
|
|
ConsumerSpec(
|
|
name = "input",
|
|
schema = EntityContexts,
|
|
handler = self.on_message,
|
|
)
|
|
)
|
|
|
|
self.register_specification(
|
|
EmbeddingsClientSpec(
|
|
request_name = "embeddings-request",
|
|
response_name = "embeddings-response",
|
|
)
|
|
)
|
|
|
|
self.register_specification(
|
|
ProducerSpec(
|
|
name = "output",
|
|
schema = GraphEmbeddings
|
|
)
|
|
)
|
|
|
|
async def on_message(self, msg, consumer, flow):
|
|
|
|
v = msg.value()
|
|
logger.info(f"Indexing {v.metadata.id}...")
|
|
|
|
try:
|
|
|
|
# Collect all contexts for batch embedding
|
|
contexts = [entity.context for entity in v.entities]
|
|
|
|
# Single batch embedding call
|
|
all_vectors = await flow("embeddings-request").embed(
|
|
texts=contexts
|
|
)
|
|
|
|
# Pair results with entities
|
|
entities = [
|
|
EntityEmbeddings(
|
|
entity=entity.entity,
|
|
vector=vector,
|
|
chunk_id=entity.chunk_id, # Provenance: source chunk
|
|
)
|
|
for entity, vector in zip(v.entities, all_vectors)
|
|
]
|
|
|
|
# Send in batches to avoid oversized messages
|
|
for i in range(0, len(entities), self.batch_size):
|
|
batch = entities[i:i + self.batch_size]
|
|
r = GraphEmbeddings(
|
|
metadata=v.metadata,
|
|
entities=batch,
|
|
)
|
|
await flow("output").send(r)
|
|
|
|
except Exception as e:
|
|
logger.error("Exception occurred", exc_info=True)
|
|
|
|
# Retry
|
|
raise e
|
|
|
|
logger.info("Done.")
|
|
|
|
@staticmethod
|
|
def add_args(parser):
|
|
|
|
parser.add_argument(
|
|
'--batch-size',
|
|
type=int,
|
|
default=default_batch_size,
|
|
help=f'Maximum entities per output message (default: {default_batch_size})'
|
|
)
|
|
|
|
FlowProcessor.add_args(parser)
|
|
|
|
def run():
|
|
|
|
Processor.launch(default_ident, __doc__)
|
|
|