apunkt/trustgraph
1
0
Fork 0
mirror of https://github.com/trustgraph-ai/trustgraph.git synced 2026-07-11 06:12:11 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Implementation

Browse source
This commit is contained in:
Cyber MacGeddon 2026-03-08 18:34:22 +00:00
parent 87ef89ef7b
commit 3d4f859aa2
8 changed files with 73 additions and 47 deletions
Download patch file Download diff file Expand all files Collapse all files

6
trustgraph-flow/trustgraph/embeddings/document_embeddings/embeddings.py
View file

@ -62,11 +62,13 @@ class Processor(FlowProcessor):
resp = await flow("embeddings-request").request(
EmbeddingsRequest(
text = v.chunk
texts=[v.chunk]
)
)
vectors = resp.vectors
# vectors[0] is the vector set for the first (only) text
# vectors[0][0] is the first vector in that set
vectors = resp.vectors[0][0] if resp.vectors else []
embeds = [
ChunkEmbeddings(

26
trustgraph-flow/trustgraph/embeddings/graph_embeddings/embeddings.py
View file

@ -58,23 +58,25 @@ class Processor(FlowProcessor):
v = msg.value()
logger.info(f"Indexing {v.metadata.id}...")
entities = []
try:
for entity in v.entities:
# Collect all contexts for batch embedding
contexts = [entity.context for entity in v.entities]
vectors = await flow("embeddings-request").embed(
text = entity.context
)
# Single batch embedding call
all_vectors = await flow("embeddings-request").embed(
texts=contexts
)
entities.append(
EntityEmbeddings(
entity=entity.entity,
vectors=vectors,
chunk_id=entity.chunk_id, # Provenance: source chunk
)
# Pair results with entities
entities = [
EntityEmbeddings(
entity=entity.entity,
vectors=vectors[0], # First vector from the set
chunk_id=entity.chunk_id, # Provenance: source chunk
)
for entity, vectors in zip(v.entities, all_vectors)
]
# Send in batches to avoid oversized messages
for i in range(0, len(entities), self.batch_size):

14
trustgraph-flow/trustgraph/embeddings/row_embeddings/embeddings.py
View file

@ -200,15 +200,23 @@ class Processor(CollectionConfigHandler, FlowProcessor):
embeddings_list = []
try:
for text, (index_name, index_value) in texts_to_embed.items():
vectors = await flow("embeddings-request").embed(text=text)
# Collect texts and metadata for batch embedding
texts = list(texts_to_embed.keys())
metadata = list(texts_to_embed.values())
# Single batch embedding call
all_vectors = await flow("embeddings-request").embed(texts=texts)
# Pair results with metadata
for text, (index_name, index_value), vectors in zip(
texts, metadata, all_vectors
):
embeddings_list.append(
RowIndexEmbedding(
index_name=index_name,
index_value=index_value,
text=text,
vectors=vectors
vectors=vectors[0] # First vector from the set
)
)