Skip to content

armgit / txtai

Public Imported from GitHub
0

💡 All-in-one AI framework for semantic search, LLM orchestration and language model workflows

Update default vector model logic for embeddings.batchtransform

davidmezzetti9a670cd5ecbe
diff --git a/src/python/txtai/embeddings/base.py b/src/python/txtai/embeddings/base.py
index d7e02aa7..6bf5e5b9 100644
--- a/src/python/txtai/embeddings/base.py
+++ b/src/python/txtai/embeddings/base.py
@@ -315,7 +315,7 @@ class Embeddings:
self.defaults()
# Default vector model, if necessary
- model = self.model if self.model else next(iter(self.models.values()))
+ model = self.model if self.model else self.indexes.model()
# Convert documents into sentence embeddings
embeddings = model.batchtransform(Stream(self)(documents), category)
@@ -877,7 +877,7 @@ class Embeddings:
if "indexes" in self.config and self.models is None:
self.models = {}
- # Load and store uncached model
+ # Load vector model
return VectorsFactory.create(self.config, self.scoring, self.models)
def loadquery(self):
diff --git a/src/python/txtai/embeddings/index/indexes.py b/src/python/txtai/embeddings/index/indexes.py
index e3c3c1f2..abff9b96 100644
--- a/src/python/txtai/embeddings/index/indexes.py
+++ b/src/python/txtai/embeddings/index/indexes.py
@@ -18,7 +18,7 @@ class Indexes:
Args:
embeddings: embeddings instance
- indexes: list of subindexes to add
+ indexes: dict of subindexes to add
"""
self.embeddings = embeddings
@@ -83,6 +83,18 @@ class Indexes:
return list(self.indexes.keys())[0]
+ def model(self):
+ """
+ Scans indexes and gets the first vector model.
+
+ Returns:
+ Vectors
+ """
+
+ # Return first vector model
+ matches = [index.model for index in self.indexes.values() if index.model]
+ return matches[0] if matches else None
+
def insert(self, documents, index=None):
"""
Inserts a batch of documents into each subindex.
diff --git a/test/python/testdatabase/testrdbms.py b/test/python/testdatabase/testrdbms.py
index 66c29f4c..cb62c5f2 100644
--- a/test/python/testdatabase/testrdbms.py
+++ b/test/python/testdatabase/testrdbms.py
@@ -685,6 +685,9 @@ class Common:
)
embeddings.index(data)
+ # Test transform
+ self.assertEqual(embeddings.transform("feel good story").shape, (768,))
+
# Run search
result = embeddings.search("feel good story", 1)[0]
self.assertEqual(result["text"], data[4][1])
diff --git a/test/python/testembeddings.py b/test/python/testembeddings.py
index 1a96c2f6..a9bfc68f 100644
--- a/test/python/testembeddings.py
+++ b/test/python/testembeddings.py
@@ -388,6 +388,9 @@ class TestEmbeddings(unittest.TestCase):
embeddings = Embeddings({"defaults": False, "indexes": {"index1": {"path": "sentence-transformers/nli-mpnet-base-v2"}}})
embeddings.index(data)
+ # Test transform
+ self.assertEqual(embeddings.transform("feel good story").shape, (768,))
+
# Run search
uid = embeddings.search("feel good story", 1)[0][0]
self.assertEqual(uid, 4)