Enrich documents with embeddings instead of recreating documents (#42)

deepset-ai · Oct 19, 2023 · 8d15314 · 8d15314
1 parent 955f8f9
commit 8d15314
Showing 1 changed file with 2 additions and 6 deletions.
diff --git a/components/instructor-embedders/instructor_embedders/instructor_document_embedder.py b/components/instructor-embedders/instructor_embedders/instructor_document_embedder.py
@@ -126,11 +126,7 @@ def run(self, documents: List[Document]):
             normalize_embeddings=self.normalize_embeddings,
         )
 
-        documents_with_embeddings = []
         for doc, emb in zip(documents, embeddings):
-            doc_as_dict = doc.to_dict()
-            doc_as_dict["embedding"] = emb
-            del doc_as_dict["id"]
-            documents_with_embeddings.append(Document.from_dict(doc_as_dict))
+            doc.embedding = emb
 
-        return {"documents": documents_with_embeddings}
+        return {"documents": documents}