iris-vector-graph 1.26.0__tar.gz → 1.26.2__tar.gz

{iris_vector_graph-1.26.0 → iris_vector_graph-1.26.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: iris-vector-graph
-Version: 1.26.0
+Version: 1.26.2
 Summary: Transactional Graph + Vector retrieval system for InterSystems IRIS with hybrid search, openCypher, and GraphQL APIs
 Project-URL: Homepage, https://github.com/isc-tdyar/iris-vector-graph
 Project-URL: Documentation, https://github.com/isc-tdyar/iris-vector-graph/tree/main/docs

{iris_vector_graph-1.26.0 → iris_vector_graph-1.26.2}/iris_src/src/Graph/KG/PLAIDSearch.cls

@@ -29,6 +29,29 @@ ClassMethod StoreCentroids(name As %String, centroidsJSON As %String)
 	If K > 0 Set ^PLAID(name, "meta", "dim") = dim
 }
 
+ClassMethod StoreDocTokensBatch(name As %String, batchJSON As %String)
+{
+	Set batch = ##class(%DynamicArray).%FromJSON(batchJSON)
+	Set nDocs = batch.%Size()
+	For d = 0:1:(nDocs-1) {
+		Set docObj = batch.%Get(d)
+		Set docId = docObj.%Get("id")
+		Set tokArr = docObj.%Get("tokens")
+		Set nTok = tokArr.%Size()
+		For t = 0:1:(nTok-1) {
+			Set tArr = tokArr.%Get(t)
+			Set dim = tArr.%Size()
+			Set vec = ""
+			For i = 0:1:(dim-1) {
+				Set $vector(vec, i+1, "double") = +tArr.%Get(i)
+			}
+			Set ^PLAID(name, "docTokens", docId, t) = vec
+		}
+		Set ^PLAID(name, "meta", "nDocs") = $Get(^PLAID(name, "meta", "nDocs"), 0) + 1
+		Set ^PLAID(name, "meta", "totalTokens") = $Get(^PLAID(name, "meta", "totalTokens"), 0) + nTok
+	}
+}
+
 ClassMethod StoreDocTokens(name As %String, docId As %String, tokensJSON As %String)
 {
 	Set arr = ##class(%DynamicArray).%FromJSON(tokensJSON)

{iris_vector_graph-1.26.0 → iris_vector_graph-1.26.2}/iris_vector_graph/engine.py

@@ -8,7 +8,6 @@ that can be used across any domain.
 """
 
 import json
-import numpy as np
 from pathlib import Path
 from typing import List, Tuple, Optional, Dict, Any
 import logging
@@ -977,7 +976,7 @@ class IRISGraphEngine:
         """
         cursor = self.conn.cursor()
         try:
-            # Parse query vector from JSON string
+            import numpy as np
             query_array = np.array(json.loads(query_vector))
 
             # Get embeddings with optional label filter (optimized query)
@@ -1542,12 +1541,19 @@ class IRISGraphEngine:
         centroids_json = json.dumps(kmeans.cluster_centers_.tolist())
         iris_obj.classMethodVoid("Graph.KG.PLAIDSearch", "StoreCentroids", name, centroids_json)
 
-        for doc in docs:
-            tokens_json = json.dumps([[float(v) for v in tok] for tok in doc["tokens"]])
-            iris_obj.classMethodVoid("Graph.KG.PLAIDSearch", "StoreDocTokens", name, doc["id"], tokens_json)
-
-        assignments_json = json.dumps(doc_token_map)
-        iris_obj.classMethodVoid("Graph.KG.PLAIDSearch", "BuildInvertedIndex", name, assignments_json)
+        BATCH_SIZE = 50
+        for i in range(0, len(docs), BATCH_SIZE):
+            batch = docs[i : i + BATCH_SIZE]
+            batch_json = json.dumps([
+                {"id": doc["id"], "tokens": [[float(v) for v in tok] for tok in doc["tokens"]]}
+                for doc in batch
+            ])
+            iris_obj.classMethodVoid("Graph.KG.PLAIDSearch", "StoreDocTokensBatch", name, batch_json)
+
+        ASSIGN_CHUNK = 5000
+        for i in range(0, len(doc_token_map), ASSIGN_CHUNK):
+            chunk_json = json.dumps(doc_token_map[i : i + ASSIGN_CHUNK])
+            iris_obj.classMethodVoid("Graph.KG.PLAIDSearch", "BuildInvertedIndex", name, chunk_json)
 
         return json.loads(str(iris_obj.classMethodValue("Graph.KG.PLAIDSearch", "Info", name)))
 

{iris_vector_graph-1.26.0 → iris_vector_graph-1.26.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "iris-vector-graph"
-version = "1.26.0"
+version = "1.26.2"
 description = "Transactional Graph + Vector retrieval system for InterSystems IRIS with hybrid search, openCypher, and GraphQL APIs"
 readme = "README.md"
 license = "MIT"