💡 All-in-one AI framework for semantic search, LLM orchestration and language model workflows
Python agentsaiai-agentsembeddingsinformation-retrievallanguage-modellarge-language-modelsllmnlppythonragretrieval-augmented-generationsearchsearch-enginesemantic-searchsentence-embeddingstransformerstxtaivector-databasevector-search
Modify variable name
davidmezzettid9ffcf6e5c73
diff --git a/src/python/txtai/embeddings/search.py b/src/python/txtai/embeddings/search.pyindex a2c614e1..93bbf795 100644--- a/src/python/txtai/embeddings/search.py+++ b/src/python/txtai/embeddings/search.py@@ -92,7 +92,7 @@ class Search:limit = max(limit, self.limit(queries))# Extract embeddings queries as single batch across all queries- equeries, search, window = [], [], 0+ equeries, search, candidates = [], [], 0for x, query in enumerate(queries):if "similar" in query:# Get handle to similar queries@@ -100,14 +100,14 @@ class Search:# Store the query index and similar text argument (first argument)equeries.append((x, params[0]))- # Second argument is similarity candidate window+ # Second argument is similar candidate limitif len(params) > 1 and params[1].isdigit():- # Get largest window value across all queries- window = int(params[1]) if int(params[1]) > window else window+ # Get largest candidate limit across all queries+ candidates = int(params[1]) if int(params[1]) > candidates else candidates# Bulk approximate nearest neighbor searchif equeries:- search = self.search([query for _, query in equeries], limit * 10 if not window else window)+ search = self.search([query for _, query in equeries], limit * 10 if not candidates else candidates)# Combine approximate nearest neighbor search results with database search resultsresults = []