MULLER Vectorized Hybrid Search Engine

[heathersherry]

MULLER provides a comprehensive suite of query functionalities tailored for AI data lakes:

• Comparison Operators: Supports exact and range matching using >,<, >=, and <= for numerical types (int/float) where the tensor htype is generic.
• Equality and Inequality: Supports == and != for int, float, str, and bool types (generic or text htypes). Users can optionally build inverted indexes to significantly accelerate retrieval performance.
• Range Queries: Supports the BETWEEN keyword for numerical types (int/float). This feature requires an inverted index.
• Full-Text Search: Supports the CONTAINS operator for str types (text htype), backed by an inverted index. For Chinese text, tokenization is handled by the open-source Jieba tokenizer.
• Pattern Matching: Supports LIKE for regular expression matching on str types (text htype).
• Boolean Logic: Supports complex query compositions using AND, OR, and NOT logical connectors.
• Pagination: Supports query results with OFFSET and LIMIT clauses for efficient data sampling.
• Data Aggregation: Supports standard SQL-like aggregation workflows, including SELECT, GROUP BY, and ORDER BY, alongside aggregate functions such as COUNT, AVG, MIN, MAX, and SUM.
• Vector Similarity Search: Supports high-dimensional vector similarity retrieval based on IVFPQ, HNSW and DISKANN for AI-centric embedding analysis.