Powered by Dipamkara

Abstract

With the rapid development of big data and artificial intelligence technologies, the demand for effective processing and retrieval of vector data is growing. Against this backdrop, I have developed the Bhakti vector database, aiming to provide a lightweight and easy-to-deploy solution to meet the storage and semantic search needs of small and medium-sized datasets. Bhakti supports a variety of similarity calculation methods and a domain-specific language (DSL) for document-based pattern matching pre-filtering, facilitating migration of data with its portable data files, flexible data management and seamless integration with Python3. Furthermore, I propose a memory-enhanced large language model dialogue solution based on the Bhakti database, which can assign different weights to the question and answer in dialogue history, achieving fine-grained control over the semantic importance of each segment in a single dialogue history. Through experimental validation, my method shows significant performance in the application of semantic search and question-answering systems. Although there are limitations in processing large datasets, such as not supporting approximate calculation methods like HNSW, the lightweight nature of Bhakti gives it a clear advantage in scenarios involving small and medium-sized datasets.

Citation

If you are incorporating Bhakti into your research, please remember to properly cite it to acknowledge its contribution to your work.

如果您正在將 Bhakti 整合到您的研究中，請務必正確引用它，以聲明它對您工作的貢獻.

@article{wu2025bhakti,
  author = {Zihao Wu},
  title = {Bhakti: A Lightweight Vector Database Management System for Endowing Large Language Models with Semantic Search Capabilities and Memory},
  journal = {arXiv preprint},
  year = {2025},
  eprint = {2504.01553},
  archivePrefix = {arXiv},
  primaryClass = {cs.DB},
  url = {https://arxiv.org/abs/2504.01553}
}

Installation

• From PYPI

  pip install bhakti

• From Github

  Download .whl first then run

  pip install ./bhakti-X.X.X-py3-none-any.whl

Quick Start

Before all, make sure you've successfully installed Bhakti :)

• Run Bhakti Server

  • To begin, create a path for storing data

    mkdir -p /path/to/db

  • Start server using shell command

    1. Create configuration file (.yaml)

       # bhakti.yaml
       DIMENSION: 1024
       DB_PATH: /path/to/db
       DB_ENGINE: dipamkara # optional, default to dipamkara
       CACHED: false # optional, default to false
       HOST: 0.0.0.0 # optional, default to 0.0.0.0
       PORT: 23860 # optional, default to 23860
       EOF: <eof> # optional, default to <eof>
       TIMEOUT: 4.0 # optional, default to 4.0 seconds
       BUFFER_SIZE: 256 # optional, default to 256 bytes
       VERBOSE: false # optional, default to false

    2. Run bhakti in shell

       # bash
       bhakti ./bhakti.yaml

  • Start server using Python

    # main.py
    from bhakti import BhaktiServer
    from bhakti.database import DBEngine
    
    if __name__ == '__main__':
        bhakti_server = BhaktiServer(
            dimension=1024,  # required, only vectors with 1024 dimensions are acceptable
            db_path='/path/to/db',  # required, path where stores data, portable
            db_engine=DBEngine.DIPAMKARA,  # optional, default to dipamkara
            cached=False,  # optional, default to false
            host='0.0.0.0',  # optional, default to 0.0.0.0
            port=23860,  # optional, default to 23860
            eof=b'<eof>',  # optional, default to b'<eof>'
            timeout=4.0,  # optional, default to 4.0 seconds
            buffer_size=256,  # optional, default to 256 bytes
            verbose=False  # optional, default to false
        )
        # run server
        bhakti_server.run()

• Interact With A Bhakti Client

  Currently, Python(>=3.10) is supported

  # main.py
  import asyncio
  import numpy as np
  from bhakti import BhaktiClient
  from bhakti.database import Metric
  from bhakti.database import DBEngine
  
  
  async def main():
      client = BhaktiClient(
          server='127.0.0.1',  # optional, default to 127.0.0.1
          port=23860,  # optional, default to 23860
          eof=b'<eof>',  # optional, default to b'<eof>'
          timeout=4.0,  # optional, default to 4.0 seconds
          buffer_size=256,  # optional, default to 256 bytes
          db_engine=DBEngine.DIPAMKARA,  # optional, default to dipamkara
          verbose=False  # optional, default to false
      )
      vector = np.random.randn(1024)
      await client.create(vector=vector, document={'age': 31, 'gender': 'male'})
      await client.create_index('age')
      await client.create_index('gender')
      results = await client.find_documents_by_vector_indexed(
          query='age <= 31 && gender != "female"', 
          vector=vector,
          metric=Metric.EUCLIDEAN_Z_SCORE, 
          top_k=3
      )
      print(results)
      
  
  if __name__ == '__main__':
      asyncio.run(main())

• Projects Related

  • Bhakti based memory retrieval dialogue solution