tiledb-vector-search 0.0.7

TileDB Vector Search Python client

Prerequisites

1. Create and activate a Python environment with pybind11 available

a. Download minimamba: b. mamba env create -f environment.yml c. mamba activate tiledbvspy

Build Instructions

Build and install with pip

See setup.py for supported configuration overrides, including the libtiledb path.

pip install -e .

Direct build, then export PYTHONPATH

For development purposes, at the moment it is simpler to build directly and export the extension directory into PYTHONPATH.

1. Run the top level cmake build with -DTILEDB_VS_PYTHON=ON

mkdir fvp-build
cd fvp-build
cmake -DTILEDB_VS_PYTHON=ON ~/work/git/feature-vector-type
<cmake --build . or make -j4 etc.>

2. Look at fvp-build/python. There should be a Python extension module:

pydev ❯ pwd
/Users/inorton/work/bld/fvp-build/python

~/work/bld/fvp-build/python
pydev ❯ ls
CMakeFiles                       cmake_install.cmake              tiledbvspy.cpython-310-darwin.so

3. Export the extension path as PYTHONPATH

export PYTHONPATH= ~/work/bld/fvp-build/python

3. Run Python, and import tiledbvspy

import tiledbvspy