remove print statements now that the join is working.

updated the precomputed TF-IDF dataframe to not include the base path to the docs folder

fixing indentation format

added debug print statements to check why we get no results

added debug print statements to check why we get no results

point to the actual location of the module files

use valid columns for the result

tracking the new block embeddings

trying to remove old embeddings file from LFS

look for column named 'file' not 'path'

fixed an error where the polars chains weren't wrapped in parens

return results as HTML links

adding the search page back #3

removed debugging messages now that search is working

making sure all the correct arguments are there in the function calls

pass IDF DTM to the function factory

removed some cruft from app.py

the factory has to actually return the function

corrected the name of the fasttext model binary

send row names to be included in the results

build the search functions on app launch, rather than per-query

add the data files necessary for search to version control

switched to the new search functions using sentence-transformers

add sentence_transformers requirement

added search URL to the README

renamed modules that do the search

changes TF-IDF search to cosine similarity from dot product

changes TF-IDF search to cosine similarity from dot product

add the rank-combined column before using it

being more selective with the columns that print in the output

need to used vectorized strip_prefix to modify a polars column

shortening the name of files

just return the dist of results directly

trying to get a response

testing a simpler response

return results as JSON

render result to a jinja table

apparently need pyarrow for pandas to_html and the huggingface environment-builder is too stupid to install required packages

working out how to format the result

add pandas to requirements

trying to figure out why the last print failed

switch tfidf search to use file list saved by joblib

reshape with shape, not size

reduce files list to 1-d

convert output to string

testing separate query methods

trying to test app.py

trying to test app.py

fixed a typo and added a test endpoint

use block_embeddings_df from the compressed serialized parquet file