mirror of
https://github.com/khoj-ai/khoj.git
synced 2024-11-30 10:53:02 +01:00
c17a0fd05b
- It's more of a hassle to not let word filter go stale on entry updates - Generating index on 120K lines of notes takes 1s. Loading from file takes 0.2s. For less content load time difference will be even smaller - Let go of startup time improvement for simplicity for now
82 lines
No EOL
2.9 KiB
Python
82 lines
No EOL
2.9 KiB
Python
# External Packages
|
|
import pytest
|
|
|
|
# Internal Packages
|
|
from src.search_type import image_search, text_search
|
|
from src.utils.config import SearchType
|
|
from src.utils.rawconfig import ContentConfig, TextContentConfig, ImageContentConfig, SearchConfig, TextSearchConfig, ImageSearchConfig
|
|
from src.processor.org_mode.org_to_jsonl import org_to_jsonl
|
|
from src.search_filter.date_filter import DateFilter
|
|
from src.search_filter.word_filter import WordFilter
|
|
from src.search_filter.file_filter import FileFilter
|
|
|
|
|
|
@pytest.fixture(scope='session')
|
|
def search_config(tmp_path_factory) -> SearchConfig:
|
|
model_dir = tmp_path_factory.mktemp('data')
|
|
|
|
search_config = SearchConfig()
|
|
|
|
search_config.symmetric = TextSearchConfig(
|
|
encoder = "sentence-transformers/all-MiniLM-L6-v2",
|
|
cross_encoder = "cross-encoder/ms-marco-MiniLM-L-6-v2",
|
|
model_directory = model_dir
|
|
)
|
|
|
|
search_config.asymmetric = TextSearchConfig(
|
|
encoder = "sentence-transformers/multi-qa-MiniLM-L6-cos-v1",
|
|
cross_encoder = "cross-encoder/ms-marco-MiniLM-L-6-v2",
|
|
model_directory = model_dir
|
|
)
|
|
|
|
search_config.image = ImageSearchConfig(
|
|
encoder = "sentence-transformers/clip-ViT-B-32",
|
|
model_directory = model_dir
|
|
)
|
|
|
|
return search_config
|
|
|
|
|
|
@pytest.fixture(scope='session')
|
|
def model_dir(search_config: SearchConfig):
|
|
model_dir = search_config.asymmetric.model_directory
|
|
|
|
# Generate Image Embeddings from Test Images
|
|
content_config = ContentConfig()
|
|
content_config.image = ImageContentConfig(
|
|
input_directories = ['tests/data/images'],
|
|
embeddings_file = model_dir.joinpath('image_embeddings.pt'),
|
|
batch_size = 10,
|
|
use_xmp_metadata = False)
|
|
|
|
image_search.setup(content_config.image, search_config.image, regenerate=False)
|
|
|
|
# Generate Notes Embeddings from Test Notes
|
|
content_config.org = TextContentConfig(
|
|
input_files = None,
|
|
input_filter = 'tests/data/org/*.org',
|
|
compressed_jsonl = model_dir.joinpath('notes.jsonl.gz'),
|
|
embeddings_file = model_dir.joinpath('note_embeddings.pt'))
|
|
|
|
filters = [DateFilter(), WordFilter(), FileFilter()]
|
|
text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False, filters=filters)
|
|
|
|
return model_dir
|
|
|
|
|
|
@pytest.fixture(scope='session')
|
|
def content_config(model_dir) -> ContentConfig:
|
|
content_config = ContentConfig()
|
|
content_config.org = TextContentConfig(
|
|
input_files = None,
|
|
input_filter = 'tests/data/org/*.org',
|
|
compressed_jsonl = model_dir.joinpath('notes.jsonl.gz'),
|
|
embeddings_file = model_dir.joinpath('note_embeddings.pt'))
|
|
|
|
content_config.image = ImageContentConfig(
|
|
input_directories = ['tests/data/images'],
|
|
embeddings_file = model_dir.joinpath('image_embeddings.pt'),
|
|
batch_size = 1,
|
|
use_xmp_metadata = False)
|
|
|
|
return content_config |