klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-06 05:39:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

Setup Filters when configuring Text Search for each Search Type

Browse Source 
- Allows enabling different filters for different Text Search Types
- Use FileFilter in Text Search on Org Files
This commit is contained in:
Debanjum Singh Solanky
2022-09-05 01:05:13 +03:00
parent 1f9fd28b34
commit 092b9e329d
5 changed files with 33 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/configure.py
View File

@@ -14,6 +14,9 @@ from src.utils.config import SearchType, SearchModels, ProcessorConfigModel, Con
from src.utils import state from src.utils import state
from src.utils.helpers import resolve_absolute_path from src.utils.helpers import resolve_absolute_path
from src.utils.rawconfig import FullConfig, ProcessorConfig from src.utils.rawconfig import FullConfig, ProcessorConfig
from src.search_filter.date_filter import DateFilter
from src.search_filter.word_filter import WordFilter
from src.search_filter.file_filter import FileFilter
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
@@ -39,23 +42,25 @@ def configure_server(args, required=False):
def configure_search(model: SearchModels, config: FullConfig, regenerate: bool, t: SearchType = None): def configure_search(model: SearchModels, config: FullConfig, regenerate: bool, t: SearchType = None):
# Initialize Org Notes Search # Initialize Org Notes Search
if (t == SearchType.Org or t == None) and config.content_type.org: if (t == SearchType.Org or t == None) and config.content_type.org:
filter_directory = resolve_absolute_path(config.content_type.org.compressed_jsonl.parent)
filters = [DateFilter(), WordFilter(filter_directory, search_type=SearchType.Org), FileFilter()]
# Extract Entries, Generate Notes Embeddings # Extract Entries, Generate Notes Embeddings
model.orgmode_search = text_search.setup(org_to_jsonl, config.content_type.org, search_config=config.search_type.asymmetric, search_type=SearchType.Org, regenerate=regenerate) model.orgmode_search = text_search.setup(org_to_jsonl, config.content_type.org, search_config=config.search_type.asymmetric, regenerate=regenerate, filters=filters)
# Initialize Org Music Search # Initialize Org Music Search
if (t == SearchType.Music or t == None) and config.content_type.music: if (t == SearchType.Music or t == None) and config.content_type.music:
# Extract Entries, Generate Music Embeddings # Extract Entries, Generate Music Embeddings
model.music_search = text_search.setup(org_to_jsonl, config.content_type.music, search_config=config.search_type.asymmetric, search_type=SearchType.Music, regenerate=regenerate) model.music_search = text_search.setup(org_to_jsonl, config.content_type.music, search_config=config.search_type.asymmetric, regenerate=regenerate)
# Initialize Markdown Search # Initialize Markdown Search
if (t == SearchType.Markdown or t == None) and config.content_type.markdown: if (t == SearchType.Markdown or t == None) and config.content_type.markdown:
# Extract Entries, Generate Markdown Embeddings # Extract Entries, Generate Markdown Embeddings
model.markdown_search = text_search.setup(markdown_to_jsonl, config.content_type.markdown, search_config=config.search_type.asymmetric, search_type=SearchType.Markdown, regenerate=regenerate) model.markdown_search = text_search.setup(markdown_to_jsonl, config.content_type.markdown, search_config=config.search_type.asymmetric, regenerate=regenerate)
# Initialize Ledger Search # Initialize Ledger Search
if (t == SearchType.Ledger or t == None) and config.content_type.ledger: if (t == SearchType.Ledger or t == None) and config.content_type.ledger:
# Extract Entries, Generate Ledger Embeddings # Extract Entries, Generate Ledger Embeddings
model.ledger_search = text_search.setup(beancount_to_jsonl, config.content_type.ledger, search_config=config.search_type.symmetric, search_type=SearchType.Ledger, regenerate=regenerate) model.ledger_search = text_search.setup(beancount_to_jsonl, config.content_type.ledger, search_config=config.search_type.symmetric, regenerate=regenerate)
# Initialize Image Search # Initialize Image Search
if (t == SearchType.Image or t == None) and config.content_type.image: if (t == SearchType.Image or t == None) and config.content_type.image:

9
src/search_type/text_search.py
View File

@@ -7,13 +7,12 @@ import time
# External Packages # External Packages
import torch import torch
from sentence_transformers import SentenceTransformer, CrossEncoder, util from sentence_transformers import SentenceTransformer, CrossEncoder, util
from src.search_filter.date_filter import DateFilter from src.search_filter.base_filter import BaseFilter
from src.search_filter.word_filter import WordFilter
# Internal Packages # Internal Packages
from src.utils import state from src.utils import state
from src.utils.helpers import get_absolute_path, resolve_absolute_path, load_model from src.utils.helpers import get_absolute_path, resolve_absolute_path, load_model
from src.utils.config import SearchType, TextSearchModel from src.utils.config import TextSearchModel
from src.utils.rawconfig import TextSearchConfig, TextContentConfig from src.utils.rawconfig import TextSearchConfig, TextContentConfig
from src.utils.jsonl import load_jsonl from src.utils.jsonl import load_jsonl
@@ -153,7 +152,7 @@ def collate_results(hits, entries, count=5):
in hits[0:count]] in hits[0:count]]
def setup(text_to_jsonl, config: TextContentConfig, search_config: TextSearchConfig, search_type: SearchType, regenerate: bool) -> TextSearchModel: def setup(text_to_jsonl, config: TextContentConfig, search_config: TextSearchConfig, regenerate: bool, filters: list[BaseFilter] = []) -> TextSearchModel:
# Initialize Model # Initialize Model
bi_encoder, cross_encoder, top_k = initialize_model(search_config) bi_encoder, cross_encoder, top_k = initialize_model(search_config)
@@ -170,8 +169,6 @@ def setup(text_to_jsonl, config: TextContentConfig, search_config: TextSearchCon
config.embeddings_file = resolve_absolute_path(config.embeddings_file) config.embeddings_file = resolve_absolute_path(config.embeddings_file)
corpus_embeddings = compute_embeddings(entries, bi_encoder, config.embeddings_file, regenerate=regenerate) corpus_embeddings = compute_embeddings(entries, bi_encoder, config.embeddings_file, regenerate=regenerate)
filter_directory = resolve_absolute_path(config.compressed_jsonl.parent)
filters = [DateFilter(), WordFilter(filter_directory, search_type=search_type)]
for filter in filters: for filter in filters:
filter.load(entries, regenerate=regenerate) filter.load(entries, regenerate=regenerate)

14
tests/conftest.py
View File

@@ -1,4 +1,4 @@
# Standard Packages # External Packages
import pytest import pytest
# Internal Packages # Internal Packages
@@ -6,10 +6,13 @@ from src.search_type import image_search, text_search
from src.utils.config import SearchType from src.utils.config import SearchType
from src.utils.rawconfig import ContentConfig, TextContentConfig, ImageContentConfig, SearchConfig, TextSearchConfig, ImageSearchConfig 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.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') @pytest.fixture(scope='session')
def search_config(tmp_path_factory): def search_config(tmp_path_factory) -> SearchConfig:
model_dir = tmp_path_factory.mktemp('data') model_dir = tmp_path_factory.mktemp('data')
search_config = SearchConfig() search_config = SearchConfig()
@@ -35,7 +38,7 @@ def search_config(tmp_path_factory):
@pytest.fixture(scope='session') @pytest.fixture(scope='session')
def model_dir(search_config): def model_dir(search_config: SearchConfig):
model_dir = search_config.asymmetric.model_directory model_dir = search_config.asymmetric.model_directory
# Generate Image Embeddings from Test Images # Generate Image Embeddings from Test Images
@@ -55,13 +58,14 @@ def model_dir(search_config):
compressed_jsonl = model_dir.joinpath('notes.jsonl.gz'), compressed_jsonl = model_dir.joinpath('notes.jsonl.gz'),
embeddings_file = model_dir.joinpath('note_embeddings.pt')) embeddings_file = model_dir.joinpath('note_embeddings.pt'))
text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) filters = [DateFilter(), WordFilter(model_dir, search_type=SearchType.Org), FileFilter()]
text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False, filters=filters)
return model_dir return model_dir
@pytest.fixture(scope='session') @pytest.fixture(scope='session')
def content_config(model_dir): def content_config(model_dir) -> ContentConfig:
content_config = ContentConfig() content_config = ContentConfig()
content_config.org = TextContentConfig( content_config.org = TextContentConfig(
input_files = None, input_files = None,

12
tests/test_client.py
View File

@@ -4,7 +4,6 @@ from PIL import Image
# External Packages # External Packages
from fastapi.testclient import TestClient from fastapi.testclient import TestClient
import pytest
# Internal Packages # Internal Packages
from src.main import app from src.main import app
@@ -12,7 +11,8 @@ from src.utils.config import SearchType
from src.utils.state import model, config from src.utils.state import model, config
from src.search_type import text_search, image_search from src.search_type import text_search, image_search
from src.utils.rawconfig import ContentConfig, SearchConfig from src.utils.rawconfig import ContentConfig, SearchConfig
from src.processor.org_mode import org_to_jsonl from src.processor.org_mode.org_to_jsonl import org_to_jsonl
from src.search_filter.word_filter import WordFilter
# Arrange # Arrange
@@ -116,7 +116,7 @@ def test_image_search(content_config: ContentConfig, search_config: SearchConfig
# ---------------------------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------------------------
def test_notes_search(content_config: ContentConfig, search_config: SearchConfig): def test_notes_search(content_config: ContentConfig, search_config: SearchConfig):
# Arrange # Arrange
model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False)
user_query = "How to git install application?" user_query = "How to git install application?"
# Act # Act
@@ -132,7 +132,8 @@ def test_notes_search(content_config: ContentConfig, search_config: SearchConfig
# ---------------------------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------------------------
def test_notes_search_with_include_filter(content_config: ContentConfig, search_config: SearchConfig): def test_notes_search_with_include_filter(content_config: ContentConfig, search_config: SearchConfig):
# Arrange # Arrange
model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) filters = [WordFilter(content_config.org.compressed_jsonl.parent, search_type=SearchType.Org)]
model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False, filters=filters)
user_query = 'How to git install application? +"Emacs"' user_query = 'How to git install application? +"Emacs"'
# Act # Act
@@ -148,7 +149,8 @@ def test_notes_search_with_include_filter(content_config: ContentConfig, search_
# ---------------------------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------------------------
def test_notes_search_with_exclude_filter(content_config: ContentConfig, search_config: SearchConfig): def test_notes_search_with_exclude_filter(content_config: ContentConfig, search_config: SearchConfig):
# Arrange # Arrange
model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) filters = [WordFilter(content_config.org.compressed_jsonl.parent, search_type=SearchType.Org)]
model.orgmode_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False, filters=filters)
user_query = 'How to git install application? -"clone"' user_query = 'How to git install application? -"clone"'
# Act # Act

10
tests/test_text_search.py
View File

@@ -14,7 +14,7 @@ from src.processor.org_mode.org_to_jsonl import org_to_jsonl
def test_asymmetric_setup(content_config: ContentConfig, search_config: SearchConfig): def test_asymmetric_setup(content_config: ContentConfig, search_config: SearchConfig):
# Act # Act
# Regenerate notes embeddings during asymmetric setup # Regenerate notes embeddings during asymmetric setup
notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=True) notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=True)
# Assert # Assert
assert len(notes_model.entries) == 10 assert len(notes_model.entries) == 10
@@ -24,7 +24,7 @@ def test_asymmetric_setup(content_config: ContentConfig, search_config: SearchCo
# ---------------------------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------------------------
def test_asymmetric_search(content_config: ContentConfig, search_config: SearchConfig): def test_asymmetric_search(content_config: ContentConfig, search_config: SearchConfig):
# Arrange # Arrange
model.notes_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) model.notes_search = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False)
query = "How to git install application?" query = "How to git install application?"
# Act # Act
@@ -47,7 +47,7 @@ def test_asymmetric_search(content_config: ContentConfig, search_config: SearchC
# ---------------------------------------------------------------------------------------------------- # ----------------------------------------------------------------------------------------------------
def test_asymmetric_reload(content_config: ContentConfig, search_config: SearchConfig): def test_asymmetric_reload(content_config: ContentConfig, search_config: SearchConfig):
# Arrange # Arrange
initial_notes_model= text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) initial_notes_model= text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False)
assert len(initial_notes_model.entries) == 10 assert len(initial_notes_model.entries) == 10
assert len(initial_notes_model.corpus_embeddings) == 10 assert len(initial_notes_model.corpus_embeddings) == 10
@@ -60,11 +60,11 @@ def test_asymmetric_reload(content_config: ContentConfig, search_config: SearchC
f.write("\n* A Chihuahua doing Tango\n- Saw a super cute video of a chihuahua doing the Tango on Youtube\n") f.write("\n* A Chihuahua doing Tango\n- Saw a super cute video of a chihuahua doing the Tango on Youtube\n")
# regenerate notes jsonl, model embeddings and model to include entry from new file # regenerate notes jsonl, model embeddings and model to include entry from new file
regenerated_notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=True) regenerated_notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=True)
# Act # Act
# reload embeddings, entries, notes model from previously generated notes jsonl and model embeddings files # reload embeddings, entries, notes model from previously generated notes jsonl and model embeddings files
initial_notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, SearchType.Org, regenerate=False) initial_notes_model = text_search.setup(org_to_jsonl, content_config.org, search_config.asymmetric, regenerate=False)
# Assert # Assert
assert len(regenerated_notes_model.entries) == 11 assert len(regenerated_notes_model.entries) == 11