highperfocused/danswer
1
0
Fork 0
mirror of https://github.com/danswer-ai/danswer.git synced 2025-07-03 11:11:45 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add image support for search (#4090)

Browse Source 
* add support for image search

* quick fix up

* k

* k

* k

* k

* nit

* quick fix for connector tests
This commit is contained in:
pablonyx
2025-03-05 09:44:18 -08:00
committed by GitHub
parent f731beca1f
commit 20f2b9b2bb
36 changed files with 1857 additions and 589 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
backend/onyx/document_index/vespa/app_config/schemas/danswer_chunk.sd
View File

@ -55,6 +55,9 @@ schema DANSWER_CHUNK_NAME {
field blurb type string {
indexing: summary | attribute
}
field image_file_name type string {
indexing: summary | attribute
}
# https://docs.vespa.ai/en/attributes.html potential enum store for speed, but probably not worth it
field source_type type string {
indexing: summary | attribute

3
backend/onyx/document_index/vespa/chunk_retrieval.py
View File

@ -31,6 +31,7 @@ from onyx.document_index.vespa_constants import DOC_UPDATED_AT
from onyx.document_index.vespa_constants import DOCUMENT_ID
from onyx.document_index.vespa_constants import DOCUMENT_ID_ENDPOINT
from onyx.document_index.vespa_constants import HIDDEN
from onyx.document_index.vespa_constants import IMAGE_FILE_NAME
from onyx.document_index.vespa_constants import LARGE_CHUNK_REFERENCE_IDS
from onyx.document_index.vespa_constants import MAX_ID_SEARCH_QUERY_SIZE
from onyx.document_index.vespa_constants import MAX_OR_CONDITIONS
@ -130,6 +131,7 @@ def _vespa_hit_to_inference_chunk(
section_continuation=fields[SECTION_CONTINUATION],
document_id=fields[DOCUMENT_ID],
source_type=fields[SOURCE_TYPE],
image_file_name=fields.get(IMAGE_FILE_NAME),
title=fields.get(TITLE),
semantic_identifier=fields[SEMANTIC_IDENTIFIER],
boost=fields.get(BOOST, 1),
@ -211,6 +213,7 @@ def _get_chunks_via_visit_api(
# Check if the response contains any documents
response_data = response.json()
if "documents" in response_data:
for document in response_data["documents"]:
if filters.access_control_list:

3
backend/onyx/document_index/vespa/indexing_utils.py
View File

@ -32,6 +32,7 @@ from onyx.document_index.vespa_constants import DOCUMENT_ID
from onyx.document_index.vespa_constants import DOCUMENT_ID_ENDPOINT
from onyx.document_index.vespa_constants import DOCUMENT_SETS
from onyx.document_index.vespa_constants import EMBEDDINGS
from onyx.document_index.vespa_constants import IMAGE_FILE_NAME
from onyx.document_index.vespa_constants import LARGE_CHUNK_REFERENCE_IDS
from onyx.document_index.vespa_constants import METADATA
from onyx.document_index.vespa_constants import METADATA_LIST
@ -198,13 +199,13 @@ def _index_vespa_chunk(
# which only calls VespaIndex.update
ACCESS_CONTROL_LIST: {acl_entry: 1 for acl_entry in chunk.access.to_acl()},
DOCUMENT_SETS: {document_set: 1 for document_set in chunk.document_sets},
IMAGE_FILE_NAME: chunk.image_file_name,
BOOST: chunk.boost,
}
if multitenant:
if chunk.tenant_id:
vespa_document_fields[TENANT_ID] = chunk.tenant_id
vespa_url = f"{DOCUMENT_ID_ENDPOINT.format(index_name=index_name)}/{vespa_chunk_id}"
logger.debug(f'Indexing to URL "{vespa_url}"')
res = http_client.post(

2
backend/onyx/document_index/vespa_constants.py
View File

@ -77,6 +77,7 @@ PRIMARY_OWNERS = "primary_owners"
SECONDARY_OWNERS = "secondary_owners"
RECENCY_BIAS = "recency_bias"
HIDDEN = "hidden"
IMAGE_FILE_NAME = "image_file_name"
# Specific to Vespa, needed for highlighting matching keywords / section
CONTENT_SUMMARY = "content_summary"
@ -94,6 +95,7 @@ YQL_BASE = (
f"{SEMANTIC_IDENTIFIER}, "
f"{TITLE}, "
f"{SECTION_CONTINUATION}, "
f"{IMAGE_FILE_NAME}, "
f"{BOOST}, "
f"{HIDDEN}, "
f"{DOC_UPDATED_AT}, "