Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Vespa schema changes for query control & general quality of life #163

Draft
wants to merge 11 commits into
base: main
Choose a base branch
from
+67 −133
Draft

Vespa schema changes for query control & general quality of life #163

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
9048c5b
set document language to english
kdutia Dec 18, 2024
9b8ea4a
remove unused 'exact' rank-profile
kdutia Dec 18, 2024
59084c5
add weights with defaults to hybrid profile
kdutia Dec 18, 2024
401105e
hybrid_no_closeness schema inherits from hybrid
kdutia Dec 18, 2024
6a7c6b0
remove hybrid_no_description_embedding rank-profile
kdutia Dec 18, 2024
780bfa4
search_summary_with_tokens inherits search_summary
kdutia Dec 18, 2024
7e2411d
add all features to rank profile summary-features
kdutia Dec 18, 2024
4f238b1
add nativerank profiles
kdutia Dec 18, 2024
7ab8260
add field variants with bolding
kdutia Dec 18, 2024
d6e21d8
bump version to 1.12.0
kdutia Dec 18, 2024
3aec5fc
add summary to family_name_bolding & family_description_bolding
kdutia Dec 18, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/cpr_sdk/version.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
_MAJOR = "1"
_MINOR = "11"
_MINOR = "12"
_PATCH = "0"
_SUFFIX = ""

Expand Down
87 changes: 31 additions & 56 deletions tests/local_vespa/test_app/schemas/document_passage.sd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,15 @@ schema document_passage {
stemming: none
}

field language type string {
indexing: "en" | set_language
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looking at the docs for this:

The recommended use is to have one field in the document containing the language code, and that field should be the first field in the document, as it will only affect the fields defined after it in the schema.

https://docs.vespa.ai/en/reference/indexing-language-reference.html#set_language

This feels weird to me to have a document config at a field level, but it is what it is! I'm wondering if we need to move the language field above text_block_not_stemmed?

}

field text_block_bolding type string {
indexing: input text_block | summary | index
bolding: true
}

document document_passage {

field search_weights_ref type reference<search_weights> {
Expand Down Expand Up @@ -139,91 +148,57 @@ schema document_passage {
summary concepts {}
}

document-summary search_summary_with_tokens {
summary family_name {}
summary family_description {}
summary family_import_id {}
summary family_slug {}
summary family_category {}
summary family_publication_ts {}
summary family_geography {}
summary family_geographies {}
summary family_source {}
summary document_import_id {}
summary document_slug {}
summary document_languages {}
summary document_content_type {}
summary document_cdn_object {}
summary document_source_url {}
summary corpus_import_id {}
summary corpus_type_name {}
summary metadata {}
summary text_block {}
summary text_block_id {}
summary text_block_type {}
summary text_block_page {}
summary text_block_coords {}
summary concepts {}
document-summary search_summary_with_tokens inherits search_summary {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ohh this is nice!

summary text_block_tokens {
source: text_block
tokens
}
}

rank-profile exact inherits default {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice cleaning this up now we use exact_not_stemmed instead 🎉

function text_score() {
expression: attribute(passage_weight) * fieldMatch(text_block)
}
first-phase {
expression: text_score()
}
summary-features: text_score() fieldMatch(text_block)
}

rank-profile exact_not_stemmed inherits default {
function text_score() {
expression: attribute(passage_weight) * fieldMatch(text_block_not_stemmed)
expression: fieldMatch(text_block_not_stemmed)
}
first-phase {
expression: text_score()
expression: attribute(passage_weight) * text_score()
}
summary-features: text_score() fieldMatch(text_block)
}

rank-profile hybrid_no_closeness inherits default {
function text_score() {
expression: attribute(passage_weight) * bm25(text_block)
}
first-phase {
expression: text_score()
}
summary-features: text_score() bm25(text_block)
summary-features: attribute(passage_weight) text_score()
}

rank-profile hybrid inherits default {
inputs {
query(query_embedding) tensor<float>(x[768])
query(passage_bm25_weight) double: 1.0
query(passage_closeness_weight) double: 1.0
}
function text_score() {
expression: attribute(passage_weight) * (bm25(text_block) + closeness(text_embedding))
expression: query(passage_bm25_weight) * bm25(text_block) + query(passage_closeness_weight) * closeness(text_embedding)
}
first-phase {
expression: text_score()
expression: attribute(passage_weight) * text_score()
}
summary-features: text_score() bm25(text_block) closeness(text_embedding)
summary-features: text_score() bm25(text_block) closeness(text_embedding) attribute(passage_weight)
}

rank-profile hybrid_custom_weight inherits default {
rank-profile hybrid_nativerank inherits default {
inputs {
query(query_embedding) tensor<float>(x[768])
query(bm25_weight) double
query(passage_nativerank_weight) double: 1.0
query(passage_closeness_weight) double: 1.0
}
function text_score() {
expression: attribute(passage_weight) * (query(bm25_weight) * bm25(text_block) + closeness(text_embedding))
expression: query(passage_nativerank_weight) * nativeRank(text_block) + query(passage_closeness_weight) * closeness(text_embedding)
}
first-phase {
expression: text_score()
expression: attribute(passage_weight) * text_score()
}
summary-features: text_score() bm25(text_block) closeness(text_embedding)
summary-features: text_score() nativeRank(text_block) closeness(text_embedding) attribute(passage_weight)
}

rank-profile hybrid_no_closeness inherits hybrid {
inputs {
query(passage_closeness_weight) double: 0.0
}
}

}
111 changes: 35 additions & 76 deletions tests/local_vespa/test_app/schemas/family_document.sd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,20 @@ schema family_document {
stemming: none
}

field family_name_bolding type string {
indexing: input family_name_index | summary | index
bolding: true
}

field family_description_bolding type string {
indexing: input family_description_index | summary | index
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Any reason to hang onto the index attribute for these when they come from an index field?

bolding: true
}

field language type string {
indexing: "en" | set_language
}

Comment on lines +23 to +26
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As on the document passage, wondering if this needs to be at the top of the doc

document family_document {

field search_weights_ref type reference<search_weights> {
Expand Down Expand Up @@ -168,95 +182,61 @@ schema family_document {
fields: family_name_index, family_description_index
}

rank-profile exact inherits default {
function name_score() {
expression: attribute(name_weight) * fieldMatch(family_name_index)
}
function description_score() {
expression: attribute(description_weight) * fieldMatch(family_description_index)
}
first-phase {
expression: name_score() + description_score()
}
summary-features: name_score() description_score()
}

rank-profile exact_not_stemmed inherits default {
function name_score() {
expression: attribute(name_weight) * fieldMatch(family_name_not_stemmed)
expression: fieldMatch(family_name_not_stemmed)
}
function description_score() {
expression: attribute(description_weight) * fieldMatch(family_description_not_stemmed)
expression: fieldMatch(family_description_not_stemmed)
}
first-phase {
expression: name_score() + description_score()
expression: attribute(name_weight) * name_score() + attribute(description_weight) * description_score()
}
summary-features: name_score() description_score()
}

rank-profile hybrid_no_closeness inherits default {
function name_score() {
expression: attribute(name_weight) * bm25(family_name_index)
}
function description_score() {
expression: attribute(description_weight) * bm25(family_description_index)
}
first-phase {
expression: name_score() + description_score()
}
summary-features: name_score() description_score()
summary-features: name_score() description_score() attribute(name_weight) attribute(description_weight)
}

rank-profile hybrid inherits default {
inputs {
query(query_embedding) tensor<float>(x[768])
query(description_bm25_weight) double: 1.0
query(description_closeness_weight) double: 1.0
}
function name_score() {
expression: attribute(name_weight) * bm25(family_name_index)
expression: bm25(family_name_index)
}
function description_score() {
expression: attribute(description_weight) * (bm25(family_description_index) + closeness(family_description_embedding))
expression: query(description_bm25_weight) * bm25(family_description_index) + query(descriptione_closeness_weight) * closeness(family_description_embedding)
}
first-phase {
expression: name_score() + description_score()
expression: (attribute(name_weight) * name_score()) + (attribute(description_weight) * description_score())
}
summary-features: name_score() description_score()
summary-features: name_score() description_score() bm25(family_name_index) bm25(family_description_index) closeness(family_description_embedding)
}

rank-profile hybrid_no_description_embedding inherits default {
rank-profile hybrid_nativerank inherits default {
inputs {
query(query_embedding) tensor<float>(x[768])
query(description_nativerank_weight) double: 1.0
query(description_closeness_weight) double: 1.0
}
function name_score() {
expression: attribute(name_weight) * bm25(family_name_index)
expression: nativeRank(family_name_index)
}
function description_score() {
expression: attribute(description_weight) * bm25(family_description_index)
expression: query(description_nativerank_weight) * nativeRank(family_description_index) + query(descriptione_closeness_weight) * closeness(family_description_embedding)
}
first-phase {
expression: name_score() + description_score()
expression: (attribute(name_weight) * name_score()) + (attribute(description_weight) * description_score())
}
summary-features: name_score() description_score()
summary-features: name_score() description_score() nativeRank(family_name_index) nativeRank(family_description_index) closeness(family_description_embedding)
}

rank-profile hybrid_custom_weight inherits default {
rank-profile hybrid_no_closeness inherits hybrid {
inputs {
query(query_embedding) tensor<float>(x[768])
query(bm25_weight) double
}
function name_score() {
expression: attribute(name_weight) * bm25(family_name_index)
}
function description_score() {
expression: attribute(description_weight) * bm25(family_description_index)
query(description_closeness_weight) double: 0.0
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Getting my head around this, does setting this to 0.0 make it have no effect?

}
first-phase {
expression: name_score() + description_score()
}
summary-features: name_score() description_score()
}



document-summary search_summary {
summary family_name {}
summary family_description {}
Expand All @@ -281,28 +261,7 @@ schema family_document {
summary collection_summary {}
}

document-summary search_summary_with_tokens {
summary family_name {}
summary family_description {}
summary family_import_id {}
summary family_slug {}
summary family_category {}
summary family_publication_ts {}
summary family_geography {}
summary family_geographies {}
summary family_source {}
summary document_import_id {}
summary document_title {}
summary document_slug {}
summary document_languages {}
summary document_content_type {}
summary document_cdn_object {}
summary document_source_url {}
summary metadata {}
summary corpus_import_id {}
summary corpus_type_name {}
summary collection_title {}
summary collection_summary {}
document-summary search_summary_with_tokens inherits search_summary {
summary family_name_index {}
summary family_name_index_tokens {
source: family_name_index
Expand Down
Loading