Added support for the semantic_text field and semantic query type
#1881
Conversation
|
|
||
| name: str | ||
| summary: str | ||
| content: Any = dsl.mapped_field( |
There was a problem hiding this comment.
I'm typing the semantic text field as Any because it has different types during ingest and search. On ingest it is a plain string, while on search it is returned as an object with the original string in the text attribute. The object also includes an inference attribute with the autogenerated chunks and their embeddings.
miguelgrinberg
force-pushed
the
semantic-text-support
branch
from
99d091f to
bb99b1c
Compare
examples/async/semantic_text.py
Outdated
|
|
||
|
|
||
| async def search(query: str) -> dsl.AsyncSearch[WorkplaceDoc]: | ||
| return WorkplaceDoc.search()[:5].query( |
There was a problem hiding this comment.
nit: Doing [:5] before the query call is a bit confusing IMO.
examples/async/semantic_text.py
Outdated
|
|
||
| async def search(query: str) -> dsl.AsyncSearch[WorkplaceDoc]: | ||
| return WorkplaceDoc.search()[:5].query( | ||
| "semantic", |
There was a problem hiding this comment.
Any reason not to use the class here? One advantage is that it avoids typos.
There was a problem hiding this comment.
No reason at all, except that historically it has been preferred to use the names. I actually prefer the classes myself and started using them on my tests. I'll update this.
…#1881) (#1882) * Added support for the `semantic_text` field and `semantic` query type * Fix nltk code... again * feedback (cherry picked from commit 7fa4f8c) Co-authored-by: Miguel Grinberg <[email protected]>
…elastic#1881) * Added support for the `semantic_text` field and `semantic` query type * Fix nltk code... again * feedback
This change adds support for semantic text, introduced in Elasticsearch 8.15, plus an example application. Unfortunately running this application requires a somewhat beefy ES instance, so I'm not going to add integration tests.