mirror of
https://github.com/discourse/discourse-ai.git
synced 2025-10-24 02:58:39 +00:00
2c0f535bab
* FEATURE: HyDE-powered semantic search. It relies on the new outlet added on discourse/discourse#23390 to display semantic search results in an unobtrusive way. We'll use a HyDE-backed approach for semantic search, which consists on generating an hypothetical document from a given keywords, which gets transformed into a vector and used in a asymmetric similarity topic search. This PR also reorganizes the internals to have less moving parts, maintaining one hierarchy of DAOish classes for vector-related operations like transformations and querying. Completions and vectors created by HyDE will remain cached on Redis for now, but we could later use Postgres instead. * Missing translation and rate limiting --------- Co-authored-by: Roman Rizzi <rizziromanalejandro@gmail.com>
106 lines
3.0 KiB
Ruby
106 lines
3.0 KiB
Ruby
# frozen_string_literal: true
|
|
|
|
require_relative "../../../support/embeddings_generation_stubs"
|
|
require_relative "../../../support/openai_completions_inference_stubs"
|
|
|
|
RSpec.describe DiscourseAi::Embeddings::SemanticSearch do
|
|
fab!(:post) { Fabricate(:post) }
|
|
fab!(:user) { Fabricate(:user) }
|
|
|
|
let(:query) { "test_query" }
|
|
let(:subject) { described_class.new(Guardian.new(user)) }
|
|
|
|
describe "#search_for_topics" do
|
|
let(:hypothetical_post) { "This is an hypothetical post generated from the keyword test_query" }
|
|
|
|
before do
|
|
SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com"
|
|
|
|
prompt = DiscourseAi::Embeddings::HydeGenerators::OpenAi.new.prompt(query)
|
|
OpenAiCompletionsInferenceStubs.stub_response(prompt, hypothetical_post)
|
|
|
|
hyde_embedding = [0.049382, 0.9999]
|
|
EmbeddingsGenerationStubs.discourse_service(
|
|
SiteSetting.ai_embeddings_model,
|
|
hypothetical_post,
|
|
hyde_embedding,
|
|
)
|
|
end
|
|
|
|
after { described_class.clear_cache_for(query) }
|
|
|
|
def stub_candidate_ids(candidate_ids)
|
|
DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2
|
|
.any_instance
|
|
.expects(:asymmetric_topics_similarity_search)
|
|
.returns(candidate_ids)
|
|
end
|
|
|
|
it "returns the first post of a topic included in the asymmetric search results" do
|
|
stub_candidate_ids([post.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).to contain_exactly(post)
|
|
end
|
|
|
|
describe "applies different scopes to the candidates" do
|
|
context "when the topic is not visible" do
|
|
it "returns an empty list" do
|
|
post.topic.update!(visible: false)
|
|
stub_candidate_ids([post.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).to be_empty
|
|
end
|
|
end
|
|
|
|
context "when the post is not public" do
|
|
it "returns an empty list" do
|
|
pm_post = Fabricate(:private_message_post)
|
|
stub_candidate_ids([pm_post.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).to be_empty
|
|
end
|
|
end
|
|
|
|
context "when the post type is not visible" do
|
|
it "returns an empty list" do
|
|
post.update!(post_type: Post.types[:whisper])
|
|
stub_candidate_ids([post.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).to be_empty
|
|
end
|
|
end
|
|
|
|
context "when the post is not the first post in the topic" do
|
|
it "returns an empty list" do
|
|
reply = Fabricate(:reply)
|
|
reply.topic.first_post.trash!
|
|
stub_candidate_ids([reply.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).to be_empty
|
|
end
|
|
end
|
|
|
|
context "when the post is not a candidate" do
|
|
it "doesn't include it in the results" do
|
|
post_2 = Fabricate(:post)
|
|
stub_candidate_ids([post.topic_id])
|
|
|
|
posts = subject.search_for_topics(query)
|
|
|
|
expect(posts).not_to include(post_2)
|
|
end
|
|
end
|
|
end
|
|
end
|
|
end
|