From 534b0df391ad6d1ae54c6674453aec5a6c69203f Mon Sep 17 00:00:00 2001
From: Roman Rizzi <roman@discourse.org>
Date: Mon, 16 Dec 2024 09:55:39 -0300
Subject: [PATCH] REFACTOR: Separation of concerns for embedding generation.
 (#1027)

In a previous refactor, we moved the responsibility of querying and storing embeddings into the `Schema` class. Now, it's time for embedding generation.

The motivation behind these changes is to isolate vector characteristics in simple objects to later replace them with a DB-backed version, similar to what we did with LLM configs.
---
 app/jobs/regular/generate_embeddings.rb       |   4 +-
 app/jobs/regular/generate_rag_embeddings.rb   |   4 +-
 app/jobs/scheduled/embeddings_backfill.rb     |  27 +--
 lib/ai_bot/personas/persona.rb                |   6 +-
 lib/ai_bot/tool_runner.rb                     |   5 +-
 lib/ai_helper/semantic_categorizer.rb         |   6 +-
 lib/embeddings/schema.rb                      |  44 +++--
 lib/embeddings/semantic_related.rb            |   3 +-
 lib/embeddings/semantic_search.rb             |  18 +-
 lib/embeddings/strategies/truncation.rb       |  17 +-
 lib/embeddings/vector.rb                      |  76 +++++++++
 .../all_mpnet_base_v2.rb                      |   8 -
 lib/embeddings/vector_representations/base.rb |  92 ++--------
 .../vector_representations/bge_large_en.rb    |  21 +--
 .../vector_representations/bge_m3.rb          |   9 -
 .../vector_representations/gemini.rb          |   8 -
 .../multilingual_e5_large.rb                  |  31 ++--
 .../text_embedding_3_large.rb                 |   8 -
 .../text_embedding_3_small.rb                 |   8 -
 .../text_embedding_ada_002.rb                 |   8 -
 .../modules/ai_bot/personas/persona_spec.rb   |  12 +-
 .../ai_helper/semantic_categorizer_spec.rb    |   8 +-
 .../jobs/generate_embeddings_spec.rb          |  26 ++-
 spec/lib/modules/embeddings/schema_spec.rb    |  12 +-
 .../embeddings/semantic_search_spec.rb        |   4 +-
 .../embeddings/strategies/truncation_spec.rb  |  13 +-
 .../all_mpnet_base_v2_spec.rb                 |  17 --
 .../vector_representations/gemini_spec.rb     |  18 --
 .../multilingual_e5_large_spec.rb             |  21 ---
 .../text_embedding_3_large_spec.rb            |  22 ---
 .../text_embedding_3_small_spec.rb            |  15 --
 .../text_embedding_ada_002_spec.rb            |  15 --
 .../vector_rep_shared_examples.rb             | 115 -------------
 spec/lib/modules/embeddings/vector_spec.rb    | 160 ++++++++++++++++++
 spec/models/rag_document_fragment_spec.rb     |   6 +-
 .../embeddings/embeddings_controller_spec.rb  |   4 +-
 36 files changed, 375 insertions(+), 496 deletions(-)
 create mode 100644 lib/embeddings/vector.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/all_mpnet_base_v2_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/gemini_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/multilingual_e5_large_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/text_embedding_3_large_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/text_embedding_3_small_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/text_embedding_ada_002_spec.rb
 delete mode 100644 spec/lib/modules/embeddings/vector_representations/vector_rep_shared_examples.rb
 create mode 100644 spec/lib/modules/embeddings/vector_spec.rb

diff --git a/app/jobs/regular/generate_embeddings.rb b/app/jobs/regular/generate_embeddings.rb
index 4e16dc22..6cc71fed 100644
--- a/app/jobs/regular/generate_embeddings.rb
+++ b/app/jobs/regular/generate_embeddings.rb
@@ -16,9 +16,7 @@ module Jobs
       return if topic.private_message? && !SiteSetting.ai_embeddings_generate_for_pms
       return if post.raw.blank?
 
-      vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
-
-      vector_rep.generate_representation_from(target)
+      DiscourseAi::Embeddings::Vector.instance.generate_representation_from(target)
     end
   end
 end
diff --git a/app/jobs/regular/generate_rag_embeddings.rb b/app/jobs/regular/generate_rag_embeddings.rb
index 836ede8b..775b849f 100644
--- a/app/jobs/regular/generate_rag_embeddings.rb
+++ b/app/jobs/regular/generate_rag_embeddings.rb
@@ -8,11 +8,11 @@ module ::Jobs
     def execute(args)
       return if (fragments = RagDocumentFragment.where(id: args[:fragment_ids].to_a)).empty?
 
-      vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+      vector = DiscourseAi::Embeddings::Vector.instance
 
       # generate_representation_from checks compares the digest value to make sure
       # the embedding is only generated once per fragment unless something changes.
-      fragments.map { |fragment| vector_rep.generate_representation_from(fragment) }
+      fragments.map { |fragment| vector.generate_representation_from(fragment) }
 
       last_fragment = fragments.last
       target = last_fragment.target
diff --git a/app/jobs/scheduled/embeddings_backfill.rb b/app/jobs/scheduled/embeddings_backfill.rb
index 383c4c05..4fdbe5bd 100644
--- a/app/jobs/scheduled/embeddings_backfill.rb
+++ b/app/jobs/scheduled/embeddings_backfill.rb
@@ -20,7 +20,8 @@ module Jobs
 
       rebaked = 0
 
-      vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+      vector = DiscourseAi::Embeddings::Vector.instance
+      vector_def = vector.vdef
       table_name = DiscourseAi::Embeddings::Schema::TOPICS_TABLE
 
       topics =
@@ -30,19 +31,19 @@ module Jobs
           .where(deleted_at: nil)
           .order("topics.bumped_at DESC")
 
-      rebaked += populate_topic_embeddings(vector_rep, topics.limit(limit - rebaked))
+      rebaked += populate_topic_embeddings(vector, topics.limit(limit - rebaked))
 
       return if rebaked >= limit
 
       # Then, we'll try to backfill embeddings for topics that have outdated
       # embeddings, be it model or strategy version
       relation = topics.where(<<~SQL).limit(limit - rebaked)
-          #{table_name}.model_version < #{vector_rep.version}
+          #{table_name}.model_version < #{vector_def.version}
           OR
-          #{table_name}.strategy_version < #{vector_rep.strategy_version}
+          #{table_name}.strategy_version < #{vector_def.strategy_version}
         SQL
 
-      rebaked += populate_topic_embeddings(vector_rep, relation)
+      rebaked += populate_topic_embeddings(vector, relation)
 
       return if rebaked >= limit
 
@@ -54,7 +55,7 @@ module Jobs
           .where("#{table_name}.updated_at < topics.updated_at")
           .limit((limit - rebaked) / 10)
 
-      populate_topic_embeddings(vector_rep, relation, force: true)
+      populate_topic_embeddings(vector, relation, force: true)
 
       return if rebaked >= limit
 
@@ -76,7 +77,7 @@ module Jobs
         .limit(limit - rebaked)
         .pluck(:id)
         .each_slice(posts_batch_size) do |batch|
-          vector_rep.gen_bulk_reprensentations(Post.where(id: batch))
+          vector.gen_bulk_reprensentations(Post.where(id: batch))
           rebaked += batch.length
         end
 
@@ -86,14 +87,14 @@ module Jobs
       # embeddings, be it model or strategy version
       posts
         .where(<<~SQL)
-          #{table_name}.model_version < #{vector_rep.version}
+          #{table_name}.model_version < #{vector_def.version}
           OR
-          #{table_name}.strategy_version < #{vector_rep.strategy_version}
+          #{table_name}.strategy_version < #{vector_def.strategy_version}
         SQL
         .limit(limit - rebaked)
         .pluck(:id)
         .each_slice(posts_batch_size) do |batch|
-          vector_rep.gen_bulk_reprensentations(Post.where(id: batch))
+          vector.gen_bulk_reprensentations(Post.where(id: batch))
           rebaked += batch.length
         end
 
@@ -107,7 +108,7 @@ module Jobs
         .limit((limit - rebaked) / 10)
         .pluck(:id)
         .each_slice(posts_batch_size) do |batch|
-          vector_rep.gen_bulk_reprensentations(Post.where(id: batch))
+          vector.gen_bulk_reprensentations(Post.where(id: batch))
           rebaked += batch.length
         end
 
@@ -116,7 +117,7 @@ module Jobs
 
     private
 
-    def populate_topic_embeddings(vector_rep, topics, force: false)
+    def populate_topic_embeddings(vector, topics, force: false)
       done = 0
 
       topics =
@@ -126,7 +127,7 @@ module Jobs
       batch_size = 1000
 
       ids.each_slice(batch_size) do |batch|
-        vector_rep.gen_bulk_reprensentations(Topic.where(id: batch).order("topics.bumped_at DESC"))
+        vector.gen_bulk_reprensentations(Topic.where(id: batch).order("topics.bumped_at DESC"))
         done += batch.length
       end
 
diff --git a/lib/ai_bot/personas/persona.rb b/lib/ai_bot/personas/persona.rb
index d3b7cdf4..1569ce45 100644
--- a/lib/ai_bot/personas/persona.rb
+++ b/lib/ai_bot/personas/persona.rb
@@ -314,10 +314,10 @@ module DiscourseAi
 
           return nil if !consolidated_question
 
-          vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+          vector = DiscourseAi::Embeddings::Vector.instance
           reranker = DiscourseAi::Inference::HuggingFaceTextEmbeddings
 
-          interactions_vector = vector_rep.vector_from(consolidated_question)
+          interactions_vector = vector.vector_from(consolidated_question)
 
           rag_conversation_chunks = self.class.rag_conversation_chunks
           search_limit =
@@ -327,7 +327,7 @@ module DiscourseAi
               rag_conversation_chunks
             end
 
-          schema = DiscourseAi::Embeddings::Schema.for(RagDocumentFragment, vector: vector_rep)
+          schema = DiscourseAi::Embeddings::Schema.for(RagDocumentFragment, vector_def: vector.vdef)
 
           candidate_fragment_ids =
             schema
diff --git a/lib/ai_bot/tool_runner.rb b/lib/ai_bot/tool_runner.rb
index 74c92767..17440ead 100644
--- a/lib/ai_bot/tool_runner.rb
+++ b/lib/ai_bot/tool_runner.rb
@@ -141,11 +141,10 @@ module DiscourseAi
 
         return [] if upload_refs.empty?
 
-        vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
-        query_vector = vector_rep.vector_from(query)
+        query_vector = DiscourseAi::Embeddings::Vector.instance.vector_from(query)
         fragment_ids =
           DiscourseAi::Embeddings::Schema
-            .for(RagDocumentFragment, vector: vector_rep)
+            .for(RagDocumentFragment)
             .asymmetric_similarity_search(query_vector, limit: limit, offset: 0) do |builder|
               builder.join(<<~SQL, target_id: tool.id, target_type: "AiTool")
                 rag_document_fragments ON
diff --git a/lib/ai_helper/semantic_categorizer.rb b/lib/ai_helper/semantic_categorizer.rb
index fc39c264..b226b01b 100644
--- a/lib/ai_helper/semantic_categorizer.rb
+++ b/lib/ai_helper/semantic_categorizer.rb
@@ -92,10 +92,10 @@ module DiscourseAi
       private
 
       def nearest_neighbors(limit: 100)
-        vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
-        schema = DiscourseAi::Embeddings::Schema.for(Topic, vector: vector_rep)
+        vector = DiscourseAi::Embeddings::Vector.instance
+        schema = DiscourseAi::Embeddings::Schema.for(Topic, vector_def: vector.vdef)
 
-        raw_vector = vector_rep.vector_from(@text)
+        raw_vector = vector.vector_from(@text)
 
         muted_category_ids = nil
         if @user.present?
diff --git a/lib/embeddings/schema.rb b/lib/embeddings/schema.rb
index bf878fd1..3ae51c7b 100644
--- a/lib/embeddings/schema.rb
+++ b/lib/embeddings/schema.rb
@@ -14,30 +14,31 @@ module DiscourseAi
 
       def self.for(
         target_klass,
-        vector: DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+        vector_def: DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
       )
         case target_klass&.name
         when "Topic"
-          new(TOPICS_TABLE, "topic_id", vector)
+          new(TOPICS_TABLE, "topic_id", vector_def)
         when "Post"
-          new(POSTS_TABLE, "post_id", vector)
+          new(POSTS_TABLE, "post_id", vector_def)
         when "RagDocumentFragment"
-          new(RAG_DOCS_TABLE, "rag_document_fragment_id", vector)
+          new(RAG_DOCS_TABLE, "rag_document_fragment_id", vector_def)
         else
           raise ArgumentError, "Invalid target type for embeddings"
         end
       end
 
-      def initialize(table, target_column, vector)
+      def initialize(table, target_column, vector_def)
         @table = table
         @target_column = target_column
-        @vector = vector
+        @vector_def = vector_def
       end
 
-      attr_reader :table, :target_column, :vector
+      attr_reader :table, :target_column, :vector_def
 
       def find_by_embedding(embedding)
-        DB.query(<<~SQL, query_embedding: embedding, vid: vector.id, vsid: vector.strategy_id).first
+        DB.query(
+          <<~SQL,
           SELECT *
           FROM #{table}
           WHERE
@@ -46,10 +47,15 @@ module DiscourseAi
             embeddings::halfvec(#{dimensions}) #{pg_function} '[:query_embedding]'::halfvec(#{dimensions})
           LIMIT 1
         SQL
+          query_embedding: embedding,
+          vid: vector_def.id,
+          vsid: vector_def.strategy_id,
+        ).first
       end
 
       def find_by_target(target)
-        DB.query(<<~SQL, target_id: target.id, vid: vector.id, vsid: vector.strategy_id).first
+        DB.query(
+          <<~SQL,
           SELECT *
           FROM #{table}
           WHERE
@@ -58,6 +64,10 @@ module DiscourseAi
             #{target_column} = :target_id
           LIMIT 1
         SQL
+          target_id: target.id,
+          vid: vector_def.id,
+          vsid: vector_def.strategy_id,
+        ).first
       end
 
       def asymmetric_similarity_search(embedding, limit:, offset:)
@@ -87,8 +97,8 @@ module DiscourseAi
 
         builder.where(
           "model_id = :model_id AND strategy_id = :strategy_id",
-          model_id: vector.id,
-          strategy_id: vector.strategy_id,
+          model_id: vector_def.id,
+          strategy_id: vector_def.strategy_id,
         )
 
         yield(builder) if block_given?
@@ -156,7 +166,7 @@ module DiscourseAi
 
         yield(builder) if block_given?
 
-        builder.query(vid: vector.id, vsid: vector.strategy_id, target_id: record.id)
+        builder.query(vid: vector_def.id, vsid: vector_def.strategy_id, target_id: record.id)
       rescue PG::Error => e
         Rails.logger.error("Error #{e} querying embeddings for model #{name}")
         raise MissingEmbeddingError
@@ -176,10 +186,10 @@ module DiscourseAi
             updated_at = :now
           SQL
           target_id: record.id,
-          model_id: vector.id,
-          model_version: vector.version,
-          strategy_id: vector.strategy_id,
-          strategy_version: vector.strategy_version,
+          model_id: vector_def.id,
+          model_version: vector_def.version,
+          strategy_id: vector_def.strategy_id,
+          strategy_version: vector_def.strategy_version,
           digest: digest,
           embeddings: embedding,
           now: Time.zone.now,
@@ -188,7 +198,7 @@ module DiscourseAi
 
       private
 
-      delegate :dimensions, :pg_function, to: :vector
+      delegate :dimensions, :pg_function, to: :vector_def
     end
   end
 end
diff --git a/lib/embeddings/semantic_related.rb b/lib/embeddings/semantic_related.rb
index c28d05fc..54c8e572 100644
--- a/lib/embeddings/semantic_related.rb
+++ b/lib/embeddings/semantic_related.rb
@@ -13,14 +13,13 @@ module DiscourseAi
       def related_topic_ids_for(topic)
         return [] if SiteSetting.ai_embeddings_semantic_related_topics < 1
 
-        vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
         cache_for = results_ttl(topic)
 
         Discourse
           .cache
           .fetch(semantic_suggested_key(topic.id), expires_in: cache_for) do
             DiscourseAi::Embeddings::Schema
-              .for(Topic, vector: vector_rep)
+              .for(Topic)
               .symmetric_similarity_search(topic)
               .map(&:topic_id)
               .tap do |candidate_ids|
diff --git a/lib/embeddings/semantic_search.rb b/lib/embeddings/semantic_search.rb
index 108239b9..490772fa 100644
--- a/lib/embeddings/semantic_search.rb
+++ b/lib/embeddings/semantic_search.rb
@@ -30,8 +30,8 @@ module DiscourseAi
         Discourse.cache.read(embedding_key).present?
       end
 
-      def vector_rep
-        @vector_rep ||= DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+      def vector
+        @vector ||= DiscourseAi::Embeddings::Vector.instance
       end
 
       def hyde_embedding(search_term)
@@ -52,16 +52,14 @@ module DiscourseAi
 
         Discourse
           .cache
-          .fetch(embedding_key, expires_in: 1.week) { vector_rep.vector_from(hypothetical_post) }
+          .fetch(embedding_key, expires_in: 1.week) { vector.vector_from(hypothetical_post) }
       end
 
       def embedding(search_term)
         digest = OpenSSL::Digest::SHA1.hexdigest(search_term)
         embedding_key = build_embedding_key(digest, "", SiteSetting.ai_embeddings_model)
 
-        Discourse
-          .cache
-          .fetch(embedding_key, expires_in: 1.week) { vector_rep.vector_from(search_term) }
+        Discourse.cache.fetch(embedding_key, expires_in: 1.week) { vector.vector_from(search_term) }
       end
 
       # this ensures the candidate topics are over selected
@@ -84,7 +82,7 @@ module DiscourseAi
 
         over_selection_limit = limit * OVER_SELECTION_FACTOR
 
-        schema = DiscourseAi::Embeddings::Schema.for(Topic, vector: vector_rep)
+        schema = DiscourseAi::Embeddings::Schema.for(Topic, vector_def: vector.vdef)
 
         candidate_topic_ids =
           schema.asymmetric_similarity_search(
@@ -114,7 +112,7 @@ module DiscourseAi
 
         return [] if search_term.nil? || search_term.length < SiteSetting.min_search_term_length
 
-        vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+        vector = DiscourseAi::Embeddings::Vector.instance
 
         digest = OpenSSL::Digest::SHA1.hexdigest(search_term)
 
@@ -129,12 +127,12 @@ module DiscourseAi
           Discourse
             .cache
             .fetch(embedding_key, expires_in: 1.week) do
-              vector_rep.vector_from(search_term, asymetric: true)
+              vector.vector_from(search_term, asymetric: true)
             end
 
         candidate_post_ids =
           DiscourseAi::Embeddings::Schema
-            .for(Post, vector: vector_rep)
+            .for(Post, vector_def: vector.vdef)
             .asymmetric_similarity_search(
               search_term_embedding,
               limit: max_semantic_results_per_page,
diff --git a/lib/embeddings/strategies/truncation.rb b/lib/embeddings/strategies/truncation.rb
index 1ae82d80..cbe463c0 100644
--- a/lib/embeddings/strategies/truncation.rb
+++ b/lib/embeddings/strategies/truncation.rb
@@ -12,19 +12,28 @@ module DiscourseAi
           1
         end
 
-        def prepare_text_from(target, tokenizer, max_length)
+        def prepare_target_text(target, vdef)
+          max_length = vdef.max_sequence_length - 2
+
           case target
           when Topic
-            topic_truncation(target, tokenizer, max_length)
+            topic_truncation(target, vdef.tokenizer, max_length)
           when Post
-            post_truncation(target, tokenizer, max_length)
+            post_truncation(target, vdef.tokenizer, max_length)
           when RagDocumentFragment
-            tokenizer.truncate(target.fragment, max_length)
+            vdef.tokenizer.truncate(target.fragment, max_length)
           else
             raise ArgumentError, "Invalid target type"
           end
         end
 
+        def prepare_query_text(text, vdef, asymetric: false)
+          qtext = asymetric ? "#{vdef.asymmetric_query_prefix} #{text}" : text
+          max_length = vdef.max_sequence_length - 2
+
+          vdef.tokenizer.truncate(text, max_length)
+        end
+
         private
 
         def topic_information(topic)
diff --git a/lib/embeddings/vector.rb b/lib/embeddings/vector.rb
new file mode 100644
index 00000000..2fe8c72c
--- /dev/null
+++ b/lib/embeddings/vector.rb
@@ -0,0 +1,76 @@
+# frozen_string_literal: true
+
+module DiscourseAi
+  module Embeddings
+    class Vector
+      def self.instance
+        new(DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation)
+      end
+
+      def initialize(vector_definition)
+        @vdef = vector_definition
+      end
+
+      def gen_bulk_reprensentations(relation)
+        http_pool_size = 100
+        pool =
+          Concurrent::CachedThreadPool.new(
+            min_threads: 0,
+            max_threads: http_pool_size,
+            idletime: 30,
+          )
+
+        schema = DiscourseAi::Embeddings::Schema.for(relation.first.class, vector_def: vdef)
+
+        embedding_gen = vdef.inference_client
+        promised_embeddings =
+          relation
+            .map do |record|
+              prepared_text = vdef.prepare_target_text(record)
+              next if prepared_text.blank?
+
+              new_digest = OpenSSL::Digest::SHA1.hexdigest(prepared_text)
+              next if schema.find_by_target(record)&.digest == new_digest
+
+              Concurrent::Promises
+                .fulfilled_future({ target: record, text: prepared_text, digest: new_digest }, pool)
+                .then_on(pool) do |w_prepared_text|
+                  w_prepared_text.merge(embedding: embedding_gen.perform!(w_prepared_text[:text]))
+                end
+            end
+            .compact
+
+        Concurrent::Promises
+          .zip(*promised_embeddings)
+          .value!
+          .each { |e| schema.store(e[:target], e[:embedding], e[:digest]) }
+
+        pool.shutdown
+        pool.wait_for_termination
+      end
+
+      def generate_representation_from(target)
+        text = vdef.prepare_target_text(target)
+        return if text.blank?
+
+        schema = DiscourseAi::Embeddings::Schema.for(target.class, vector_def: vdef)
+
+        new_digest = OpenSSL::Digest::SHA1.hexdigest(text)
+        return if schema.find_by_target(target)&.digest == new_digest
+
+        embeddings = vdef.inference_client.perform!(text)
+
+        schema.store(target, embeddings, new_digest)
+      end
+
+      def vector_from(text, asymetric: false)
+        prepared_text = vdef.prepare_query_text(text, asymetric: asymetric)
+        return if prepared_text.blank?
+
+        vdef.inference_client.perform!(prepared_text)
+      end
+
+      attr_reader :vdef
+    end
+  end
+end
diff --git a/lib/embeddings/vector_representations/all_mpnet_base_v2.rb b/lib/embeddings/vector_representations/all_mpnet_base_v2.rb
index 7e4a2ad7..a89aab8b 100644
--- a/lib/embeddings/vector_representations/all_mpnet_base_v2.rb
+++ b/lib/embeddings/vector_representations/all_mpnet_base_v2.rb
@@ -23,10 +23,6 @@ module DiscourseAi
           end
         end
 
-        def vector_from(text, asymetric: false)
-          inference_client.perform!(text)
-        end
-
         def dimensions
           768
         end
@@ -47,10 +43,6 @@ module DiscourseAi
           "<#>"
         end
 
-        def pg_index_type
-          "halfvec_ip_ops"
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::AllMpnetBaseV2Tokenizer
         end
diff --git a/lib/embeddings/vector_representations/base.rb b/lib/embeddings/vector_representations/base.rb
index 94e33d8d..8670404b 100644
--- a/lib/embeddings/vector_representations/base.rb
+++ b/lib/embeddings/vector_representations/base.rb
@@ -21,8 +21,7 @@ module DiscourseAi
           end
 
           def current_representation
-            truncation = DiscourseAi::Embeddings::Strategies::Truncation.new
-            find_representation(SiteSetting.ai_embeddings_model).new(truncation)
+            find_representation(SiteSetting.ai_embeddings_model).new
           end
 
           def correctly_configured?
@@ -43,73 +42,6 @@ module DiscourseAi
           end
         end
 
-        def initialize(strategy)
-          @strategy = strategy
-        end
-
-        def vector_from(text, asymetric: false)
-          raise NotImplementedError
-        end
-
-        def gen_bulk_reprensentations(relation)
-          http_pool_size = 100
-          pool =
-            Concurrent::CachedThreadPool.new(
-              min_threads: 0,
-              max_threads: http_pool_size,
-              idletime: 30,
-            )
-
-          schema = DiscourseAi::Embeddings::Schema.for(relation.first.class, vector: self)
-
-          embedding_gen = inference_client
-          promised_embeddings =
-            relation
-              .map do |record|
-                prepared_text = prepare_text(record)
-                next if prepared_text.blank?
-
-                new_digest = OpenSSL::Digest::SHA1.hexdigest(prepared_text)
-                next if schema.find_by_target(record)&.digest == new_digest
-
-                Concurrent::Promises
-                  .fulfilled_future(
-                    { target: record, text: prepared_text, digest: new_digest },
-                    pool,
-                  )
-                  .then_on(pool) do |w_prepared_text|
-                    w_prepared_text.merge(embedding: embedding_gen.perform!(w_prepared_text[:text]))
-                  end
-              end
-              .compact
-
-          Concurrent::Promises
-            .zip(*promised_embeddings)
-            .value!
-            .each { |e| schema.store(e[:target], e[:embedding], e[:digest]) }
-
-          pool.shutdown
-          pool.wait_for_termination
-        end
-
-        def generate_representation_from(target, persist: true)
-          text = prepare_text(target)
-          return if text.blank?
-
-          schema = DiscourseAi::Embeddings::Schema.for(target.class, vector: self)
-
-          new_digest = OpenSSL::Digest::SHA1.hexdigest(text)
-          return if schema.find_by_target(target)&.digest == new_digest
-
-          vector = vector_from(text)
-
-          schema.store(target, vector, new_digest) if persist
-        end
-
-        def index_name(table_name)
-          "#{table_name}_#{id}_#{@strategy.id}_search"
-        end
-
         def name
           raise NotImplementedError
         end
@@ -139,26 +71,32 @@ module DiscourseAi
         end
 
         def asymmetric_query_prefix
-          raise NotImplementedError
+          ""
         end
 
         def strategy_id
-          @strategy.id
+          strategy.id
         end
 
         def strategy_version
-          @strategy.version
+          strategy.version
         end
 
-        protected
+        def prepare_query_text(text, asymetric: false)
+          strategy.prepare_query_text(text, self, asymetric: asymetric)
+        end
+
+        def prepare_target_text(target)
+          strategy.prepare_target_text(target, self)
+        end
+
+        def strategy
+          @strategy ||= DiscourseAi::Embeddings::Strategies::Truncation.new
+        end
 
         def inference_client
           raise NotImplementedError
         end
-
-        def prepare_text(record)
-          @strategy.prepare_text_from(record, tokenizer, max_sequence_length - 2)
-        end
       end
     end
   end
diff --git a/lib/embeddings/vector_representations/bge_large_en.rb b/lib/embeddings/vector_representations/bge_large_en.rb
index 923ee19e..9006ebbe 100644
--- a/lib/embeddings/vector_representations/bge_large_en.rb
+++ b/lib/embeddings/vector_representations/bge_large_en.rb
@@ -30,21 +30,6 @@ module DiscourseAi
           end
         end
 
-        def vector_from(text, asymetric: false)
-          text = "#{asymmetric_query_prefix} #{text}" if asymetric
-
-          client = inference_client
-
-          needs_truncation = client.class.name.include?("HuggingFaceTextEmbeddings")
-          text = tokenizer.truncate(text, max_sequence_length - 2) if needs_truncation
-
-          inference_client.perform!(text)
-        end
-
-        def inference_model_name
-          "baai/bge-large-en-v1.5"
-        end
-
         def dimensions
           1024
         end
@@ -65,10 +50,6 @@ module DiscourseAi
           "<#>"
         end
 
-        def pg_index_type
-          "halfvec_ip_ops"
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::BgeLargeEnTokenizer
         end
@@ -78,6 +59,8 @@ module DiscourseAi
         end
 
         def inference_client
+          inference_model_name = "baai/bge-large-en-v1.5"
+
           if SiteSetting.ai_cloudflare_workers_api_token.present?
             DiscourseAi::Inference::CloudflareWorkersAi.instance(inference_model_name)
           elsif DiscourseAi::Inference::HuggingFaceTextEmbeddings.configured?
diff --git a/lib/embeddings/vector_representations/bge_m3.rb b/lib/embeddings/vector_representations/bge_m3.rb
index d7e963fc..d3625e41 100644
--- a/lib/embeddings/vector_representations/bge_m3.rb
+++ b/lib/embeddings/vector_representations/bge_m3.rb
@@ -18,11 +18,6 @@ module DiscourseAi
           end
         end
 
-        def vector_from(text, asymetric: false)
-          truncated_text = tokenizer.truncate(text, max_sequence_length - 2)
-          inference_client.perform!(truncated_text)
-        end
-
         def dimensions
           1024
         end
@@ -43,10 +38,6 @@ module DiscourseAi
           "<#>"
         end
 
-        def pg_index_type
-          "halfvec_ip_ops"
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::BgeM3Tokenizer
         end
diff --git a/lib/embeddings/vector_representations/gemini.rb b/lib/embeddings/vector_representations/gemini.rb
index 5ac1cba7..110b8b4c 100644
--- a/lib/embeddings/vector_representations/gemini.rb
+++ b/lib/embeddings/vector_representations/gemini.rb
@@ -38,14 +38,6 @@ module DiscourseAi
           "<=>"
         end
 
-        def pg_index_type
-          "halfvec_cosine_ops"
-        end
-
-        def vector_from(text, asymetric: false)
-          inference_client.perform!(text)
-        end
-
         # There is no public tokenizer for Gemini, and from the ones we already ship in the plugin
         # OpenAI gets the closest results. Gemini Tokenizer results in ~10% less tokens, so it's safe
         # to use OpenAI tokenizer since it will overestimate the number of tokens.
diff --git a/lib/embeddings/vector_representations/multilingual_e5_large.rb b/lib/embeddings/vector_representations/multilingual_e5_large.rb
index fe611ec9..7d6894a8 100644
--- a/lib/embeddings/vector_representations/multilingual_e5_large.rb
+++ b/lib/embeddings/vector_representations/multilingual_e5_large.rb
@@ -28,19 +28,6 @@ module DiscourseAi
           end
         end
 
-        def vector_from(text, asymetric: false)
-          client = inference_client
-
-          needs_truncation = client.class.name.include?("HuggingFaceTextEmbeddings")
-          if needs_truncation
-            text = tokenizer.truncate(text, max_sequence_length - 2)
-          elsif !text.starts_with?("query:")
-            text = "query: #{text}"
-          end
-
-          client.perform!(text)
-        end
-
         def id
           3
         end
@@ -61,10 +48,6 @@ module DiscourseAi
           "<=>"
         end
 
-        def pg_index_type
-          "halfvec_cosine_ops"
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::MultilingualE5LargeTokenizer
         end
@@ -80,8 +63,18 @@ module DiscourseAi
           end
         end
 
-        def prepare_text(record)
-          prepared_text = super(record)
+        def prepare_text(text, asymetric: false)
+          prepared_text = super(text, asymetric: asymetric)
+
+          if prepared_text.present? && inference_client.class.name.include?("DiscourseClassifier")
+            return "query: #{prepared_text}"
+          end
+
+          prepared_text
+        end
+
+        def prepare_target_text(target)
+          prepared_text = super(target)
 
           if prepared_text.present? && inference_client.class.name.include?("DiscourseClassifier")
             return "query: #{prepared_text}"
diff --git a/lib/embeddings/vector_representations/text_embedding_3_large.rb b/lib/embeddings/vector_representations/text_embedding_3_large.rb
index 202d66de..d73f4cee 100644
--- a/lib/embeddings/vector_representations/text_embedding_3_large.rb
+++ b/lib/embeddings/vector_representations/text_embedding_3_large.rb
@@ -40,14 +40,6 @@ module DiscourseAi
           "<=>"
         end
 
-        def pg_index_type
-          "halfvec_cosine_ops"
-        end
-
-        def vector_from(text, asymetric: false)
-          inference_client.perform!(text)
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::OpenAiTokenizer
         end
diff --git a/lib/embeddings/vector_representations/text_embedding_3_small.rb b/lib/embeddings/vector_representations/text_embedding_3_small.rb
index 87f31185..90a3f790 100644
--- a/lib/embeddings/vector_representations/text_embedding_3_small.rb
+++ b/lib/embeddings/vector_representations/text_embedding_3_small.rb
@@ -38,14 +38,6 @@ module DiscourseAi
           "<=>"
         end
 
-        def pg_index_type
-          "halfvec_cosine_ops"
-        end
-
-        def vector_from(text, asymetric: false)
-          inference_client.perform!(text)
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::OpenAiTokenizer
         end
diff --git a/lib/embeddings/vector_representations/text_embedding_ada_002.rb b/lib/embeddings/vector_representations/text_embedding_ada_002.rb
index 1e570b98..f5340918 100644
--- a/lib/embeddings/vector_representations/text_embedding_ada_002.rb
+++ b/lib/embeddings/vector_representations/text_embedding_ada_002.rb
@@ -38,14 +38,6 @@ module DiscourseAi
           "<=>"
         end
 
-        def pg_index_type
-          "halfvec_cosine_ops"
-        end
-
-        def vector_from(text, asymetric: false)
-          inference_client.perform!(text)
-        end
-
         def tokenizer
           DiscourseAi::Tokenizer::OpenAiTokenizer
         end
diff --git a/spec/lib/modules/ai_bot/personas/persona_spec.rb b/spec/lib/modules/ai_bot/personas/persona_spec.rb
index 0ff4bd53..c58158d3 100644
--- a/spec/lib/modules/ai_bot/personas/persona_spec.rb
+++ b/spec/lib/modules/ai_bot/personas/persona_spec.rb
@@ -326,8 +326,8 @@ RSpec.describe DiscourseAi::AiBot::Personas::Persona do
       fab!(:llm_model) { Fabricate(:fake_model) }
 
       it "will run the question consolidator" do
-        vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
-        context_embedding = vector_rep.dimensions.times.map { rand(-1.0...1.0) }
+        vector_def = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+        context_embedding = vector_def.dimensions.times.map { rand(-1.0...1.0) }
         EmbeddingsGenerationStubs.discourse_service(
           SiteSetting.ai_embeddings_model,
           consolidated_question,
@@ -373,14 +373,14 @@ RSpec.describe DiscourseAi::AiBot::Personas::Persona do
     end
 
     context "when a persona has RAG uploads" do
-      let(:vector_rep) do
+      let(:vector_def) do
         DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
       end
       let(:embedding_value) { 0.04381 }
-      let(:prompt_cc_embeddings) { [embedding_value] * vector_rep.dimensions }
+      let(:prompt_cc_embeddings) { [embedding_value] * vector_def.dimensions }
 
       def stub_fragments(fragment_count, persona: ai_persona)
-        schema = DiscourseAi::Embeddings::Schema.for(RagDocumentFragment, vector: vector_rep)
+        schema = DiscourseAi::Embeddings::Schema.for(RagDocumentFragment, vector_def: vector_def)
 
         fragment_count.times do |i|
           fragment =
@@ -393,7 +393,7 @@ RSpec.describe DiscourseAi::AiBot::Personas::Persona do
             )
 
           # Similarity is determined left-to-right.
-          embeddings = [embedding_value + "0.000#{i}".to_f] * vector_rep.dimensions
+          embeddings = [embedding_value + "0.000#{i}".to_f] * vector_def.dimensions
 
           schema.store(fragment, embeddings, "test")
         end
diff --git a/spec/lib/modules/ai_helper/semantic_categorizer_spec.rb b/spec/lib/modules/ai_helper/semantic_categorizer_spec.rb
index 7a246864..8b89d800 100644
--- a/spec/lib/modules/ai_helper/semantic_categorizer_spec.rb
+++ b/spec/lib/modules/ai_helper/semantic_categorizer_spec.rb
@@ -14,9 +14,9 @@ RSpec.describe DiscourseAi::AiHelper::SemanticCategorizer do
   fab!(:category)
   fab!(:topic) { Fabricate(:topic, category: category) }
 
-  let(:vector_rep) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
+  let(:vector) { DiscourseAi::Embeddings::Vector.instance }
   let(:categorizer) { DiscourseAi::AiHelper::SemanticCategorizer.new({ text: "hello" }, user) }
-  let(:expected_embedding) { [0.0038493] * vector_rep.dimensions }
+  let(:expected_embedding) { [0.0038493] * vector.vdef.dimensions }
 
   before do
     SiteSetting.ai_embeddings_enabled = true
@@ -28,8 +28,8 @@ RSpec.describe DiscourseAi::AiHelper::SemanticCategorizer do
       "#{SiteSetting.ai_embeddings_discourse_service_api_endpoint}/api/v1/classify",
     ).to_return(status: 200, body: JSON.dump(expected_embedding))
 
-    vector_rep.generate_representation_from(topic)
-    vector_rep.generate_representation_from(muted_topic)
+    vector.generate_representation_from(topic)
+    vector.generate_representation_from(muted_topic)
   end
 
   it "respects user muted categories when making suggestions" do
diff --git a/spec/lib/modules/embeddings/jobs/generate_embeddings_spec.rb b/spec/lib/modules/embeddings/jobs/generate_embeddings_spec.rb
index 5b9f0fb3..56384d59 100644
--- a/spec/lib/modules/embeddings/jobs/generate_embeddings_spec.rb
+++ b/spec/lib/modules/embeddings/jobs/generate_embeddings_spec.rb
@@ -12,21 +12,16 @@ RSpec.describe Jobs::GenerateEmbeddings do
     fab!(:topic)
     fab!(:post) { Fabricate(:post, post_number: 1, topic: topic) }
 
-    let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-    let(:vector_rep) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
-    let(:topics_schema) { DiscourseAi::Embeddings::Schema.for(Topic, vector: vector_rep) }
-    let(:posts_schema) { DiscourseAi::Embeddings::Schema.for(Post, vector: vector_rep) }
+    let(:vector_def) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
+    let(:topics_schema) { DiscourseAi::Embeddings::Schema.for(Topic, vector_def: vector_def) }
+    let(:posts_schema) { DiscourseAi::Embeddings::Schema.for(Post, vector_def: vector_def) }
 
     it "works for topics" do
-      expected_embedding = [0.0038493] * vector_rep.dimensions
+      expected_embedding = [0.0038493] * vector_def.dimensions
 
-      text =
-        truncation.prepare_text_from(
-          topic,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-      EmbeddingsGenerationStubs.discourse_service(vector_rep.class.name, text, expected_embedding)
+      text = vector_def.prepare_target_text(topic)
+
+      EmbeddingsGenerationStubs.discourse_service(vector_def.class.name, text, expected_embedding)
 
       job.execute(target_id: topic.id, target_type: "Topic")
 
@@ -34,11 +29,10 @@ RSpec.describe Jobs::GenerateEmbeddings do
     end
 
     it "works for posts" do
-      expected_embedding = [0.0038493] * vector_rep.dimensions
+      expected_embedding = [0.0038493] * vector_def.dimensions
 
-      text =
-        truncation.prepare_text_from(post, vector_rep.tokenizer, vector_rep.max_sequence_length - 2)
-      EmbeddingsGenerationStubs.discourse_service(vector_rep.class.name, text, expected_embedding)
+      text = vector_def.prepare_target_text(post)
+      EmbeddingsGenerationStubs.discourse_service(vector_def.class.name, text, expected_embedding)
 
       job.execute(target_id: post.id, target_type: "Post")
 
diff --git a/spec/lib/modules/embeddings/schema_spec.rb b/spec/lib/modules/embeddings/schema_spec.rb
index 45c4e243..107428aa 100644
--- a/spec/lib/modules/embeddings/schema_spec.rb
+++ b/spec/lib/modules/embeddings/schema_spec.rb
@@ -1,16 +1,12 @@
 # frozen_string_literal: true
 
 RSpec.describe DiscourseAi::Embeddings::Schema do
-  subject(:posts_schema) { described_class.for(Post, vector: vector) }
+  subject(:posts_schema) { described_class.for(Post, vector_def: vector_def) }
 
-  let(:embeddings) { [0.0038490295] * vector.dimensions }
+  let(:embeddings) { [0.0038490295] * vector_def.dimensions }
   fab!(:post) { Fabricate(:post, post_number: 1) }
   let(:digest) { OpenSSL::Digest.hexdigest("SHA1", "test") }
-  let(:vector) do
-    DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2.new(
-      DiscourseAi::Embeddings::Strategies::Truncation.new,
-    )
-  end
+  let(:vector_def) { DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2.new }
 
   before { posts_schema.store(post, embeddings, digest) }
 
@@ -34,7 +30,7 @@ RSpec.describe DiscourseAi::Embeddings::Schema do
 
   describe "similarity searches" do
     fab!(:post_2) { Fabricate(:post) }
-    let(:similar_embeddings) { [0.0038490294] * vector.dimensions }
+    let(:similar_embeddings) { [0.0038490294] * vector_def.dimensions }
 
     describe "#symmetric_similarity_search" do
       before { posts_schema.store(post_2, similar_embeddings, digest) }
diff --git a/spec/lib/modules/embeddings/semantic_search_spec.rb b/spec/lib/modules/embeddings/semantic_search_spec.rb
index 67d5c275..2ed0bae9 100644
--- a/spec/lib/modules/embeddings/semantic_search_spec.rb
+++ b/spec/lib/modules/embeddings/semantic_search_spec.rb
@@ -11,8 +11,8 @@ RSpec.describe DiscourseAi::Embeddings::SemanticSearch do
 
   describe "#search_for_topics" do
     let(:hypothetical_post) { "This is an hypothetical post generated from the keyword test_query" }
-    let(:vector_rep) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
-    let(:hyde_embedding) { [0.049382] * vector_rep.dimensions }
+    let(:vector_def) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
+    let(:hyde_embedding) { [0.049382] * vector_def.dimensions }
 
     before do
       SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com"
diff --git a/spec/lib/modules/embeddings/strategies/truncation_spec.rb b/spec/lib/modules/embeddings/strategies/truncation_spec.rb
index b0498023..0b31fb11 100644
--- a/spec/lib/modules/embeddings/strategies/truncation_spec.rb
+++ b/spec/lib/modules/embeddings/strategies/truncation_spec.rb
@@ -3,8 +3,8 @@
 RSpec.describe DiscourseAi::Embeddings::Strategies::Truncation do
   subject(:truncation) { described_class.new }
 
-  describe "#prepare_text_from" do
-    context "when using vector from OpenAI" do
+  describe "#prepare_query_text" do
+    context "when using vector def from OpenAI" do
       before { SiteSetting.max_post_length = 100_000 }
 
       fab!(:topic)
@@ -20,15 +20,12 @@ RSpec.describe DiscourseAi::Embeddings::Strategies::Truncation do
       end
       fab!(:post) { Fabricate(:post, topic: topic, raw: "Surfin' bird\n" * 800) }
 
-      let(:model) do
-        DiscourseAi::Embeddings::VectorRepresentations::TextEmbeddingAda002.new(truncation)
-      end
+      let(:vector_def) { DiscourseAi::Embeddings::VectorRepresentations::TextEmbeddingAda002.new }
 
       it "truncates a topic" do
-        prepared_text =
-          truncation.prepare_text_from(topic, model.tokenizer, model.max_sequence_length)
+        prepared_text = truncation.prepare_target_text(topic, vector_def)
 
-        expect(model.tokenizer.size(prepared_text)).to be <= model.max_sequence_length
+        expect(vector_def.tokenizer.size(prepared_text)).to be <= vector_def.max_sequence_length
       end
     end
   end
diff --git a/spec/lib/modules/embeddings/vector_representations/all_mpnet_base_v2_spec.rb b/spec/lib/modules/embeddings/vector_representations/all_mpnet_base_v2_spec.rb
deleted file mode 100644
index c1cefe04..00000000
--- a/spec/lib/modules/embeddings/vector_representations/all_mpnet_base_v2_spec.rb
+++ /dev/null
@@ -1,17 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2 do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-
-  before { SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com" }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.discourse_service(described_class.name, text, expected_embedding)
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/gemini_spec.rb b/spec/lib/modules/embeddings/vector_representations/gemini_spec.rb
deleted file mode 100644
index 80f021f7..00000000
--- a/spec/lib/modules/embeddings/vector_representations/gemini_spec.rb
+++ /dev/null
@@ -1,18 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::Gemini do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-  let!(:api_key) { "test-123" }
-
-  before { SiteSetting.ai_gemini_api_key = api_key }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.gemini_service(api_key, text, expected_embedding)
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/multilingual_e5_large_spec.rb b/spec/lib/modules/embeddings/vector_representations/multilingual_e5_large_spec.rb
deleted file mode 100644
index e7af5eba..00000000
--- a/spec/lib/modules/embeddings/vector_representations/multilingual_e5_large_spec.rb
+++ /dev/null
@@ -1,21 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::MultilingualE5Large do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-
-  before { SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com" }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.discourse_service(
-      described_class.name,
-      "query: #{text}",
-      expected_embedding,
-    )
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/text_embedding_3_large_spec.rb b/spec/lib/modules/embeddings/vector_representations/text_embedding_3_large_spec.rb
deleted file mode 100644
index 5bed2863..00000000
--- a/spec/lib/modules/embeddings/vector_representations/text_embedding_3_large_spec.rb
+++ /dev/null
@@ -1,22 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::TextEmbedding3Large do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.openai_service(
-      described_class.name,
-      text,
-      expected_embedding,
-      extra_args: {
-        dimensions: 2000,
-      },
-    )
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/text_embedding_3_small_spec.rb b/spec/lib/modules/embeddings/vector_representations/text_embedding_3_small_spec.rb
deleted file mode 100644
index 1f4f01c2..00000000
--- a/spec/lib/modules/embeddings/vector_representations/text_embedding_3_small_spec.rb
+++ /dev/null
@@ -1,15 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::TextEmbedding3Small do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.openai_service(described_class.name, text, expected_embedding)
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/text_embedding_ada_002_spec.rb b/spec/lib/modules/embeddings/vector_representations/text_embedding_ada_002_spec.rb
deleted file mode 100644
index ed5a80fc..00000000
--- a/spec/lib/modules/embeddings/vector_representations/text_embedding_ada_002_spec.rb
+++ /dev/null
@@ -1,15 +0,0 @@
-# frozen_string_literal: true
-
-require_relative "vector_rep_shared_examples"
-
-RSpec.describe DiscourseAi::Embeddings::VectorRepresentations::TextEmbeddingAda002 do
-  subject(:vector_rep) { described_class.new(truncation) }
-
-  let(:truncation) { DiscourseAi::Embeddings::Strategies::Truncation.new }
-
-  def stub_vector_mapping(text, expected_embedding)
-    EmbeddingsGenerationStubs.openai_service(described_class.name, text, expected_embedding)
-  end
-
-  it_behaves_like "generates and store embedding using with vector representation"
-end
diff --git a/spec/lib/modules/embeddings/vector_representations/vector_rep_shared_examples.rb b/spec/lib/modules/embeddings/vector_representations/vector_rep_shared_examples.rb
deleted file mode 100644
index 0801860e..00000000
--- a/spec/lib/modules/embeddings/vector_representations/vector_rep_shared_examples.rb
+++ /dev/null
@@ -1,115 +0,0 @@
-# frozen_string_literal: true
-
-RSpec.shared_examples "generates and store embedding using with vector representation" do
-  let(:expected_embedding_1) { [0.0038493] * vector_rep.dimensions }
-  let(:expected_embedding_2) { [0.0037684] * vector_rep.dimensions }
-
-  let(:topics_schema) { DiscourseAi::Embeddings::Schema.for(Topic, vector: vector_rep) }
-  let(:posts_schema) { DiscourseAi::Embeddings::Schema.for(Post, vector: vector_rep) }
-
-  describe "#vector_from" do
-    it "creates a vector from a given string" do
-      text = "This is a piece of text"
-      stub_vector_mapping(text, expected_embedding_1)
-
-      expect(vector_rep.vector_from(text)).to eq(expected_embedding_1)
-    end
-  end
-
-  describe "#generate_representation_from" do
-    fab!(:topic) { Fabricate(:topic) }
-    fab!(:post) { Fabricate(:post, post_number: 1, topic: topic) }
-    fab!(:post2) { Fabricate(:post, post_number: 2, topic: topic) }
-
-    it "creates a vector from a topic and stores it in the database" do
-      text =
-        truncation.prepare_text_from(
-          topic,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-      stub_vector_mapping(text, expected_embedding_1)
-
-      vector_rep.generate_representation_from(topic)
-
-      expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
-    end
-
-    it "creates a vector from a post and stores it in the database" do
-      text =
-        truncation.prepare_text_from(
-          post2,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-      stub_vector_mapping(text, expected_embedding_1)
-
-      vector_rep.generate_representation_from(post)
-
-      expect(posts_schema.find_by_embedding(expected_embedding_1).post_id).to eq(post.id)
-    end
-  end
-
-  describe "#gen_bulk_reprensentations" do
-    fab!(:topic) { Fabricate(:topic) }
-    fab!(:post) { Fabricate(:post, post_number: 1, topic: topic) }
-    fab!(:post2) { Fabricate(:post, post_number: 2, topic: topic) }
-
-    fab!(:topic_2) { Fabricate(:topic) }
-    fab!(:post_2_1) { Fabricate(:post, post_number: 1, topic: topic_2) }
-    fab!(:post_2_2) { Fabricate(:post, post_number: 2, topic: topic_2) }
-
-    it "creates a vector for each object in the relation" do
-      text =
-        truncation.prepare_text_from(
-          topic,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-
-      text2 =
-        truncation.prepare_text_from(
-          topic_2,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-
-      stub_vector_mapping(text, expected_embedding_1)
-      stub_vector_mapping(text2, expected_embedding_2)
-
-      vector_rep.gen_bulk_reprensentations(Topic.where(id: [topic.id, topic_2.id]))
-
-      expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
-    end
-
-    it "does nothing if passed record has no content" do
-      expect { vector_rep.gen_bulk_reprensentations([Topic.new]) }.not_to raise_error
-    end
-
-    it "doesn't ask for a new embedding if digest is the same" do
-      text =
-        truncation.prepare_text_from(
-          topic,
-          vector_rep.tokenizer,
-          vector_rep.max_sequence_length - 2,
-        )
-      stub_vector_mapping(text, expected_embedding_1)
-
-      original_vector_gen = Time.zone.parse("2021-06-04 10:00")
-
-      freeze_time(original_vector_gen) do
-        vector_rep.gen_bulk_reprensentations(Topic.where(id: [topic.id]))
-      end
-      # check vector exists
-      expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
-
-      vector_rep.gen_bulk_reprensentations(Topic.where(id: [topic.id]))
-      last_update =
-        DB.query_single(
-          "SELECT updated_at FROM #{DiscourseAi::Embeddings::Schema::TOPICS_TABLE} WHERE topic_id = #{topic.id} LIMIT 1",
-        ).first
-
-      expect(last_update).to eq(original_vector_gen)
-    end
-  end
-end
diff --git a/spec/lib/modules/embeddings/vector_spec.rb b/spec/lib/modules/embeddings/vector_spec.rb
new file mode 100644
index 00000000..6542323d
--- /dev/null
+++ b/spec/lib/modules/embeddings/vector_spec.rb
@@ -0,0 +1,160 @@
+# frozen_string_literal: true
+
+RSpec.describe DiscourseAi::Embeddings::Vector do
+  shared_examples "generates and store embeddings using a vector definition" do
+    subject(:vector) { described_class.new(vdef) }
+
+    let(:expected_embedding_1) { [0.0038493] * vdef.dimensions }
+    let(:expected_embedding_2) { [0.0037684] * vdef.dimensions }
+
+    let(:topics_schema) { DiscourseAi::Embeddings::Schema.for(Topic, vector_def: vdef) }
+    let(:posts_schema) { DiscourseAi::Embeddings::Schema.for(Post, vector_def: vdef) }
+
+    fab!(:topic)
+    fab!(:post) { Fabricate(:post, post_number: 1, topic: topic) }
+    fab!(:post2) { Fabricate(:post, post_number: 2, topic: topic) }
+
+    describe "#vector_from" do
+      it "creates a vector from a given string" do
+        text = "This is a piece of text"
+        stub_vector_mapping(text, expected_embedding_1)
+
+        expect(vector.vector_from(text)).to eq(expected_embedding_1)
+      end
+    end
+
+    describe "#generate_representation_from" do
+      it "creates a vector from a topic and stores it in the database" do
+        text = vdef.prepare_target_text(topic)
+        stub_vector_mapping(text, expected_embedding_1)
+
+        vector.generate_representation_from(topic)
+
+        expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
+      end
+
+      it "creates a vector from a post and stores it in the database" do
+        text = vdef.prepare_target_text(post2)
+        stub_vector_mapping(text, expected_embedding_1)
+
+        vector.generate_representation_from(post)
+
+        expect(posts_schema.find_by_embedding(expected_embedding_1).post_id).to eq(post.id)
+      end
+    end
+
+    describe "#gen_bulk_reprensentations" do
+      fab!(:topic_2) { Fabricate(:topic) }
+      fab!(:post_2_1) { Fabricate(:post, post_number: 1, topic: topic_2) }
+      fab!(:post_2_2) { Fabricate(:post, post_number: 2, topic: topic_2) }
+
+      it "creates a vector for each object in the relation" do
+        text = vdef.prepare_target_text(topic)
+
+        text2 = vdef.prepare_target_text(topic_2)
+
+        stub_vector_mapping(text, expected_embedding_1)
+        stub_vector_mapping(text2, expected_embedding_2)
+
+        vector.gen_bulk_reprensentations(Topic.where(id: [topic.id, topic_2.id]))
+
+        expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
+      end
+
+      it "does nothing if passed record has no content" do
+        expect { vector.gen_bulk_reprensentations([Topic.new]) }.not_to raise_error
+      end
+
+      it "doesn't ask for a new embedding if digest is the same" do
+        text = vdef.prepare_target_text(topic)
+        stub_vector_mapping(text, expected_embedding_1)
+
+        original_vector_gen = Time.zone.parse("2021-06-04 10:00")
+
+        freeze_time(original_vector_gen) do
+          vector.gen_bulk_reprensentations(Topic.where(id: [topic.id]))
+        end
+        # check vector exists
+        expect(topics_schema.find_by_embedding(expected_embedding_1).topic_id).to eq(topic.id)
+
+        vector.gen_bulk_reprensentations(Topic.where(id: [topic.id]))
+
+        expect(topics_schema.find_by_target(topic).updated_at).to eq_time(original_vector_gen)
+      end
+    end
+  end
+
+  context "with text-embedding-ada-002" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::TextEmbeddingAda002.new }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.openai_service(vdef.class.name, text, expected_embedding)
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+
+  context "with all all-mpnet-base-v2" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::AllMpnetBaseV2.new }
+
+    before { SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com" }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.discourse_service(vdef.class.name, text, expected_embedding)
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+
+  context "with gemini" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::Gemini.new }
+    let(:api_key) { "test-123" }
+
+    before { SiteSetting.ai_gemini_api_key = api_key }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.gemini_service(api_key, text, expected_embedding)
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+
+  context "with multilingual-e5-large" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::MultilingualE5Large.new }
+
+    before { SiteSetting.ai_embeddings_discourse_service_api_endpoint = "http://test.com" }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.discourse_service(vdef.class.name, text, expected_embedding)
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+
+  context "with text-embedding-3-large" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::TextEmbedding3Large.new }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.openai_service(
+        vdef.class.name,
+        text,
+        expected_embedding,
+        extra_args: {
+          dimensions: 2000,
+        },
+      )
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+
+  context "with text-embedding-3-small" do
+    let(:vdef) { DiscourseAi::Embeddings::VectorRepresentations::TextEmbedding3Small.new }
+
+    def stub_vector_mapping(text, expected_embedding)
+      EmbeddingsGenerationStubs.openai_service(vdef.class.name, text, expected_embedding)
+    end
+
+    it_behaves_like "generates and store embeddings using a vector definition"
+  end
+end
diff --git a/spec/models/rag_document_fragment_spec.rb b/spec/models/rag_document_fragment_spec.rb
index 77a0061d..afb95d83 100644
--- a/spec/models/rag_document_fragment_spec.rb
+++ b/spec/models/rag_document_fragment_spec.rb
@@ -74,7 +74,7 @@ RSpec.describe RagDocumentFragment do
   end
 
   describe ".indexing_status" do
-    let(:vector_rep) { DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation }
+    let(:vector) { DiscourseAi::Embeddings::Vector.instance }
 
     fab!(:rag_document_fragment_1) do
       Fabricate(:rag_document_fragment, upload: upload_1, target: persona)
@@ -84,7 +84,7 @@ RSpec.describe RagDocumentFragment do
       Fabricate(:rag_document_fragment, upload: upload_1, target: persona)
     end
 
-    let(:expected_embedding) { [0.0038493] * vector_rep.dimensions }
+    let(:expected_embedding) { [0.0038493] * vector.vdef.dimensions }
 
     before do
       SiteSetting.ai_embeddings_enabled = true
@@ -96,7 +96,7 @@ RSpec.describe RagDocumentFragment do
         "#{SiteSetting.ai_embeddings_discourse_service_api_endpoint}/api/v1/classify",
       ).to_return(status: 200, body: JSON.dump(expected_embedding))
 
-      vector_rep.generate_representation_from(rag_document_fragment_1)
+      vector.generate_representation_from(rag_document_fragment_1)
     end
 
     it "regenerates all embeddings if ai_embeddings_model changes" do
diff --git a/spec/requests/embeddings/embeddings_controller_spec.rb b/spec/requests/embeddings/embeddings_controller_spec.rb
index 39bcf136..e79039fb 100644
--- a/spec/requests/embeddings/embeddings_controller_spec.rb
+++ b/spec/requests/embeddings/embeddings_controller_spec.rb
@@ -19,14 +19,14 @@ describe DiscourseAi::Embeddings::EmbeddingsController do
     fab!(:post_in_subcategory) { Fabricate(:post, topic: topic_in_subcategory) }
 
     def index(topic)
-      vector_rep = DiscourseAi::Embeddings::VectorRepresentations::Base.current_representation
+      vector = DiscourseAi::Embeddings::Vector.instance
 
       stub_request(:post, "https://api.openai.com/v1/embeddings").to_return(
         status: 200,
         body: JSON.dump({ data: [{ embedding: [0.1] * 1536 }] }),
       )
 
-      vector_rep.generate_representation_from(topic)
+      vector.generate_representation_from(topic)
     end
 
     def stub_embedding(query)