Roman Rizzi 534b0df391
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.
2024-12-16 09:55:39 -03:00

57 lines
1.1 KiB
Ruby

# frozen_string_literal: true
module DiscourseAi
module Embeddings
module VectorRepresentations
class TextEmbedding3Large < Base
class << self
def name
"text-embedding-3-large"
end
def correctly_configured?
SiteSetting.ai_openai_api_key.present?
end
def dependant_setting_names
%w[ai_openai_api_key]
end
end
def id
7
end
def version
1
end
def dimensions
# real dimentions are 3072, but we only support up to 2000 in the
# indexes, so we downsample to 2000 via API
2000
end
def max_sequence_length
8191
end
def pg_function
"<=>"
end
def tokenizer
DiscourseAi::Tokenizer::OpenAiTokenizer
end
def inference_client
DiscourseAi::Inference::OpenAiEmbeddings.instance(
model: self.class.name,
dimensions: dimensions,
)
end
end
end
end
end