Class: Langchain::Vectorsearch::Pgvector

Inherits:
Base
  • Object
show all
Defined in:
lib/langchainrb_overrides/vectorsearch/pgvector.rb

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(llm:) ⇒ Pgvector

Returns a new instance of Pgvector.

Parameters:

  • url (String) —

    The URL of the PostgreSQL database

  • index_name (String) —

    The name of the table to use for the index

  • llm (Object) —

    The LLM client to use

  • namespace (String) —

    The namespace to use for the index when inserting/querying



25
26
27
28
29
30
31
32
33
34
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 25

def initialize(llm:)
  # If the line below is called, the generator fails as calls to
  # LangchainrbRails.config.vectorsearch will generate an exception.
  # These happen in the template files.
  # depends_on "neighbor"

  @operator = OPERATORS[DEFAULT_OPERATOR]

  super(llm: llm)
end

Instance Attribute Details

#llm ⇒ Object (readonly)

The PostgreSQL vector search adapter

Gem requirements:

gem "pgvector", "~> 0.2"

Usage:

pgvector = Langchain::Vectorsearch::Pgvector.new(llm:)



18
19
20
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 18

def llm
  @llm
end

#model ⇒ Object

Returns the value of attribute model.



19
20
21
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 19

def model
  @model
end

#operator ⇒ Object (readonly)

The PostgreSQL vector search adapter

Gem requirements:

gem "pgvector", "~> 0.2"

Usage:

pgvector = Langchain::Vectorsearch::Pgvector.new(llm:)



18
19
20
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 18

def operator
  @operator
end

Instance Method Details

#add_texts(texts:, ids:) ⇒ Array<Integer>

Add a list of texts to the index

Parameters:

  • texts (Array<String>) —

    The texts to add to the index

  • ids (Array<String>) —

    The ids to add to the index, in the same order as the texts

Returns:

  • (Array<Integer>) —

    The the ids of the added texts.



40
41
42
43
44
45
46
47
48
49
50
51
52
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 40

def add_texts(texts:, ids:)
  embeddings = texts.map do |text|
    llm.embed(text: text).embedding
  end

  # I believe the records returned by #find must be in the
  # same order as the embeddings. I _think_ this works for uuid ids but didn't test
  # deeply.
  # TODO - implement find_each so we don't load all records into memory
  model.find(ids).each.with_index do |record, i|
    record.update_column(:embedding, embeddings[i])
  end
end

#ask(question:, k: 4) {|String| ... } ⇒ String

Ask a question and return the answer

Parameters:

  • question (String) —

    The question to ask

  • k (Integer) (defaults to: 4) —

    The number of results to have in context

Yields:

  • (String) —

    Stream responses back one String at a time

Returns:

  • (String) —

    The answer to the question



100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 100

def ask(question:, k: 4, &block)
  # Noisy as the embedding column has a lot of data
  ActiveRecord::Base.logger.silence do
    search_results = similarity_search(query: question, k: k)

    context = search_results.map do |result|
      result.as_vector
    end
    context = context.join("\n---\n")

    prompt = generate_rag_prompt(question: question, context: context)

    messages = [{role: "user", content: prompt}]
    llm.chat(messages: messages, &block)
  end
end

#create_default_schema ⇒ Object

Invoke a rake task that will create an initializer (‘config/initializers/langchain.rb`) file and db/migrations/* files



60
61
62
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 60

def create_default_schema
  Rake::Task["pgvector"].invoke
end

#destroy_default_schema ⇒ Object

Destroy default schema



65
66
67
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 65

def destroy_default_schema
  # Tell the user to rollback the migration
end

#similarity_search(query:, k: 4) ⇒ Array<Hash>

Search for similar texts in the index TODO - drop the named “query:” param so it is the same interface as #ask?

Parameters:

  • query (String) —

    The text to search for

  • k (Integer) (defaults to: 4) —

    The number of top results to return

Returns:

  • (Array<Hash>) —

    The results of the search



74
75
76
77
78
79
80
81
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 74

def similarity_search(query:, k: 4)
  embedding = llm.embed(text: query).embedding

  similarity_search_by_vector(
    embedding: embedding,
    k: k
  )
end

#similarity_search_by_vector(embedding:, k: 4) ⇒ Array<Hash>

Search for similar texts in the index by the passed in vector. You must generate your own vector using the same LLM that generated the embeddings stored in the Vectorsearch DB. TODO - drop the named “embedding:” param so it is the same interface as #ask?

Parameters:

  • embedding (Array<Float>) —

    The vector to search for

  • k (Integer) (defaults to: 4) —

    The number of top results to return

Returns:

  • (Array<Hash>) —

    The results of the search



89
90
91
92
93
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 89

def similarity_search_by_vector(embedding:, k: 4)
  model
    .nearest_neighbors(:embedding, embedding, distance: operator)
    .limit(k)
end

#update_texts(texts:, ids:) ⇒ Object 



54
55
56
 
# File 'lib/langchainrb_overrides/vectorsearch/pgvector.rb', line 54

def update_texts(texts:, ids:)
  add_texts(texts: texts, ids: ids)
end