langchainrb 0.6.17 → 0.6.19
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +15 -0
- data/README.md +18 -3
- data/lib/langchain/active_record/hooks.rb +14 -0
- data/lib/langchain/agent/react_agent.rb +1 -1
- data/lib/langchain/agent/sql_query_agent.rb +2 -2
- data/lib/langchain/chunk.rb +16 -0
- data/lib/langchain/chunker/base.rb +4 -0
- data/lib/langchain/chunker/recursive_text.rb +5 -2
- data/lib/langchain/chunker/semantic.rb +4 -1
- data/lib/langchain/chunker/sentence.rb +4 -2
- data/lib/langchain/chunker/text.rb +5 -2
- data/lib/langchain/conversation.rb +1 -1
- data/lib/langchain/llm/ai21.rb +4 -3
- data/lib/langchain/llm/anthropic.rb +3 -3
- data/lib/langchain/llm/cohere.rb +6 -5
- data/lib/langchain/llm/google_palm.rb +14 -10
- data/lib/langchain/llm/hugging_face.rb +4 -3
- data/lib/langchain/llm/llama_cpp.rb +1 -1
- data/lib/langchain/llm/ollama.rb +18 -6
- data/lib/langchain/llm/openai.rb +7 -6
- data/lib/langchain/llm/replicate.rb +6 -10
- data/lib/langchain/llm/response/ai21_response.rb +13 -0
- data/lib/langchain/llm/response/anthropic_response.rb +29 -0
- data/lib/langchain/llm/response/base_response.rb +79 -0
- data/lib/langchain/llm/response/cohere_response.rb +21 -0
- data/lib/langchain/llm/response/google_palm_response.rb +36 -0
- data/lib/langchain/llm/response/hugging_face_response.rb +13 -0
- data/lib/langchain/llm/response/ollama_response.rb +26 -0
- data/lib/langchain/llm/response/openai_response.rb +51 -0
- data/lib/langchain/llm/response/replicate_response.rb +28 -0
- data/lib/langchain/vectorsearch/base.rb +4 -7
- data/lib/langchain/vectorsearch/chroma.rb +13 -12
- data/lib/langchain/vectorsearch/elasticsearch.rb +147 -0
- data/lib/langchain/vectorsearch/hnswlib.rb +5 -5
- data/lib/langchain/vectorsearch/milvus.rb +5 -4
- data/lib/langchain/vectorsearch/pgvector.rb +12 -6
- data/lib/langchain/vectorsearch/pinecone.rb +14 -13
- data/lib/langchain/vectorsearch/qdrant.rb +9 -8
- data/lib/langchain/vectorsearch/weaviate.rb +9 -8
- data/lib/langchain/version.rb +1 -1
- data/lib/langchain.rb +5 -0
- metadata +27 -2
@@ -0,0 +1,29 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class AnthropicResponse < BaseResponse
|
5
|
+
def model
|
6
|
+
raw_response.dig("model")
|
7
|
+
end
|
8
|
+
|
9
|
+
def completion
|
10
|
+
completions.first
|
11
|
+
end
|
12
|
+
|
13
|
+
def completions
|
14
|
+
[raw_response.dig("completion")]
|
15
|
+
end
|
16
|
+
|
17
|
+
def stop_reason
|
18
|
+
raw_response.dig("stop_reason")
|
19
|
+
end
|
20
|
+
|
21
|
+
def stop
|
22
|
+
raw_response.dig("stop")
|
23
|
+
end
|
24
|
+
|
25
|
+
def log_id
|
26
|
+
raw_response.dig("log_id")
|
27
|
+
end
|
28
|
+
end
|
29
|
+
end
|
@@ -0,0 +1,79 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain
|
4
|
+
module LLM
|
5
|
+
class BaseResponse
|
6
|
+
attr_reader :raw_response, :model
|
7
|
+
|
8
|
+
def initialize(raw_response, model: nil)
|
9
|
+
@raw_response = raw_response
|
10
|
+
@model = model
|
11
|
+
end
|
12
|
+
|
13
|
+
# Returns the completion text
|
14
|
+
#
|
15
|
+
# @return [String]
|
16
|
+
#
|
17
|
+
def completion
|
18
|
+
raise NotImplementedError
|
19
|
+
end
|
20
|
+
|
21
|
+
# Returns the chat completion text
|
22
|
+
#
|
23
|
+
# @return [String]
|
24
|
+
#
|
25
|
+
def chat_completion
|
26
|
+
raise NotImplementedError
|
27
|
+
end
|
28
|
+
|
29
|
+
# Return the first embedding
|
30
|
+
#
|
31
|
+
# @return [Array<Float>]
|
32
|
+
def embedding
|
33
|
+
raise NotImplementedError
|
34
|
+
end
|
35
|
+
|
36
|
+
# Return the completion candidates
|
37
|
+
#
|
38
|
+
# @return [Array]
|
39
|
+
def completions
|
40
|
+
raise NotImplementedError
|
41
|
+
end
|
42
|
+
|
43
|
+
# Return the chat completion candidates
|
44
|
+
#
|
45
|
+
# @return [Array]
|
46
|
+
def chat_completions
|
47
|
+
raise NotImplementedError
|
48
|
+
end
|
49
|
+
|
50
|
+
# Return the embeddings
|
51
|
+
#
|
52
|
+
# @return [Array<Array>]
|
53
|
+
def embeddings
|
54
|
+
raise NotImplementedError
|
55
|
+
end
|
56
|
+
|
57
|
+
# Number of tokens utilized in the prompt
|
58
|
+
#
|
59
|
+
# @return [Integer]
|
60
|
+
def prompt_tokens
|
61
|
+
raise NotImplementedError
|
62
|
+
end
|
63
|
+
|
64
|
+
# Number of tokens utilized to generate the completion
|
65
|
+
#
|
66
|
+
# @return [Integer]
|
67
|
+
def completion_tokens
|
68
|
+
raise NotImplementedError
|
69
|
+
end
|
70
|
+
|
71
|
+
# Total number of tokens utilized
|
72
|
+
#
|
73
|
+
# @return [Integer]
|
74
|
+
def total_tokens
|
75
|
+
raise NotImplementedError
|
76
|
+
end
|
77
|
+
end
|
78
|
+
end
|
79
|
+
end
|
@@ -0,0 +1,21 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class CohereResponse < BaseResponse
|
5
|
+
def embedding
|
6
|
+
embeddings.first
|
7
|
+
end
|
8
|
+
|
9
|
+
def embeddings
|
10
|
+
raw_response.dig("embeddings")
|
11
|
+
end
|
12
|
+
|
13
|
+
def completions
|
14
|
+
raw_response.dig("generations")
|
15
|
+
end
|
16
|
+
|
17
|
+
def completion
|
18
|
+
completions&.dig(0, "text")
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|
@@ -0,0 +1,36 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class GooglePalmResponse < BaseResponse
|
5
|
+
attr_reader :prompt_tokens
|
6
|
+
|
7
|
+
def initialize(raw_response, model: nil, prompt_tokens: nil)
|
8
|
+
@prompt_tokens = prompt_tokens
|
9
|
+
super(raw_response, model: model)
|
10
|
+
end
|
11
|
+
|
12
|
+
def completion
|
13
|
+
completions&.dig(0, "output")
|
14
|
+
end
|
15
|
+
|
16
|
+
def embedding
|
17
|
+
embeddings.first
|
18
|
+
end
|
19
|
+
|
20
|
+
def completions
|
21
|
+
raw_response.dig("candidates")
|
22
|
+
end
|
23
|
+
|
24
|
+
def chat_completion
|
25
|
+
chat_completions&.dig(0, "content")
|
26
|
+
end
|
27
|
+
|
28
|
+
def chat_completions
|
29
|
+
raw_response.dig("candidates")
|
30
|
+
end
|
31
|
+
|
32
|
+
def embeddings
|
33
|
+
[raw_response.dig("embedding", "value")]
|
34
|
+
end
|
35
|
+
end
|
36
|
+
end
|
@@ -0,0 +1,26 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class OllamaResponse < BaseResponse
|
5
|
+
def initialize(raw_response, model: nil, prompt_tokens: nil)
|
6
|
+
@prompt_tokens = prompt_tokens
|
7
|
+
super(raw_response, model: model)
|
8
|
+
end
|
9
|
+
|
10
|
+
def completion
|
11
|
+
raw_response.first
|
12
|
+
end
|
13
|
+
|
14
|
+
def completions
|
15
|
+
raw_response.is_a?(String) ? [raw_response] : []
|
16
|
+
end
|
17
|
+
|
18
|
+
def embedding
|
19
|
+
embeddings.first
|
20
|
+
end
|
21
|
+
|
22
|
+
def embeddings
|
23
|
+
[raw_response&.dig("embedding")]
|
24
|
+
end
|
25
|
+
end
|
26
|
+
end
|
@@ -0,0 +1,51 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class OpenAIResponse < BaseResponse
|
5
|
+
def model
|
6
|
+
raw_response["model"]
|
7
|
+
end
|
8
|
+
|
9
|
+
def created_at
|
10
|
+
if raw_response.dig("created")
|
11
|
+
Time.at(raw_response.dig("created"))
|
12
|
+
end
|
13
|
+
end
|
14
|
+
|
15
|
+
def completion
|
16
|
+
completions&.dig(0, "message", "content")
|
17
|
+
end
|
18
|
+
|
19
|
+
def chat_completion
|
20
|
+
completion
|
21
|
+
end
|
22
|
+
|
23
|
+
def embedding
|
24
|
+
embeddings&.first
|
25
|
+
end
|
26
|
+
|
27
|
+
def completions
|
28
|
+
raw_response.dig("choices")
|
29
|
+
end
|
30
|
+
|
31
|
+
def chat_completions
|
32
|
+
raw_response.dig("choices")
|
33
|
+
end
|
34
|
+
|
35
|
+
def embeddings
|
36
|
+
raw_response.dig("data")&.map { |datum| datum.dig("embedding") }
|
37
|
+
end
|
38
|
+
|
39
|
+
def prompt_tokens
|
40
|
+
raw_response.dig("usage", "prompt_tokens")
|
41
|
+
end
|
42
|
+
|
43
|
+
def completion_tokens
|
44
|
+
raw_response.dig("usage", "completion_tokens")
|
45
|
+
end
|
46
|
+
|
47
|
+
def total_tokens
|
48
|
+
raw_response.dig("usage", "total_tokens")
|
49
|
+
end
|
50
|
+
end
|
51
|
+
end
|
@@ -0,0 +1,28 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::LLM
|
4
|
+
class ReplicateResponse < BaseResponse
|
5
|
+
def completions
|
6
|
+
# Response comes back as an array of strings, e.g.: ["Hi", "how ", "are ", "you?"]
|
7
|
+
# The first array element is missing a space at the end, so we add it manually
|
8
|
+
raw_response.output[0] += " "
|
9
|
+
[raw_response.output.join]
|
10
|
+
end
|
11
|
+
|
12
|
+
def completion
|
13
|
+
completions.first
|
14
|
+
end
|
15
|
+
|
16
|
+
def created_at
|
17
|
+
Time.parse(raw_response.created_at)
|
18
|
+
end
|
19
|
+
|
20
|
+
def embedding
|
21
|
+
embeddings.first
|
22
|
+
end
|
23
|
+
|
24
|
+
def embeddings
|
25
|
+
[raw_response.output]
|
26
|
+
end
|
27
|
+
end
|
28
|
+
end
|
@@ -1,7 +1,5 @@
|
|
1
1
|
# frozen_string_literal: true
|
2
2
|
|
3
|
-
require "forwardable"
|
4
|
-
|
5
3
|
module Langchain::Vectorsearch
|
6
4
|
# = Vector Databases
|
7
5
|
# A vector database a type of database that stores data as high-dimensional vectors, which are mathematical representations of features or attributes. Each vector has a certain number of dimensions, which can range from tens to thousands, depending on the complexity and granularity of the data.
|
@@ -9,11 +7,13 @@ module Langchain::Vectorsearch
|
|
9
7
|
# == Available vector databases
|
10
8
|
#
|
11
9
|
# - {Langchain::Vectorsearch::Chroma}
|
10
|
+
# - {Langchain::Vectorsearch::Elasticsearch}
|
11
|
+
# - {Langchain::Vectorsearch::Hnswlib}
|
12
12
|
# - {Langchain::Vectorsearch::Milvus}
|
13
|
+
# - {Langchain::Vectorsearch::Pgvector}
|
13
14
|
# - {Langchain::Vectorsearch::Pinecone}
|
14
15
|
# - {Langchain::Vectorsearch::Qdrant}
|
15
16
|
# - {Langchain::Vectorsearch::Weaviate}
|
16
|
-
# - {Langchain::Vectorsearch::Pgvector}
|
17
17
|
#
|
18
18
|
# == Usage
|
19
19
|
#
|
@@ -150,9 +150,6 @@ module Langchain::Vectorsearch
|
|
150
150
|
raise NotImplementedError, "#{self.class.name} does not support asking questions"
|
151
151
|
end
|
152
152
|
|
153
|
-
def_delegators :llm,
|
154
|
-
:default_dimension
|
155
|
-
|
156
153
|
# HyDE-style prompt
|
157
154
|
#
|
158
155
|
# @param [String] User's question
|
@@ -184,7 +181,7 @@ module Langchain::Vectorsearch
|
|
184
181
|
.flatten
|
185
182
|
.map do |path|
|
186
183
|
data = Langchain::Loader.new(path)&.load&.chunks
|
187
|
-
data.map { |chunk| chunk
|
184
|
+
data.map { |chunk| chunk.text }
|
188
185
|
end
|
189
186
|
|
190
187
|
texts.flatten!
|
@@ -5,7 +5,8 @@ module Langchain::Vectorsearch
|
|
5
5
|
#
|
6
6
|
# Wrapper around Chroma DB
|
7
7
|
#
|
8
|
-
# Gem requirements:
|
8
|
+
# Gem requirements:
|
9
|
+
# gem "chroma-db", "~> 0.6.0"
|
9
10
|
#
|
10
11
|
# Usage:
|
11
12
|
# chroma = Langchain::Vectorsearch::Chroma.new(url:, index_name:, llm:, llm_api_key:, api_key: nil)
|
@@ -28,15 +29,16 @@ module Langchain::Vectorsearch
|
|
28
29
|
end
|
29
30
|
|
30
31
|
# Add a list of texts to the index
|
31
|
-
# @param texts [Array] The list of texts to add
|
32
|
+
# @param texts [Array<String>] The list of texts to add
|
33
|
+
# @param ids [Array<String>] The list of ids to use for the texts (optional)
|
34
|
+
# @param metadatas [Array<Hash>] The list of metadata to use for the texts (optional)
|
32
35
|
# @return [Hash] The response from the server
|
33
|
-
def add_texts(texts:, ids: [])
|
36
|
+
def add_texts(texts:, ids: [], metadatas: [])
|
34
37
|
embeddings = Array(texts).map.with_index do |text, i|
|
35
38
|
::Chroma::Resources::Embedding.new(
|
36
39
|
id: ids[i] ? ids[i].to_s : SecureRandom.uuid,
|
37
|
-
embedding: llm.embed(text: text),
|
38
|
-
|
39
|
-
metadata: {}, # metadatas[index],
|
40
|
+
embedding: llm.embed(text: text).embedding,
|
41
|
+
metadata: metadatas[i] || {},
|
40
42
|
document: text # Do we actually need to store the whole original document?
|
41
43
|
)
|
42
44
|
end
|
@@ -45,13 +47,12 @@ module Langchain::Vectorsearch
|
|
45
47
|
collection.add(embeddings)
|
46
48
|
end
|
47
49
|
|
48
|
-
def update_texts(texts:, ids:)
|
50
|
+
def update_texts(texts:, ids:, metadatas: [])
|
49
51
|
embeddings = Array(texts).map.with_index do |text, i|
|
50
52
|
::Chroma::Resources::Embedding.new(
|
51
53
|
id: ids[i].to_s,
|
52
|
-
embedding: llm.embed(text: text),
|
53
|
-
|
54
|
-
metadata: [], # metadatas[index],
|
54
|
+
embedding: llm.embed(text: text).embedding,
|
55
|
+
metadata: metadatas[i] || {},
|
55
56
|
document: text # Do we actually need to store the whole original document?
|
56
57
|
)
|
57
58
|
end
|
@@ -85,7 +86,7 @@ module Langchain::Vectorsearch
|
|
85
86
|
query:,
|
86
87
|
k: 4
|
87
88
|
)
|
88
|
-
embedding = llm.embed(text: query)
|
89
|
+
embedding = llm.embed(text: query).embedding
|
89
90
|
|
90
91
|
similarity_search_by_vector(
|
91
92
|
embedding: embedding,
|
@@ -94,7 +95,7 @@ module Langchain::Vectorsearch
|
|
94
95
|
end
|
95
96
|
|
96
97
|
# Search for similar texts by embedding
|
97
|
-
# @param embedding [Array] The embedding to search for
|
98
|
+
# @param embedding [Array<Float>] The embedding to search for
|
98
99
|
# @param k [Integer] The number of results to return
|
99
100
|
# @return [Chroma::Resources::Embedding] The response from the server
|
100
101
|
def similarity_search_by_vector(
|
@@ -0,0 +1,147 @@
|
|
1
|
+
# frozen_string_literal: true
|
2
|
+
|
3
|
+
module Langchain::Vectorsearch
|
4
|
+
class Elasticsearch < Base
|
5
|
+
#
|
6
|
+
# Wrapper around Elasticsearch vector search capabilities.
|
7
|
+
#
|
8
|
+
# Setting up Elasticsearch:
|
9
|
+
# 1. Get Elasticsearch up and running with Docker: https://www.elastic.co/guide/en/elasticsearch/reference/current/docker.html
|
10
|
+
# 2. Copy the HTTP CA certificate SHA-256 fingerprint and set the ELASTICSEARCH_CA_FINGERPRINT environment variable
|
11
|
+
# 3. Set the ELASTICSEARCH_URL environment variable
|
12
|
+
#
|
13
|
+
# Gem requirements:
|
14
|
+
# gem "elasticsearch", "~> 8.0.0"
|
15
|
+
#
|
16
|
+
# Usage:
|
17
|
+
# llm = Langchain::LLM::OpenAI.new(api_key: ENV["OPENAI_API_KEY"])
|
18
|
+
# es = Langchain::Vectorsearch::Elasticsearch.new(
|
19
|
+
# url: ENV["ELASTICSEARCH_URL"],
|
20
|
+
# index_name: "docs",
|
21
|
+
# llm: llm,
|
22
|
+
# es_options: {
|
23
|
+
# transport_options: {ssl: {verify: false}},
|
24
|
+
# ca_fingerprint: ENV["ELASTICSEARCH_CA_FINGERPRINT"]
|
25
|
+
# }
|
26
|
+
# )
|
27
|
+
#
|
28
|
+
# es.create_default_schema
|
29
|
+
# es.add_texts(texts: ["..."])
|
30
|
+
# es.similarity_search(text: "...")
|
31
|
+
#
|
32
|
+
attr_accessor :es_client, :index_name, :options
|
33
|
+
|
34
|
+
def initialize(url:, index_name:, llm:, api_key: nil, es_options: {})
|
35
|
+
require "elasticsearch"
|
36
|
+
|
37
|
+
@options = {
|
38
|
+
url: url,
|
39
|
+
request_timeout: 20,
|
40
|
+
log: false
|
41
|
+
}.merge(es_options)
|
42
|
+
|
43
|
+
@es_client = ::Elasticsearch::Client.new(**options)
|
44
|
+
@index_name = index_name
|
45
|
+
|
46
|
+
super(llm: llm)
|
47
|
+
end
|
48
|
+
|
49
|
+
def add_texts(texts: [])
|
50
|
+
body = texts.map do |text|
|
51
|
+
[
|
52
|
+
{index: {_index: index_name}},
|
53
|
+
{input: text, input_vector: llm.embed(text: text).embedding}
|
54
|
+
]
|
55
|
+
end.flatten
|
56
|
+
|
57
|
+
es_client.bulk(body: body)
|
58
|
+
end
|
59
|
+
|
60
|
+
def update_texts(texts: [], ids: [])
|
61
|
+
body = texts.map.with_index do |text, i|
|
62
|
+
[
|
63
|
+
{index: {_index: index_name, _id: ids[i]}},
|
64
|
+
{input: text, input_vector: llm.embed(text: text).embedding}
|
65
|
+
]
|
66
|
+
end.flatten
|
67
|
+
|
68
|
+
es_client.bulk(body: body)
|
69
|
+
end
|
70
|
+
|
71
|
+
def create_default_schema
|
72
|
+
es_client.indices.create(
|
73
|
+
index: index_name,
|
74
|
+
body: default_schema
|
75
|
+
)
|
76
|
+
end
|
77
|
+
|
78
|
+
def delete_default_schema
|
79
|
+
es_client.indices.delete(
|
80
|
+
index: index_name
|
81
|
+
)
|
82
|
+
end
|
83
|
+
|
84
|
+
def default_vector_settings
|
85
|
+
{type: "dense_vector", dims: llm.default_dimension}
|
86
|
+
end
|
87
|
+
|
88
|
+
def vector_settings
|
89
|
+
options[:vector_settings] || default_vector_settings
|
90
|
+
end
|
91
|
+
|
92
|
+
def default_schema
|
93
|
+
{
|
94
|
+
mappings: {
|
95
|
+
properties: {
|
96
|
+
input: {
|
97
|
+
type: "text"
|
98
|
+
},
|
99
|
+
input_vector: vector_settings
|
100
|
+
}
|
101
|
+
}
|
102
|
+
}
|
103
|
+
end
|
104
|
+
|
105
|
+
def default_query(query_vector)
|
106
|
+
{
|
107
|
+
script_score: {
|
108
|
+
query: {match_all: {}},
|
109
|
+
script: {
|
110
|
+
source: "cosineSimilarity(params.query_vector, 'input_vector') + 1.0",
|
111
|
+
params: {
|
112
|
+
query_vector: query_vector
|
113
|
+
}
|
114
|
+
}
|
115
|
+
}
|
116
|
+
}
|
117
|
+
end
|
118
|
+
|
119
|
+
# TODO: Implement this
|
120
|
+
# def ask()
|
121
|
+
# end
|
122
|
+
|
123
|
+
def similarity_search(text: "", k: 10, query: {})
|
124
|
+
if text.empty? && query.empty?
|
125
|
+
raise "Either text or query should pass as an argument"
|
126
|
+
end
|
127
|
+
|
128
|
+
if query.empty?
|
129
|
+
query_vector = llm.embed(text: text).embedding
|
130
|
+
|
131
|
+
query = default_query(query_vector)
|
132
|
+
end
|
133
|
+
|
134
|
+
es_client.search(body: {query: query, size: k}).body
|
135
|
+
end
|
136
|
+
|
137
|
+
def similarity_search_by_vector(embedding: [], k: 10, query: {})
|
138
|
+
if embedding.empty? && query.empty?
|
139
|
+
raise "Either embedding or query should pass as an argument"
|
140
|
+
end
|
141
|
+
|
142
|
+
query = default_query(embedding) if query.empty?
|
143
|
+
|
144
|
+
es_client.search(body: {query: query, size: k}).body
|
145
|
+
end
|
146
|
+
end
|
147
|
+
end
|
@@ -35,15 +35,15 @@ module Langchain::Vectorsearch
|
|
35
35
|
#
|
36
36
|
# Add a list of texts and corresponding IDs to the index
|
37
37
|
#
|
38
|
-
# @param texts [Array] The list of texts to add
|
39
|
-
# @param ids [Array] The list of corresponding IDs (integers) to the texts
|
38
|
+
# @param texts [Array<String>] The list of texts to add
|
39
|
+
# @param ids [Array<Integer>] The list of corresponding IDs (integers) to the texts
|
40
40
|
# @return [Boolean] The response from the HNSW library
|
41
41
|
#
|
42
42
|
def add_texts(texts:, ids:)
|
43
43
|
resize_index(texts.size)
|
44
44
|
|
45
45
|
Array(texts).each_with_index do |text, i|
|
46
|
-
embedding = llm.embed(text: text)
|
46
|
+
embedding = llm.embed(text: text).embedding
|
47
47
|
|
48
48
|
client.add_point(embedding, ids[i])
|
49
49
|
end
|
@@ -64,7 +64,7 @@ module Langchain::Vectorsearch
|
|
64
64
|
query:,
|
65
65
|
k: 4
|
66
66
|
)
|
67
|
-
embedding = llm.embed(text: query)
|
67
|
+
embedding = llm.embed(text: query).embedding
|
68
68
|
|
69
69
|
similarity_search_by_vector(
|
70
70
|
embedding: embedding,
|
@@ -75,7 +75,7 @@ module Langchain::Vectorsearch
|
|
75
75
|
#
|
76
76
|
# Search for the K nearest neighbors of a given vector
|
77
77
|
#
|
78
|
-
# @param embedding [Array] The embedding to search for
|
78
|
+
# @param embedding [Array<Float>] The embedding to search for
|
79
79
|
# @param k [Integer] The number of results to return
|
80
80
|
# @return [Array] Results in the format `[[id1, distance3], [id2, distance2]]`
|
81
81
|
#
|
@@ -5,7 +5,8 @@ module Langchain::Vectorsearch
|
|
5
5
|
#
|
6
6
|
# Wrapper around Milvus REST APIs.
|
7
7
|
#
|
8
|
-
# Gem requirements:
|
8
|
+
# Gem requirements:
|
9
|
+
# gem "milvus", "~> 0.9.2"
|
9
10
|
#
|
10
11
|
# Usage:
|
11
12
|
# milvus = Langchain::Vectorsearch::Milvus.new(url:, index_name:, llm:, api_key:)
|
@@ -32,7 +33,7 @@ module Langchain::Vectorsearch
|
|
32
33
|
}, {
|
33
34
|
field_name: "vectors",
|
34
35
|
type: ::Milvus::DATA_TYPES["float_vector"],
|
35
|
-
field: Array(texts).map { |text| llm.embed(text: text) }
|
36
|
+
field: Array(texts).map { |text| llm.embed(text: text).embedding }
|
36
37
|
}
|
37
38
|
]
|
38
39
|
)
|
@@ -70,7 +71,7 @@ module Langchain::Vectorsearch
|
|
70
71
|
type_params: [
|
71
72
|
{
|
72
73
|
key: "dim",
|
73
|
-
value: default_dimension.to_s
|
74
|
+
value: llm.default_dimension.to_s
|
74
75
|
}
|
75
76
|
]
|
76
77
|
}
|
@@ -111,7 +112,7 @@ module Langchain::Vectorsearch
|
|
111
112
|
end
|
112
113
|
|
113
114
|
def similarity_search(query:, k: 4)
|
114
|
-
embedding = llm.embed(text: query)
|
115
|
+
embedding = llm.embed(text: query).embedding
|
115
116
|
|
116
117
|
similarity_search_by_vector(
|
117
118
|
embedding: embedding,
|
@@ -5,10 +5,12 @@ module Langchain::Vectorsearch
|
|
5
5
|
#
|
6
6
|
# The PostgreSQL vector search adapter
|
7
7
|
#
|
8
|
-
# Gem requirements:
|
8
|
+
# Gem requirements:
|
9
|
+
# gem "sequel", "~> 5.68.0"
|
10
|
+
# gem "pgvector", "~> 0.2"
|
9
11
|
#
|
10
12
|
# Usage:
|
11
|
-
#
|
13
|
+
# pgvector = Langchain::Vectorsearch::Pgvector.new(url:, index_name:, llm:, namespace: nil)
|
12
14
|
#
|
13
15
|
|
14
16
|
# The operators supported by the PostgreSQL vector search adapter
|
@@ -52,7 +54,7 @@ module Langchain::Vectorsearch
|
|
52
54
|
# the added or updated texts.
|
53
55
|
def upsert_texts(texts:, ids:)
|
54
56
|
data = texts.zip(ids).flat_map do |(text, id)|
|
55
|
-
{id: id, content: text, vectors: llm.embed(text: text).to_s, namespace: namespace}
|
57
|
+
{id: id, content: text, vectors: llm.embed(text: text).embedding.to_s, namespace: namespace}
|
56
58
|
end
|
57
59
|
# @db[table_name.to_sym].multi_insert(data, return: :primary_key)
|
58
60
|
@db[table_name.to_sym]
|
@@ -70,7 +72,7 @@ module Langchain::Vectorsearch
|
|
70
72
|
def add_texts(texts:, ids: nil)
|
71
73
|
if ids.nil? || ids.empty?
|
72
74
|
data = texts.map do |text|
|
73
|
-
{content: text, vectors: llm.embed(text: text).to_s, namespace: namespace}
|
75
|
+
{content: text, vectors: llm.embed(text: text).embedding.to_s, namespace: namespace}
|
74
76
|
end
|
75
77
|
|
76
78
|
@db[table_name.to_sym].multi_insert(data, return: :primary_key)
|
@@ -91,7 +93,7 @@ module Langchain::Vectorsearch
|
|
91
93
|
def create_default_schema
|
92
94
|
db.run "CREATE EXTENSION IF NOT EXISTS vector"
|
93
95
|
namespace_column = @namespace_column
|
94
|
-
vector_dimension = default_dimension
|
96
|
+
vector_dimension = llm.default_dimension
|
95
97
|
db.create_table? table_name.to_sym do
|
96
98
|
primary_key :id
|
97
99
|
text :content
|
@@ -110,7 +112,7 @@ module Langchain::Vectorsearch
|
|
110
112
|
# @param k [Integer] The number of top results to return
|
111
113
|
# @return [Array<Hash>] The results of the search
|
112
114
|
def similarity_search(query:, k: 4)
|
113
|
-
embedding = llm.embed(text: query)
|
115
|
+
embedding = llm.embed(text: query).embedding
|
114
116
|
|
115
117
|
similarity_search_by_vector(
|
116
118
|
embedding: embedding,
|
@@ -150,3 +152,7 @@ module Langchain::Vectorsearch
|
|
150
152
|
end
|
151
153
|
end
|
152
154
|
end
|
155
|
+
|
156
|
+
# Rails connection when configuring vectorsearch
|
157
|
+
# Update READMEs
|
158
|
+
# Rails migration to create a migration
|