Class: Langchain::LLM::AwsBedrock
- Defined in:
- lib/langchain/llm/aws_bedrock.rb
Overview
LLM interface for Aws Bedrock APIs: docs.aws.amazon.com/bedrock/
Gem requirements:
gem 'aws-sdk-bedrockruntime', '~> 1.1'
Usage:
llm = Langchain::LLM::AwsBedrock.new(default_options: {})
Constant Summary collapse
- DEFAULTS =
{ chat_model: "anthropic.claude-3-5-sonnet-20240620-v1:0", completion_model: "anthropic.claude-v2:1", embedding_model: "amazon.titan-embed-text-v1", max_tokens_to_sample: 300, temperature: 1, top_k: 250, top_p: 0.999, stop_sequences: ["\n\nHuman:"], return_likelihoods: "NONE" }.freeze
- SUPPORTED_COMPLETION_PROVIDERS =
%i[ anthropic ai21 cohere meta ].freeze
- SUPPORTED_CHAT_COMPLETION_PROVIDERS =
%i[ anthropic ai21 mistral ].freeze
- SUPPORTED_EMBEDDING_PROVIDERS =
%i[ amazon cohere ].freeze
Instance Attribute Summary collapse
-
#client ⇒ Object
readonly
Returns the value of attribute client.
-
#defaults ⇒ Object
readonly
Returns the value of attribute defaults.
Instance Method Summary collapse
-
#chat(params = {}) {|Hash| ... } ⇒ Langchain::LLM::AnthropicResponse
Generate a chat completion for a given prompt Currently only configured to work with the Anthropic provider and the claude-3 model family.
-
#complete(prompt:, model: , **params) ⇒ Langchain::LLM::AnthropicResponse
Generate a completion for a given prompt.
-
#embed(text:, **params) ⇒ Langchain::LLM::AwsTitanResponse
Generate an embedding for a given text.
-
#initialize(aws_client_options: {}, default_options: {}) ⇒ AwsBedrock
constructor
A new instance of AwsBedrock.
Methods inherited from Base
#chat_parameters, #default_dimension, #default_dimensions, #summarize
Methods included from DependencyHelper
Constructor Details
#initialize(aws_client_options: {}, default_options: {}) ⇒ AwsBedrock
Returns a new instance of AwsBedrock.
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 45 def initialize(aws_client_options: {}, default_options: {}) depends_on "aws-sdk-bedrockruntime", req: "aws-sdk-bedrockruntime" @client = ::Aws::BedrockRuntime::Client.new(**) @defaults = DEFAULTS.merge() chat_parameters.update( model: {default: @defaults[:chat_model]}, temperature: {}, max_tokens: {default: @defaults[:max_tokens_to_sample]}, metadata: {}, system: {} ) chat_parameters.ignore(:n, :user) chat_parameters.remap(stop: :stop_sequences) end |
Instance Attribute Details
#client ⇒ Object (readonly)
Returns the value of attribute client.
25 26 27 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 25 def client @client end |
#defaults ⇒ Object (readonly)
Returns the value of attribute defaults.
25 26 27 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 25 def defaults @defaults end |
Instance Method Details
#chat(params = {}) {|Hash| ... } ⇒ Langchain::LLM::AnthropicResponse
Generate a chat completion for a given prompt Currently only configured to work with the Anthropic provider and the claude-3 model family
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 128 def chat(params = {}, &block) parameters = chat_parameters.to_params(params) parameters = compose_parameters(parameters, parameters[:model]) unless SUPPORTED_CHAT_COMPLETION_PROVIDERS.include?(provider_name(parameters[:model])) raise "Chat provider #{parameters[:model]} is not supported." end if block response_chunks = [] client.invoke_model_with_response_stream( model_id: parameters[:model], body: parameters.except(:model).to_json, content_type: "application/json", accept: "application/json" ) do |stream| stream.on_event do |event| chunk = JSON.parse(event.bytes) response_chunks << chunk yield chunk end end response_from_chunks(response_chunks) else response = client.invoke_model({ model_id: parameters[:model], body: parameters.except(:model).to_json, content_type: "application/json", accept: "application/json" }) parse_response(response, parameters[:model]) end end |
#complete(prompt:, model: , **params) ⇒ Langchain::LLM::AnthropicResponse
Generate a completion for a given prompt
91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 91 def complete( prompt:, model: @defaults[:completion_model], **params ) raise "Completion provider #{model} is not supported." unless SUPPORTED_COMPLETION_PROVIDERS.include?(provider_name(model)) parameters = compose_parameters(params, model) parameters[:prompt] = wrap_prompt prompt response = client.invoke_model({ model_id: model, body: parameters.to_json, content_type: "application/json", accept: "application/json" }) parse_response(response, model) end |
#embed(text:, **params) ⇒ Langchain::LLM::AwsTitanResponse
Generate an embedding for a given text
69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
# File 'lib/langchain/llm/aws_bedrock.rb', line 69 def (text:, **params) raise "Completion provider #{} is not supported." unless SUPPORTED_EMBEDDING_PROVIDERS.include?() parameters = params.merge(text:) response = client.invoke_model({ model_id: @defaults[:embedding_model], body: parameters.to_json, content_type: "application/json", accept: "application/json" }) response end |