rubycanusellm 0.3.0 → 0.3.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 76737cbccfa63deb73a8a13e57c7202d42eb0689595ddae1b1cedf122f262e91
-  data.tar.gz: 282fda500fb7f36bebcd7e4fc3683a25e264e483b63e264da5aeb21b5cf229ef
+  metadata.gz: bb98a91b1d857f47d5ed3905d0fc520175297366268125d382fe928a91a993a3
+  data.tar.gz: '0284438324a2fc028176c0405a3af011a0457617ca930abc7cbcc4012f204fa1'
 SHA512:
-  metadata.gz: 9ffb10c5216314c8ae733b3be09391197a8e785308c42555764e7bc0144e958638dfb8a253e5d818ed6ba77eb69ee07e1b9279f4a7aafd1aca6e7f0e26f8c1b1
-  data.tar.gz: de69c64ab2ee56df386974bd9f99ffbcbc374cd4661b6394ee3295338f02ecee6cab439fbcd0a6bd3b8d9e54b045ef065fda7a8e290d4f6c6a95cce69cdc7949
+  metadata.gz: 9b7d8455fca4b21d60597e8d49a765a5a210d6c314b855cceb9ca21c46924555ea2841a7bda802fb55cdfd7ad49721d9a49d2a67edafd37f8427da98ba1fe522
+  data.tar.gz: d7c5818f91bc9a7973b20df9a465deaa4c4878e59e48c0859ef18bc4f0a5c6e4e261ac96458d9a05af90a901cf2b6614fb2fe4588339e1598c6a9e9b403aad11

data/CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## [0.3.1] - 2026-04-03
+
+### Added
+
+- Voyage AI provider for embeddings (recommended by Anthropic)
+- Configurable embedding provider (`embedding_provider`, `embedding_api_key`)
+- Anthropic users can now use Voyage AI or OpenAI for embeddings
+- Embedding validation with clear error messages
+
 ## [0.2.0] - 2026-04-01
 
 ### Added

data/README.md

@@ -63,10 +63,11 @@ That's it. Same code, different provider.
 
 ## Supported Providers
 
-| Provider | Models | Status |
-|----------|--------|--------|
-| OpenAI | gpt-4o-mini, gpt-4o, etc. | ✅ |
-| Anthropic | claude-sonnet-4-20250514, etc. | ✅ |
+| Provider | Models | Status | Notes |
+|----------|--------|--------|-------|
+| OpenAI | gpt-4o-mini, gpt-4o, etc. | ✅ | Chat + Embeddings |
+| Anthropic | claude-sonnet-4-20250514, etc. | ✅ | Chat only |
+| Voyage AI | voyage-3.5, voyage-4, etc. | ✅ | Embeddings only |
 
 ## API Reference
 
@@ -77,6 +78,8 @@ RubyCanUseLLM.configure do |config|
   config.api_key = "your-key"        # required
   config.model = "gpt-4o-mini"       # optional, has sensible defaults
   config.timeout = 30                # optional, default 30s
+  config.embedding_provider = :voyage  # optional, for separate embedding provider
+  config.embedding_api_key = "key"     # required when embedding_provider is set
 end
 ```
 
@@ -118,6 +121,56 @@ response.total_tokens  # 15
 response.raw           # original provider response
 ```
 
+### Embeddings
+
+```ruby
+response = RubyCanUseLLM.embed("Hello world")
+response.embedding  # [0.1, 0.2, ...]
+response.tokens     # 3
+response.model      # "text-embedding-3-small"
+```
+
+**OpenAI users** — embeddings work out of the box, no extra config needed:
+```ruby
+RubyCanUseLLM.configure do |config|
+  config.provider = :openai
+  config.api_key = ENV["OPENAI_API_KEY"]
+end
+
+RubyCanUseLLM.embed("Hello world")
+```
+
+**Anthropic users with Voyage AI** (recommended by Anthropic):
+```ruby
+RubyCanUseLLM.configure do |config|
+  config.provider = :anthropic
+  config.api_key = ENV["ANTHROPIC_API_KEY"]
+  config.embedding_provider = :voyage
+  config.embedding_api_key = ENV["VOYAGE_API_KEY"]
+end
+
+RubyCanUseLLM.embed("Hello world")
+```
+
+**Anthropic users with OpenAI for embeddings:**
+```ruby
+RubyCanUseLLM.configure do |config|
+  config.provider = :anthropic
+  config.api_key = ENV["ANTHROPIC_API_KEY"]
+  config.embedding_provider = :openai
+  config.embedding_api_key = ENV["OPENAI_API_KEY"]
+end
+
+RubyCanUseLLM.embed("Hello world")
+```
+
+**Cosine similarity:**
+```ruby
+a = RubyCanUseLLM.embed("cat")
+b = RubyCanUseLLM.embed("dog")
+a.cosine_similarity(b.embedding)  # 0.87
+```
+
 ### Error Handling
 ```ruby
 begin
@@ -150,7 +203,9 @@ end
 - [x] `generate:completion` command
 - [x] v0.1.0 release
 - [x] Streaming support
-- [ ] Embeddings + `generate:embedding`
+- [x] Embeddings + configurable embedding provider
+- [x] Voyage AI provider (embeddings)
+- [ ] `generate:embedding` command
 - [ ] Mistral and Ollama providers
 - [ ] Tool calling
 

data/lib/rubycanusellm/configuration.rb

@@ -2,15 +2,18 @@
 
 module RubyCanUseLLM
   class Configuration
-    SUPPORTED_PROVIDERS = %i[openai anthropic].freeze
+    SUPPORTED_PROVIDERS = %i[openai anthropic voyage].freeze
+    EMBEDDING_PROVIDERS = %i[openai voyage].freeze
 
-    attr_accessor :provider, :api_key, :model, :timeout
+    attr_accessor :provider, :api_key, :model, :timeout, :embedding_provider, :embedding_api_key
 
     def initialize
       @provider = nil
       @api_key = nil
       @model = nil
       @timeout = 30
+      @embedding_provider = nil
+      @embedding_api_key = nil
     end
 
     def validate!
@@ -18,5 +21,15 @@ module RubyCanUseLLM
       raise Error, "api_key is required" if api_key.nil? || api_key.empty?
       raise Error, "Unknown provider: #{provider}. Supported: #{SUPPORTED_PROVIDERS.join(", ")}" unless SUPPORTED_PROVIDERS.include?(provider)
     end
+
+    def validate_embedding!
+      effective = embedding_provider || provider
+      unless EMBEDDING_PROVIDERS.include?(effective)
+        raise Error, "#{provider} does not support embeddings. Set config.embedding_provider to :openai or :voyage and provide config.embedding_api_key"
+      end
+      if embedding_provider && (embedding_api_key.nil? || embedding_api_key.empty?)
+        raise Error, "embedding_api_key is required when embedding_provider is set"
+      end
+    end
   end
 end

data/lib/rubycanusellm/providers/voyage.rb

@@ -0,0 +1,66 @@
+require "net/http"
+require "json"
+
+module RubyCanUseLLM
+  module Providers
+    class Voyage < Base
+      API_URL = "https://api.voyageai.com/v1/embeddings"
+
+      def chat(messages, **options)
+        raise Error, "Voyage AI only supports embeddings, not chat completions"
+      end
+
+      def embed(text, **options)
+        body = {
+          model: options[:model] || "voyage-3.5",
+          input: text
+        }
+        response = embedding_request(body)
+        parse_embedding(response)
+      end
+
+      private
+
+      def embedding_request(body)
+        uri = URI(API_URL)
+        http = Net::HTTP.new(uri.host, uri.port)
+        http.use_ssl = true
+        http.read_timeout = config.timeout
+
+        req = Net::HTTP::Post.new(uri)
+        req["Authorization"] = "Bearer #{config.api_key}"
+        req["Content-Type"] = "application/json"
+        req.body = body.to_json
+
+        handle_response(http.request(req))
+      rescue Net::ReadTimeout, Net::OpenTimeout
+        raise TimeoutError, "Request to Voyage AI timed out after #{config.timeout}s"
+      end
+
+      def handle_response(response)
+        case response.code.to_i
+        when 200
+          JSON.parse(response.body)
+        when 401
+          raise AuthenticationError, "Invalid Voyage AI API key"
+        when 429
+          raise RateLimitError, "Voyage AI rate limit exceeded"
+        else
+          raise ProviderError, "Voyage AI error (#{response.code}): #{response.body}"
+        end
+      end
+
+      def parse_embedding(data)
+        embedding = data.dig("data", 0, "embedding")
+        usage = data["usage"]
+
+        EmbeddingResponse.new(
+          embedding: embedding,
+          model: data["model"],
+          tokens: usage["total_tokens"],
+          raw: data
+        )
+      end
+    end
+  end
+end

data/lib/rubycanusellm/templates/config.rb.tt

@@ -13,4 +13,10 @@ RubyCanUseLLM.configure do |config|
 
   # Request timeout in seconds (default: 30)
   # config.timeout = 30
+
+  # Embedding provider (optional, defaults to main provider)
+  # Anthropic doesn't support embeddings natively.
+  # Use :voyage (recommended by Anthropic) or :openai for embeddings.
+  # config.embedding_provider = :voyage
+  # config.embedding_api_key = ENV["VOYAGE_API_KEY"]
 end

data/lib/rubycanusellm/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Rubycanusellm
-  VERSION = "0.3.0"
+  VERSION = "0.3.1"
 end

data/lib/rubycanusellm.rb

@@ -8,12 +8,14 @@ require_relative "rubycanusellm/chunk"
 require_relative "rubycanusellm/providers/base"
 require_relative "rubycanusellm/providers/openai"
 require_relative "rubycanusellm/providers/anthropic"
+require_relative "rubycanusellm/providers/voyage"
 require_relative "rubycanusellm/embedding_response"
 
 module RubyCanUseLLM
   PROVIDERS = {
     openai: Providers::OpenAI,
-    anthropic: Providers::Anthropic
+    anthropic: Providers::Anthropic,
+    voyage: Providers::Voyage
   }.freeze
 
   class << self
@@ -28,6 +30,7 @@ module RubyCanUseLLM
     def reset!
       @configuration = Configuration.new
       @client = nil
+      @embedding_client = nil
     end
 
     def client
@@ -37,12 +40,27 @@ module RubyCanUseLLM
       end.new(configuration)
     end
 
+    def embedding_client
+      configuration.validate_embedding!
+      if configuration.embedding_provider
+        @embedding_client ||= begin
+          cfg = Configuration.new
+          cfg.provider = configuration.embedding_provider
+          cfg.api_key = configuration.embedding_api_key
+          cfg.timeout = configuration.timeout
+          PROVIDERS[configuration.embedding_provider].new(cfg)
+        end
+      else
+        client
+      end
+    end
+
     def chat(messages, **options, &block)
       client.chat(messages, **options, &block)
     end
 
     def embed(text, **options)
-      client.embed(text, **options)
+      embedding_client.embed(text, **options)
     end
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: rubycanusellm
 version: !ruby/object:Gem::Version
-  version: 0.3.0
+  version: 0.3.1
 platform: ruby
 authors:
 - Juan Manuel Guzman Nava
@@ -36,6 +36,7 @@ files:
 - lib/rubycanusellm/providers/anthropic.rb
 - lib/rubycanusellm/providers/base.rb
 - lib/rubycanusellm/providers/openai.rb
+- lib/rubycanusellm/providers/voyage.rb
 - lib/rubycanusellm/response.rb
 - lib/rubycanusellm/templates/completion.rb.tt
 - lib/rubycanusellm/templates/config.rb.tt