llm.rb 0.11.0 → 0.13.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ce6a4e56ce25f397337733009249edb930cade1641bd84d7939ef0d349c6e92a
-  data.tar.gz: b9506c346dd19af655f342e9b6a64aed25cd3f4ecffb6bb32cb36e1221e71c6a
+  metadata.gz: f83248fa3bce285d75952b726cc7d1b097e61b61fb77edd55f39168922a8614c
+  data.tar.gz: 52fa691fad7eaa2e77e24245aba97ea469688f8307f869d533fac47707aa6caf
 SHA512:
-  metadata.gz: a4a53b8e5aeaae2cd26f5303366ce1974ecfdb4176a76c5aa01f46cb9997d10a402de782b0b14f97bf3c65dc4c200b976c1279a037ce29ae44ab64fd1799ffd9
-  data.tar.gz: 96128a756147d8cad13e85f91247fdb43b6bf5c469348318c27f27518670c61b6466546d65bf1f553b086b2a42b15a5225e0ac5355b680ba46214b32bc20d754
+  metadata.gz: d17e104601295bbbd8dd5bd34fe5378e96367aacb832ddb81b7f0fdfd3fa54f39d98dff6568f3190a4d979de4382a6ebd2eba8ee2fb2cce0e67db2478be9a1df
+  data.tar.gz: d2e6bb0f507bdcf140bd7f44f9edd40c057f061a9b4bc7123a87bdf6fa40a6f16a99642965bdaf81aaaeab5e995a01587c79cd47e19c6a2b5f86e456fa4813a9

data/README.md

@@ -1,9 +1,9 @@
 ## About
 
 llm.rb is a zero-dependency Ruby toolkit for Large Language Models that
-includes OpenAI, Gemini, Anthropic, DeepSeek, Ollama, and LlamaCpp. The
-toolkit includes full support for chat, streaming, tool calling, audio,
-images, files, and JSON Schema generation.
+includes OpenAI, Gemini, Anthropic, xAI (grok), DeepSeek, Ollama, and
+LlamaCpp. The toolkit includes full support for chat, streaming, tool calling,
+audio, images, files, and structured outputs (JSON Schema).
 
 ## Features
 
@@ -22,12 +22,41 @@ images, files, and JSON Schema generation.
 - 🗣️ Text-to-speech, transcription, and translation
 - 🖼️ Image generation, editing, and variation support
 - 📎 File uploads and prompt-aware file interaction
-- 💡 Multimodal prompts (text, images, PDFs, URLs, files)
+- 💡 Multimodal prompts (text, documents, audio, images, videos, URLs, etc)
 
-#### Miscellaneous
+#### Embeddings
 - 🧮 Text embeddings and vector support
-- 🔌 Retrieve models dynamically for introspection and selection
-- 🧱 Includes support for OpenAI's responses, moderations, and vector stores APIs
+- 🧱 Includes support for OpenAI's vector stores API
+
+#### Miscellaneous
+- 📜 Model management and selection
+- 🔧 Includes support for OpenAI's responses, moderations, and vector stores APIs
+
+## Matrix
+
+While the Features section above gives you the high-level picture, the table below
+breaks things down by provider, so you can see exactly what’s supported where.
+
+
+| Feature / Provider                  | OpenAI | Anthropic | Gemini | DeepSeek | xAI (Grok) | Ollama | LlamaCpp |
+|--------------------------------------|:------:|:---------:|:------:|:--------:|:----------:|:------:|:--------:|
+| **Chat Completions**                 | ✅     | ✅        | ✅     | ✅       | ✅         | ✅     | ✅       |
+| **Streaming**                        | ✅     | ✅        | ✅     | ✅       | ✅         | ✅     | ✅       |
+| **Tool Calling**                     | ✅     | ✅        | ✅     | ✅       | ✅         | ✅     | ✅       |
+| **JSON Schema / Structured Output**  | ✅     | ❌        | ✅     | ❌       | ✅         | ✅*    | ✅*      |
+| **Audio (TTS / Transcribe / Translate)** | ✅  | ❌        | ✅     | ❌       | ❌         | ❌     | ❌       |
+| **Image Generation & Editing**       | ✅     | ❌        | ✅     | ❌       | ✅         | ❌     | ❌       |
+| **File Uploads**                     | ✅     | ❌        | ✅     | ❌       | ❌         | ❌     | ❌       |
+| **Multimodal Prompts** *(text, documents, audio, images, videos, URLs, etc)* | ✅     | ✅        | ✅     | ✅       | ✅         | ✅     | ✅       |
+| **Embeddings**                       | ✅     | ✅        | ✅     | ✅       | ❌         | ✅     | ✅       |
+| **Models API**                       | ✅     | ✅        | ✅     | ✅       | ✅         | ✅     | ✅       |
+| **Local Model Support**              | ❌     | ❌        | ❌     | ❌       | ❌         | ✅     | ✅       |
+| **Vector Stores (RAG)**               | ✅     | ❌        | ❌     | ❌       | ❌         | ❌     | ❌       |
+| **Responses**                        | ✅     | ❌        | ❌     | ❌       | ❌         | ❌     | ❌       |
+| **Moderations**                      | ✅     | ❌        | ❌     | ❌       | ❌         | ❌     | ❌       |
+
+\* JSON Schema support in Ollama/LlamaCpp depends on the model, not the API.
+
 
 ## Examples
 
@@ -49,6 +78,7 @@ require "llm"
 llm = LLM.openai(key: "yourapikey")
 llm = LLM.gemini(key: "yourapikey")
 llm = LLM.anthropic(key: "yourapikey")
+llm = LLM.xai(key: "yourapikey")
 llm = LLM.deepseek(key: "yourapikey")
 
 ##
@@ -78,14 +108,14 @@ is made before sending a request to the LLM:
 #!/usr/bin/env ruby
 require "llm"
 
-llm  = LLM.openai(key: ENV["OPENAI_SECRET"])
+llm  = LLM.openai(key: ENV["KEY"])
 bot  = LLM::Bot.new(llm)
-url  = "https://upload.wikimedia.org/wikipedia/commons/thumb/9/9a/Cognac_glass.jpg/500px-Cognac_glass.jpg"
+url  = "https://en.wikipedia.org/wiki/Special:FilePath/Cognac_glass.jpg"
 msgs = bot.chat do |prompt|
   prompt.system "Your task is to answer all user queries"
   prompt.user ["Tell me about this URL", URI(url)]
-  prompt.user ["Tell me about this pdf", File.open("spec/fixtures/documents/freebsd.sysctl.pdf", "r")]
-  prompt.user "Is the URL and PDF similar to each other?"
+  prompt.user ["Tell me about this PDF", File.open("handbook.pdf", "rb")]
+  prompt.user "Are the URL and PDF similar to each other?"
 end
 
 # At this point, we execute a single request
@@ -110,14 +140,14 @@ to process a response in the same way:
 #!/usr/bin/env ruby
 require "llm"
 
-llm = LLM.openai(key: ENV["OPENAI_SECRET"])
+llm = LLM.openai(key: ENV["KEY"])
 bot = LLM::Bot.new(llm)
-url = "https://upload.wikimedia.org/wikipedia/commons/thumb/9/9a/Cognac_glass.jpg/500px-Cognac_glass.jpg"
+url = "https://en.wikipedia.org/wiki/Special:FilePath/Cognac_glass.jpg"
 bot.chat(stream: $stdout) do |prompt|
   prompt.system "Your task is to answer all user queries"
   prompt.user ["Tell me about this URL", URI(url)]
-  prompt.user ["Tell me about this pdf", File.open("spec/fixtures/documents/freebsd.sysctl.pdf", "r")]
-  prompt.user "Is the URL and PDF similar to each other?"
+  prompt.user ["Tell me about this PDF", File.open("handbook.pdf", "rb")]
+  prompt.user "Are the URL and PDF similar to each other?"
 end.to_a
 ```
 
@@ -137,7 +167,7 @@ require "llm"
 ##
 # Objects
 llm = LLM.openai(key: ENV["KEY"])
-schema = llm.schema.object(answer: llm.schema.integer.required)
+schema = llm.schema.object(probability: llm.schema.integer.required)
 bot = LLM::Bot.new(llm, schema:)
 bot.chat "Does the earth orbit the sun?", role: :user
 bot.messages.find(&:assistant?).content! # => {probability: 1}
@@ -245,7 +275,7 @@ to represent a file stored with the LLM, and so on. These are objects you
 can throw at the prompt and have them be understood automatically.
 
 A prompt can also have multiple parts, and in that case, an array is given
-as a prompt. Each element is considered to part of the prompt:
+as a prompt. Each element is considered to be part of the prompt:
 
 ```ruby
 #!/usr/bin/env ruby
@@ -463,8 +493,15 @@ over or doesn't cover at all. The API reference is available at
 
 ### Guides
 
-The [docs/](docs/) directory contains some additional documentation that
-didn't quite make it into the README.
+* [An introduction to RAG with llm.rb](https://0x1eef.github.io/posts/an-introduction-to-rag-with-llm.rb/) –
+  a blog post that implements the RAG pattern in 32 lines of Ruby code
+* [docs/](docs/) – the docs directory contains additional guides
+
+
+## See also
+
+* [llm-shell](https://github.com/llmrb/llm-shell) – a shell that uses llm.rb to
+  provide a command-line interface to LLMs.
 
 ## Install
 

data/lib/llm/bot/conversable.rb

@@ -12,8 +12,12 @@ class LLM::Bot
     # @param [Hash] params
     # @return [void]
     def async_response(prompt, params = {})
-      role = params.delete(:role)
-      @messages << [LLM::Message.new(role, prompt), @params.merge(params), :respond]
+      if Array === prompt and prompt.empty?
+        @messages
+      else
+        role = params.delete(:role)
+        @messages << [LLM::Message.new(role, prompt), @params.merge(params), :respond]
+      end
     end
 
     ##
@@ -22,8 +26,12 @@ class LLM::Bot
     # @param [Hash] params
     # @return [void]
     def async_completion(prompt, params = {})
-      role = params.delete(:role)
-      @messages.push [LLM::Message.new(role, prompt), @params.merge(params), :complete]
+      if Array === prompt and prompt.empty?
+        @messages
+      else
+        role = params.delete(:role)
+        @messages << [LLM::Message.new(role, prompt), @params.merge(params), :complete]
+      end
     end
   end
 end

data/lib/llm/bot/prompt/completion.rb

@@ -27,5 +27,23 @@ module LLM::Bot::Prompt
       params = defaults.merge(params)
       bot.chat prompt, params.merge(role: :user)
     end
+
+    ##
+    # @param [String] prompt
+    # @param [Hash] params (see LLM::Provider#complete)
+    # @return [LLM::Bot]
+    def assistant(prompt, params = {})
+      params = defaults.merge(params)
+      bot.chat prompt, params.merge(role: :assistant)
+    end
+
+    ##
+    # @param [String] prompt
+    # @param [Hash] params (see LLM::Provider#complete)
+    # @return [LLM::Bot]
+    def model(prompt, params = {})
+      params = defaults.merge(params)
+      bot.chat prompt, params.merge(role: :model)
+    end
   end
 end

data/lib/llm/bot/prompt/respond.rb

@@ -36,5 +36,14 @@ module LLM::Bot::Prompt
       params = defaults.merge(params)
       bot.respond prompt, params.merge(role: :user)
     end
+
+    ##
+    # @param [String] prompt
+    # @param [Hash] params (see LLM::Provider#complete)
+    # @return [LLM::Bot]
+    def assistant(prompt, params = {})
+      params = defaults.merge(params)
+      bot.chat prompt, params.merge(role: :assistant)
+    end
   end
 end

data/lib/llm/bot.rb

@@ -3,33 +3,26 @@
 module LLM
   ##
   # {LLM::Bot LLM::Bot} provides an object that can maintain a
-  # a conversation. A conversation can use the chat completions API
+  # conversation. A conversation can use the chat completions API
   # that all LLM providers support or the responses API that currently
   # only OpenAI supports.
   #
-  # @example example #1
+  # @example
   #   #!/usr/bin/env ruby
   #   require "llm"
   #
-  #   llm  = LLM.openai(ENV["KEY"])
+  #   llm  = LLM.openai(key: ENV["KEY"])
   #   bot  = LLM::Bot.new(llm)
+  #   url  = "https://en.wikipedia.org/wiki/Special:FilePath/Cognac_glass.jpg"
   #   msgs = bot.chat do |prompt|
-  #     prompt.user "What programming language should I learn next ?"
-  #     prompt.user "Can you recommend a good book ?"
-  #     prompt.user "Can you suggest a fun project to practice ?"
+  #     prompt.system "Your task is to answer all user queries"
+  #     prompt.user ["Tell me about this URL", URI(url)]
+  #     prompt.user ["Tell me about this PDF", File.open("handbook.pdf", "rb")]
+  #     prompt.user "Are the URL and PDF similar to each other?"
   #   end
-  #   msgs.each { print "[#{_1.role}]", _1.content, "\n" }
   #
-  # @example example #2
-  #   #!/usr/bin/env ruby
-  #   require "llm"
-  #
-  #   llm = LLM.openai(ENV["KEY"])
-  #   bot = LLM::Bot.new(llm)
-  #   bot.chat "What programming language should I learn next ?", role: :user
-  #   bot.chat "Can you recommend a good book ?", role: :user
-  #   bot.chat "Can you suggest a fun project to practice ?", role: :user
-  #   bot.messages.each { print "[#{_1.role}]", _1.content, "\n" }
+  #   # At this point, we execute a single request
+  #   msgs.each { print "[#{_1.role}] ", _1.content, "\n" }
   class Bot
     require_relative "bot/prompt/completion"
     require_relative "bot/prompt/respond"

data/lib/llm/buffer.rb

@@ -48,7 +48,14 @@ module LLM
     end
 
     ##
-    # @param [[LLM::Message, Hash]] item
+    # Returns the last message in the buffer
+    # @return [LLM::Message, nil]
+    def last
+      to_a[-1]
+    end
+
+    ##
+    # @param [[LLM::Message, Hash, Symbol]] item
     #  A message and its parameters
     # @return [void]
     def <<(item)
@@ -73,6 +80,13 @@ module LLM
       "completed_count=#{@completed.size} pending_count=#{@pending.size}>"
     end
 
+    ##
+    # Returns true when the buffer is empty
+    # @return [Boolean]
+    def empty?
+      @pending.empty? and @completed.empty?
+    end
+
     private
 
     def empty!

data/lib/llm/error.rb

@@ -31,6 +31,10 @@ module LLM
   # HTTPTooManyRequests
   RateLimitError = Class.new(ResponseError)
 
+  ##
+  # HTTPServerError
+  ServerError = Class.new(ResponseError)
+
   ##
   # When an given an input object that is not understood
   FormatError = Class.new(Error)

data/lib/llm/file.rb

@@ -1,9 +1,10 @@
 # frozen_string_literal: true
 
 ##
-# The {LLM::File LLM::File} class represents a local file. It can
-# be used as a prompt with certain providers (eg: Ollama, Gemini),
-# and as an input with certain methods
+# {LLM::File LLM::File} represents a local file. It can be used
+# as a prompt with certain providers (eg: Ollama, Gemini),
+# and as an input with certain methods. It is usually not necessary
+# to create an instance of LLM::File directly.
 class LLM::File
   ##
   # @return [String]

data/lib/llm/function.rb

@@ -1,8 +1,8 @@
 # frozen_string_literal: true
 
 ##
-# The {LLM::Function LLM::Function} class represents a
-# local function that can be called by an LLM.
+# The {LLM::Function LLM::Function} class represents a local
+# function that can be called by an LLM.
 #
 # @example example #1
 #   LLM.function(:system) do |fn|
@@ -53,7 +53,7 @@ class LLM::Function
   # @yieldparam [LLM::Function] self The function object
   def initialize(name, &b)
     @name = name
-    @schema = JSON::Schema.new
+    @schema = LLM::Schema.new
     @called = false
     @cancelled = false
     yield(self)
@@ -72,7 +72,7 @@ class LLM::Function
   end
 
   ##
-  # @yieldparam [JSON::Schema] schema The schema object
+  # @yieldparam [LLM::Schema] schema The schema object
   # @return [void]
   def params
     @params = yield(@schema)

data/lib/llm/mime.rb

@@ -7,11 +7,8 @@ class LLM::Mime
   # Lookup a mime type
   # @return [String, nil]
   def self.[](key)
-    if key.respond_to?(:path)
-      types[File.extname(key.path)]
-    else
-      types[key]
-    end
+    key = key.respond_to?(:path) ? File.extname(key.path) : key
+    types[key] || "application/octet-stream"
   end
 
   ##
@@ -24,6 +21,15 @@ class LLM::Mime
       ".jpg" => "image/jpeg",
       ".jpeg" => "image/jpeg",
       ".webp" => "image/webp",
+      ".gif" => "image/gif",
+      ".bmp" => "image/bmp",
+      ".tif" => "image/tiff",
+      ".tiff" => "image/tiff",
+      ".svg" => "image/svg+xml",
+      ".ico" => "image/x-icon",
+      ".apng" => "image/apng",
+      ".jfif" => "image/jpeg",
+      ".heic" => "image/heic",
 
       # Videos
       ".flv" => "video/x-flv",
@@ -34,6 +40,12 @@ class LLM::Mime
       ".webm" => "video/webm",
       ".wmv" => "video/x-ms-wmv",
       ".3gp" => "video/3gpp",
+      ".avi" => "video/x-msvideo",
+      ".mkv" => "video/x-matroska",
+      ".ogv" => "video/ogg",
+      ".m4v" => "video/x-m4v",
+      ".m2ts" => "video/mp2t",
+      ".mts" => "video/mp2t",
 
       # Audio
       ".aac" => "audio/aac",
@@ -45,10 +57,80 @@ class LLM::Mime
       ".pcm" => "audio/L16",
       ".wav" => "audio/wav",
       ".weba" => "audio/webm",
+      ".oga" => "audio/ogg",
+      ".ogg" => "audio/ogg",
+      ".mid" => "audio/midi",
+      ".midi" => "audio/midi",
+      ".aiff" => "audio/aiff",
+      ".aif" => "audio/aiff",
+      ".amr" => "audio/amr",
+      ".mka" => "audio/x-matroska",
+      ".caf" => "audio/x-caf",
 
       # Documents
       ".pdf" => "application/pdf",
-      ".txt" => "text/plain"
+      ".txt" => "text/plain",
+      ".md" => "text/markdown",
+      ".markdown" => "text/markdown",
+      ".mkd" => "text/markdown",
+      ".doc" => "application/msword",
+      ".docx" => "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+      ".xls" => "application/vnd.ms-excel",
+      ".xlsx" => "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+      ".ppt" => "application/vnd.ms-powerpoint",
+      ".pptx" => "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+      ".csv" => "text/csv",
+      ".json" => "application/json",
+      ".xml" => "application/xml",
+      ".html" => "text/html",
+      ".htm" => "text/html",
+      ".odt" => "application/vnd.oasis.opendocument.text",
+      ".odp" => "application/vnd.oasis.opendocument.presentation",
+      ".ods" => "application/vnd.oasis.opendocument.spreadsheet",
+      ".rtf" => "application/rtf",
+      ".epub" => "application/epub+zip",
+
+      # Code
+      ".js" => "application/javascript",
+      ".jsx" => "text/jsx",
+      ".ts" => "application/typescript",
+      ".tsx" => "text/tsx",
+      ".css" => "text/css",
+      ".c" => "text/x-c",
+      ".cpp" => "text/x-c++",
+      ".h" => "text/x-c",
+      ".rb" => "text/x-ruby",
+      ".py" => "text/x-python",
+      ".java" => "text/x-java-source",
+      ".sh" => "application/x-sh",
+      ".php" => "application/x-httpd-php",
+      ".yml" => "text/yaml",
+      ".yaml" => "text/yaml",
+      ".go" => "text/x-go",
+      ".rs" => "text/rust",
+
+      # Fonts
+      ".woff" => "font/woff",
+      ".woff2" => "font/woff2",
+      ".ttf" => "font/ttf",
+      ".otf" => "font/otf",
+
+      # Archives
+      ".zip" => "application/zip",
+      ".tar" => "application/x-tar",
+      ".gz" => "application/gzip",
+      ".bz2" => "application/x-bzip2",
+      ".xz" => "application/x-xz",
+      ".rar" => "application/vnd.rar",
+      ".7z" => "application/x-7z-compressed",
+      ".tar.gz" => "application/gzip",
+      ".tar.bz2" => "application/x-bzip2",
+      ".apk" => "application/vnd.android.package-archive",
+      ".exe" => "application/x-msdownload",
+
+      # Others
+      ".ics" => "text/calendar",
+      ".vcf" => "text/vcard"
     }
   end
 end

data/lib/llm/multipart.rb

@@ -4,7 +4,6 @@
 ##
 # @private
 class LLM::Multipart
-  require "llm"
   require "securerandom"
 
   ##

data/lib/llm/object.rb

@@ -1,10 +1,9 @@
 # frozen_string_literal: true
 
 ##
-# The {LLM::Object LLM::Object} class encapsulates a Hash object, and it
-# allows a consumer to get and set Hash keys via regular methods. It is
+# The {LLM::Object LLM::Object} class encapsulates a Hash object. It is
 # similar in spirit to OpenStruct, and it was introduced after OpenStruct
-# became a bundled gem (and not a default gem) in Ruby 3.5.
+# became a bundled gem rather than a default gem in Ruby 3.5.
 class LLM::Object < BasicObject
   require_relative "object/builder"
   require_relative "object/kernel"

data/lib/llm/provider.rb

@@ -52,7 +52,7 @@ class LLM::Provider
   ##
   # Provides an interface to the chat completions API
   # @example
-  #   llm = LLM.openai(ENV["KEY"])
+  #   llm = LLM.openai(key: ENV["KEY"])
   #   messages = [{role: "system", content: "Your task is to answer all of my questions"}]
   #   res = llm.complete("5 + 2 ?", messages:)
   #   print "[#{res.choices[0].role}]", res.choices[0].content, "\n"
@@ -198,9 +198,9 @@ class LLM::Provider
 
   ##
   # Returns an object that can generate a JSON schema
-  # @return [JSON::Schema]
+  # @return [LLM::Schema]
   def schema
-    @schema ||= JSON::Schema.new
+    @schema ||= LLM::Schema.new
   end
 
   ##

data/lib/llm/providers/anthropic/error_handler.rb

@@ -22,6 +22,8 @@ class LLM::Anthropic
     #  Raises a subclass of {LLM::Error LLM::Error}
     def raise_error!
       case res
+      when Net::HTTPServerError
+        raise LLM::ServerError.new { _1.response = res }, "Server error"
       when Net::HTTPUnauthorized
         raise LLM::UnauthorizedError.new { _1.response = res }, "Authentication error"
       when Net::HTTPTooManyRequests

data/lib/llm/providers/anthropic/models.rb

@@ -11,7 +11,7 @@ class LLM::Anthropic
   #   #!/usr/bin/env ruby
   #   require "llm"
   #
-  #   llm = LLM.anthropic(ENV["KEY"])
+  #   llm = LLM.anthropic(key: ENV["KEY"])
   #   res = llm.models.all
   #   res.each do |model|
   #     print "id: ", model.id, "\n"
@@ -28,7 +28,7 @@ class LLM::Anthropic
     ##
     # List all models
     # @example
-    #   llm = LLM.anthropic(ENV["KEY"])
+    #   llm = LLM.anthropic(key: ENV["KEY"])
     #   res = llm.models.all
     #   res.each do |model|
     #     print "id: ", model.id, "\n"

data/lib/llm/providers/anthropic.rb

@@ -3,7 +3,16 @@
 module LLM
   ##
   # The Anthropic class implements a provider for
-  # [Anthropic](https://www.anthropic.com)
+  # [Anthropic](https://www.anthropic.com).
+  #
+  # @example
+  #   #!/usr/bin/env ruby
+  #   require "llm"
+  #
+  #   llm = LLM.anthropic(key: ENV["KEY"])
+  #   bot = LLM::Bot.new(llm)
+  #   bot.chat ["Tell me about this photo", File.open("/images/dog.jpg", "rb")]
+  #   bot.messages.select(&:assistant?).each { print "[#{_1.role}]", _1.content, "\n" }
   class Anthropic < Provider
     require_relative "anthropic/response/completion"
     require_relative "anthropic/format"

data/lib/llm/providers/deepseek.rb

@@ -6,8 +6,17 @@ module LLM
   ##
   # The DeepSeek class implements a provider for
   # [DeepSeek](https://deepseek.com)
-  # through its OpenAI-compatible API provided via
+  # through its OpenAI-compatible API available via
   # their [web platform](https://platform.deepseek.com).
+  #
+  # @example
+  #   #!/usr/bin/env ruby
+  #   require "llm"
+  #
+  #   llm = LLM.deepseek(key: ENV["KEY"])
+  #   bot = LLM::Bot.new(llm)
+  #   bot.chat ["Tell me about this photo", File.open("/images/cat.jpg", "rb")]
+  #   bot.messages.select(&:assistant?).each { print "[#{_1.role}]", _1.content, "\n" }
   class DeepSeek < OpenAI
     require_relative "deepseek/format"
     include DeepSeek::Format

data/lib/llm/providers/gemini/audio.rb

@@ -8,7 +8,7 @@ class LLM::Gemini
   #   #!/usr/bin/env ruby
   #   require "llm"
   #
-  #   llm = LLM.gemini(ENV["KEY"])
+  #   llm = LLM.gemini(key: ENV["KEY"])
   #   res = llm.audio.create_transcription(input: "/audio/rocket.mp3")
   #   res.text # => "A dog on a rocket to the moon"
   class Audio
@@ -30,7 +30,7 @@ class LLM::Gemini
     ##
     # Create an audio transcription
     # @example
-    #   llm = LLM.gemini(ENV["KEY"])
+    #   llm = LLM.gemini(key: ENV["KEY"])
     #   res = llm.audio.create_transcription(file: "/audio/rocket.mp3")
     #   res.text # => "A dog on a rocket to the moon"
     # @see https://ai.google.dev/gemini-api/docs/audio Gemini docs
@@ -52,7 +52,7 @@ class LLM::Gemini
     # Create an audio translation (in English)
     # @example
     #   # Arabic => English
-    #   llm = LLM.gemini(ENV["KEY"])
+    #   llm = LLM.gemini(key: ENV["KEY"])
     #   res = llm.audio.create_translation(file: "/audio/bismillah.mp3")
     #   res.text # => "In the name of Allah, the Beneficent, the Merciful."
     # @see https://ai.google.dev/gemini-api/docs/audio Gemini docs