llm.rb 4.9.0 → 4.11.0

data/data/zai.json

@@ -8,19 +8,17 @@
   "name": "Z.AI",
   "doc": "https://docs.z.ai/guides/overview/pricing",
   "models": {
-    "glm-5": {
-      "id": "glm-5",
-      "name": "GLM-5",
-      "family": "glm",
+    "glm-4.7-flash": {
+      "id": "glm-4.7-flash",
+      "name": "GLM-4.7-Flash",
+      "family": "glm-flash",
       "attachment": false,
       "reasoning": true,
       "tool_call": true,
-      "interleaved": {
-        "field": "reasoning_content"
-      },
       "temperature": true,
-      "release_date": "2026-02-11",
-      "last_updated": "2026-02-11",
+      "knowledge": "2025-04",
+      "release_date": "2026-01-19",
+      "last_updated": "2026-01-19",
       "modalities": {
         "input": [
           "text"
@@ -31,58 +29,66 @@
       },
       "open_weights": true,
       "cost": {
-        "input": 1,
-        "output": 3.2,
-        "cache_read": 0.2,
+        "input": 0,
+        "output": 0,
+        "cache_read": 0,
         "cache_write": 0
       },
       "limit": {
-        "context": 204800,
+        "context": 200000,
         "output": 131072
       }
     },
-    "glm-4.5-air": {
-      "id": "glm-4.5-air",
-      "name": "GLM-4.5-Air",
-      "family": "glm-air",
-      "attachment": false,
+    "glm-5v-turbo": {
+      "id": "glm-5v-turbo",
+      "name": "glm-5v-turbo",
+      "family": "glm",
+      "attachment": true,
       "reasoning": true,
       "tool_call": true,
+      "interleaved": {
+        "field": "reasoning_content"
+      },
       "temperature": true,
-      "knowledge": "2025-04",
-      "release_date": "2025-07-28",
-      "last_updated": "2025-07-28",
+      "release_date": "2026-04-01",
+      "last_updated": "2026-04-01",
       "modalities": {
         "input": [
-          "text"
+          "text",
+          "image",
+          "video",
+          "pdf"
         ],
         "output": [
           "text"
         ]
       },
-      "open_weights": true,
+      "open_weights": false,
       "cost": {
-        "input": 0.2,
-        "output": 1.1,
-        "cache_read": 0.03,
+        "input": 1.2,
+        "output": 4,
+        "cache_read": 0.24,
         "cache_write": 0
       },
       "limit": {
-        "context": 131072,
-        "output": 98304
+        "context": 200000,
+        "output": 131072
       }
     },
-    "glm-4.5": {
-      "id": "glm-4.5",
-      "name": "GLM-4.5",
+    "glm-5-turbo": {
+      "id": "glm-5-turbo",
+      "name": "GLM-5-Turbo",
       "family": "glm",
       "attachment": false,
       "reasoning": true,
       "tool_call": true,
+      "interleaved": {
+        "field": "reasoning_content"
+      },
+      "structured_output": true,
       "temperature": true,
-      "knowledge": "2025-04",
-      "release_date": "2025-07-28",
-      "last_updated": "2025-07-28",
+      "release_date": "2026-03-16",
+      "last_updated": "2026-03-16",
       "modalities": {
         "input": [
           "text"
@@ -91,22 +97,22 @@
           "text"
         ]
       },
-      "open_weights": true,
+      "open_weights": false,
       "cost": {
-        "input": 0.6,
-        "output": 2.2,
-        "cache_read": 0.11,
+        "input": 1.2,
+        "output": 4,
+        "cache_read": 0.24,
         "cache_write": 0
       },
       "limit": {
-        "context": 131072,
-        "output": 98304
+        "context": 200000,
+        "output": 131072
       }
     },
-    "glm-4.5-flash": {
-      "id": "glm-4.5-flash",
-      "name": "GLM-4.5-Flash",
-      "family": "glm-flash",
+    "glm-4.5": {
+      "id": "glm-4.5",
+      "name": "GLM-4.5",
+      "family": "glm",
       "attachment": false,
       "reasoning": true,
       "tool_call": true,
@@ -124,9 +130,9 @@
       },
       "open_weights": true,
       "cost": {
-        "input": 0,
-        "output": 0,
-        "cache_read": 0,
+        "input": 0.6,
+        "output": 2.2,
+        "cache_read": 0.11,
         "cache_write": 0
       },
       "limit": {
@@ -134,9 +140,9 @@
         "output": 98304
       }
     },
-    "glm-4.7-flash": {
-      "id": "glm-4.7-flash",
-      "name": "GLM-4.7-Flash",
+    "glm-4.7-flashx": {
+      "id": "glm-4.7-flashx",
+      "name": "GLM-4.7-FlashX",
       "family": "glm-flash",
       "attachment": false,
       "reasoning": true,
@@ -155,9 +161,9 @@
       },
       "open_weights": true,
       "cost": {
-        "input": 0,
-        "output": 0,
-        "cache_read": 0,
+        "input": 0.07,
+        "output": 0.4,
+        "cache_read": 0.01,
         "cache_write": 0
       },
       "limit": {
@@ -196,20 +202,48 @@
         "output": 131072
       }
     },
-    "glm-4.7": {
-      "id": "glm-4.7",
-      "name": "GLM-4.7",
+    "glm-4.6v": {
+      "id": "glm-4.6v",
+      "name": "GLM-4.6V",
       "family": "glm",
-      "attachment": false,
+      "attachment": true,
       "reasoning": true,
       "tool_call": true,
-      "interleaved": {
-        "field": "reasoning_content"
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2025-12-08",
+      "last_updated": "2025-12-08",
+      "modalities": {
+        "input": [
+          "text",
+          "image",
+          "video"
+        ],
+        "output": [
+          "text"
+        ]
       },
+      "open_weights": true,
+      "cost": {
+        "input": 0.3,
+        "output": 0.9
+      },
+      "limit": {
+        "context": 128000,
+        "output": 32768
+      }
+    },
+    "glm-4.5-flash": {
+      "id": "glm-4.5-flash",
+      "name": "GLM-4.5-Flash",
+      "family": "glm-flash",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
       "temperature": true,
       "knowledge": "2025-04",
-      "release_date": "2025-12-22",
-      "last_updated": "2025-12-22",
+      "release_date": "2025-07-28",
+      "last_updated": "2025-07-28",
       "modalities": {
         "input": [
           "text"
@@ -220,19 +254,19 @@
       },
       "open_weights": true,
       "cost": {
-        "input": 0.6,
-        "output": 2.2,
-        "cache_read": 0.11,
+        "input": 0,
+        "output": 0,
+        "cache_read": 0,
         "cache_write": 0
       },
       "limit": {
-        "context": 204800,
-        "output": 131072
+        "context": 131072,
+        "output": 98304
       }
     },
-    "glm-5-turbo": {
-      "id": "glm-5-turbo",
-      "name": "GLM-5-Turbo",
+    "glm-5": {
+      "id": "glm-5",
+      "name": "GLM-5",
       "family": "glm",
       "attachment": false,
       "reasoning": true,
@@ -240,10 +274,9 @@
       "interleaved": {
         "field": "reasoning_content"
       },
-      "structured_output": true,
       "temperature": true,
-      "release_date": "2026-03-16",
-      "last_updated": "2026-03-16",
+      "release_date": "2026-02-11",
+      "last_updated": "2026-02-11",
       "modalities": {
         "input": [
           "text"
@@ -252,18 +285,49 @@
           "text"
         ]
       },
-      "open_weights": false,
+      "open_weights": true,
       "cost": {
-        "input": 1.2,
-        "output": 4,
-        "cache_read": 0.24,
+        "input": 1,
+        "output": 3.2,
+        "cache_read": 0.2,
         "cache_write": 0
       },
       "limit": {
-        "context": 200000,
+        "context": 204800,
         "output": 131072
       }
     },
+    "glm-4.5-air": {
+      "id": "glm-4.5-air",
+      "name": "GLM-4.5-Air",
+      "family": "glm-air",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2025-07-28",
+      "last_updated": "2025-07-28",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0.2,
+        "output": 1.1,
+        "cache_read": 0.03,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 131072,
+        "output": 98304
+      }
+    },
     "glm-4.5v": {
       "id": "glm-4.5v",
       "name": "GLM-4.5V",
@@ -295,22 +359,23 @@
         "output": 16384
       }
     },
-    "glm-4.6v": {
-      "id": "glm-4.6v",
-      "name": "GLM-4.6V",
+    "glm-4.7": {
+      "id": "glm-4.7",
+      "name": "GLM-4.7",
       "family": "glm",
-      "attachment": true,
+      "attachment": false,
       "reasoning": true,
       "tool_call": true,
+      "interleaved": {
+        "field": "reasoning_content"
+      },
       "temperature": true,
       "knowledge": "2025-04",
-      "release_date": "2025-12-08",
-      "last_updated": "2025-12-08",
+      "release_date": "2025-12-22",
+      "last_updated": "2025-12-22",
       "modalities": {
         "input": [
-          "text",
-          "image",
-          "video"
+          "text"
         ],
         "output": [
           "text"
@@ -318,12 +383,14 @@
       },
       "open_weights": true,
       "cost": {
-        "input": 0.3,
-        "output": 0.9
+        "input": 0.6,
+        "output": 2.2,
+        "cache_read": 0.11,
+        "cache_write": 0
       },
       "limit": {
-        "context": 128000,
-        "output": 32768
+        "context": 204800,
+        "output": 131072
       }
     }
   }

data/lib/llm/context/deserializer.rb

@@ -12,7 +12,8 @@ class LLM::Context
       returns = deserialize_returns(payload["content"]) if returns.nil?
       original_tool_calls = payload["original_tool_calls"]
       usage = payload["usage"]
-      extra = {tool_calls:, original_tool_calls:, tools: @params[:tools], usage:}.compact
+      reasoning_content = payload["reasoning_content"]
+      extra = {tool_calls:, original_tool_calls:, tools: @params[:tools], usage:, reasoning_content:}.compact
       content = returns.nil? ? payload["content"] : returns
       LLM::Message.new(payload["role"], content, extra)
     end

data/lib/llm/context.rb

@@ -42,6 +42,11 @@ module LLM
     # @return [LLM::Provider]
     attr_reader :llm
 
+    ##
+    # Returns the context mode
+    # @return [Symbol]
+    attr_reader :mode
+
     ##
     # @param [LLM::Provider] llm
     #  A provider
@@ -49,10 +54,12 @@ module LLM
     #  The parameters to maintain throughout the conversation.
     #  Any parameter the provider supports can be included and
     #  not only those listed here.
+    # @option params [Symbol] :mode Defaults to :completions
     # @option params [String] :model Defaults to the provider's default model
     # @option params [Array<LLM::Function>, nil] :tools Defaults to nil
     def initialize(llm, params = {})
       @llm = llm
+      @mode = params.delete(:mode) || :completions
       @params = {model: llm.default_model, schema: nil}.compact.merge!(params)
       @messages = LLM::Buffer.new(llm)
     end
@@ -70,6 +77,7 @@ module LLM
     #   res = ctx.talk("Hello, what is your name?")
     #   puts res.messages[0].content
     def talk(prompt, params = {})
+      return respond(prompt, params) if mode == :responses
       params = params.merge(messages: @messages.to_a)
       params = @params.merge(params)
       res = @llm.complete(prompt, params)
@@ -109,7 +117,7 @@ module LLM
     # @return [String]
     def inspect
       "#<#{self.class.name}:0x#{object_id.to_s(16)} " \
-      "@llm=#{@llm.class}, @params=#{@params.inspect}, " \
+      "@llm=#{@llm.class}, @mode=#{@mode.inspect}, @params=#{@params.inspect}, " \
       "@messages=#{@messages.inspect}>"
     end
 
@@ -117,10 +125,11 @@ module LLM
     # Returns an array of functions that can be called
     # @return [Array<LLM::Function>]
     def functions
+      return_ids = returns.map(&:id)
       @messages
         .select(&:assistant?)
         .flat_map do |msg|
-          fns = msg.functions.select(&:pending?)
+          fns = msg.functions.select { _1.pending? && !return_ids.include?(_1.id) }
           fns.each do |fn|
             fn.tracer = tracer
             fn.model  = msg.model
@@ -128,6 +137,53 @@ module LLM
         end.extend(LLM::Function::Array)
     end
 
+    ##
+    # Calls a named collection of work through the context.
+    #
+    # This currently supports `:functions`, forwarding to `functions.call`.
+    #
+    # @param [Symbol] target
+    #  The work collection to call
+    # @return [Array<LLM::Function::Return>]
+    def call(target)
+      case target
+      when :functions then functions.call
+      else raise ArgumentError, "Unknown target: #{target.inspect}. Expected :functions"
+      end
+    end
+
+    ##
+    # Returns tool returns accumulated in this context
+    # @return [Array<LLM::Function::Return>]
+    def returns
+      @messages
+        .select(&:tool_return?)
+        .flat_map do |msg|
+          LLM::Function::Return === msg.content ?
+            [msg.content] :
+            [*msg.content].grep(LLM::Function::Return)
+        end
+    end
+
+    ##
+    # Waits for queued tool work to finish.
+    #
+    # This prefers queued streamed tool work when the configured stream
+    # exposes a non-empty queue. Otherwise it falls back to waiting on
+    # the context's pending functions directly.
+    #
+    # @param [Symbol] strategy
+    #  The concurrency strategy to use
+    # @return [Array<LLM::Function::Return>]
+    def wait(strategy)
+      stream = @params[:stream]
+      if LLM::Stream === stream && !stream.queue.empty?
+        stream.wait(strategy)
+      else
+        functions.wait(strategy)
+      end
+    end
+
     ##
     # Returns token usage accumulated in this context
     # @note

data/lib/llm/contract/completion.rb

@@ -50,6 +50,13 @@ module LLM::Contract
       messages.find(&:assistant?).content
     end
 
+    ##
+    # @return [String, nil]
+    #  Returns the reasoning content when the provider exposes it
+    def reasoning_content
+      messages.find(&:assistant?)&.reasoning_content
+    end
+
     ##
     # @return [Hash]
     #  Returns the LLM response after parsing it as JSON

data/lib/llm/error.rb

@@ -55,6 +55,10 @@ module LLM
   # When stuck in a tool call loop
   ToolLoopError = Class.new(Error)
 
+  ##
+  # When a tool call cannot be mapped to a local tool
+  NoSuchToolError = Class.new(Error)
+
   ##
   # When {LLM::Registry} can't map a model
   NoSuchModelError = Class.new(Error)

data/lib/llm/eventhandler.rb

@@ -42,5 +42,12 @@ module LLM
     # Returns a fully constructed response body
     # @return [LLM::Object]
     def body = @parser.body
+
+    ##
+    # Frees parser state after streaming completes.
+    # @return [void]
+    def free
+      @parser.free
+    end
   end
 end

data/lib/llm/function/registry.rb

@@ -0,0 +1,106 @@
+# frozen_string_literal: true
+
+class LLM::Function
+  ##
+  # The {LLM::Function::Registry LLM::Function::Registry} module provides
+  # shared registry behavior for functions and tools. {LLM::Tool.registry}
+  # stores {LLM::Tool LLM::Tool} subclasses, including dynamically created MCP
+  # tool subclasses, while {LLM::Function.registry} stores the functions
+  # derived from those tools.
+  #
+  # The registry overwrites older tool definitions with newer ones when they
+  # share the same tool name. In practice, tool identity is resolved by name,
+  # and LLMs generally do not allow two tools with the same name.
+  #
+  # Functions defined with {LLM.function} are not added to the function
+  # registry, since they may be closures bound to local state. Each registry
+  # decides how entries are keyed via {#registry_key}.
+  module Registry
+    ##
+    # @api private
+    def self.extended(klass)
+      klass.instance_variable_set(:@__registry, {})
+      klass.instance_variable_set(:@__names, {})
+      klass.instance_variable_set(:@__monitor, Monitor.new)
+    end
+
+    ##
+    # Returns all registered entries.
+    # @return [Array<LLM::Function, LLM::Tool>]
+    def registry
+      lock do
+        @__registry.values
+      end
+    end
+
+    ##
+    # Finds a registered entry by name.
+    # @param [String] name
+    # @return [LLM::Function, LLM::Tool, nil]
+    def find_by_name(name)
+      lock do
+        @__names[name.to_s] ||= @__registry.each_value.find do
+          tool_name(_1).to_s == name.to_s
+        end
+      end
+    end
+
+    ##
+    # Clears the registry.
+    # @return [void]
+    def clear_registry!
+      lock do
+        @__registry.clear
+        @__names.clear
+        nil
+      end
+    end
+
+    ##
+    # Registers an entry.
+    # @param [LLM::Function, LLM::Tool] entry
+    # @api private
+    def register(entry)
+      lock do
+        @__registry[registry_key(entry)] = entry
+        @__names[tool_name(entry).to_s] = entry if tool_name(entry)
+      end
+    end
+
+    ##
+    # Unregisters an entry.
+    # @param [LLM::Function, LLM::Tool] entry
+    # @api private
+    def unregister(entry)
+      lock do
+        @__registry.delete(registry_key(entry))
+        @__registry.delete(entry)
+        @__names.delete(tool_name(entry).to_s) if tool_name(entry)
+      end
+    end
+
+    ##
+    # Returns the storage key for an entry.
+    # @param [LLM::Function, LLM::Tool] entry
+    # @return [Class<LLM::Tool>, String, nil]
+    # @api private
+    def registry_key(entry)
+      tool_name(entry) ? entry.name : entry
+    end
+
+    ##
+    # Returns the tool name, or nil for tools that are not fully initialized.
+    # @param [LLM::Function, LLM::Tool] entry
+    # @return [String, nil]
+    # @api private
+    def tool_name(entry)
+      entry.respond_to?(:name) ? entry.name : nil
+    end
+
+    ##
+    # @api private
+    def lock(&)
+      @__monitor.synchronize(&)
+    end
+  end
+end