llm.rb 4.8.0 → 4.10.0

data/data/zai.json

@@ -0,0 +1,330 @@
+{
+  "id": "zai",
+  "env": [
+    "ZHIPU_API_KEY"
+  ],
+  "npm": "@ai-sdk/openai-compatible",
+  "api": "https://api.z.ai/api/paas/v4",
+  "name": "Z.AI",
+  "doc": "https://docs.z.ai/guides/overview/pricing",
+  "models": {
+    "glm-5": {
+      "id": "glm-5",
+      "name": "GLM-5",
+      "family": "glm",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
+      "interleaved": {
+        "field": "reasoning_content"
+      },
+      "temperature": true,
+      "release_date": "2026-02-11",
+      "last_updated": "2026-02-11",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 1,
+        "output": 3.2,
+        "cache_read": 0.2,
+        "cache_write": 0
+      },
+      "limit": {
+        "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-4.5": {
+      "id": "glm-4.5",
+      "name": "GLM-4.5",
+      "family": "glm",
+      "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": 0.6,
+        "output": 2.2,
+        "cache_read": 0.11,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 131072,
+        "output": 98304
+      }
+    },
+    "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": "2025-07-28",
+      "last_updated": "2025-07-28",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0,
+        "output": 0,
+        "cache_read": 0,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 131072,
+        "output": 98304
+      }
+    },
+    "glm-4.7-flash": {
+      "id": "glm-4.7-flash",
+      "name": "GLM-4.7-Flash",
+      "family": "glm-flash",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2026-01-19",
+      "last_updated": "2026-01-19",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0,
+        "output": 0,
+        "cache_read": 0,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 200000,
+        "output": 131072
+      }
+    },
+    "glm-4.6": {
+      "id": "glm-4.6",
+      "name": "GLM-4.6",
+      "family": "glm",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2025-09-30",
+      "last_updated": "2025-09-30",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0.6,
+        "output": 2.2,
+        "cache_read": 0.11,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 204800,
+        "output": 131072
+      }
+    },
+    "glm-4.7": {
+      "id": "glm-4.7",
+      "name": "GLM-4.7",
+      "family": "glm",
+      "attachment": false,
+      "reasoning": true,
+      "tool_call": true,
+      "interleaved": {
+        "field": "reasoning_content"
+      },
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2025-12-22",
+      "last_updated": "2025-12-22",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0.6,
+        "output": 2.2,
+        "cache_read": 0.11,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 204800,
+        "output": 131072
+      }
+    },
+    "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,
+      "release_date": "2026-03-16",
+      "last_updated": "2026-03-16",
+      "modalities": {
+        "input": [
+          "text"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": false,
+      "cost": {
+        "input": 1.2,
+        "output": 4,
+        "cache_read": 0.24,
+        "cache_write": 0
+      },
+      "limit": {
+        "context": 200000,
+        "output": 131072
+      }
+    },
+    "glm-4.5v": {
+      "id": "glm-4.5v",
+      "name": "GLM-4.5V",
+      "family": "glm",
+      "attachment": true,
+      "reasoning": true,
+      "tool_call": true,
+      "temperature": true,
+      "knowledge": "2025-04",
+      "release_date": "2025-08-11",
+      "last_updated": "2025-08-11",
+      "modalities": {
+        "input": [
+          "text",
+          "image",
+          "video"
+        ],
+        "output": [
+          "text"
+        ]
+      },
+      "open_weights": true,
+      "cost": {
+        "input": 0.6,
+        "output": 1.8
+      },
+      "limit": {
+        "context": 64000,
+        "output": 16384
+      }
+    },
+    "glm-4.6v": {
+      "id": "glm-4.6v",
+      "name": "GLM-4.6V",
+      "family": "glm",
+      "attachment": true,
+      "reasoning": true,
+      "tool_call": true,
+      "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
+      }
+    }
+  }
+}

data/lib/llm/agent.rb

@@ -8,7 +8,7 @@ module LLM
   #
   # **Notes:**
   # * Instructions are injected only on the first request.
-  # * An agent will automatically execute tool calls (unlike {LLM::Session LLM::Session}).
+  # * An agent will automatically execute tool calls (unlike {LLM::Context LLM::Context}).
   # * The idea originally came from RubyLLM and was adapted to llm.rb.
   #
   # @example
@@ -23,6 +23,11 @@ module LLM
   #   agent = SystemAdmin.new(llm)
   #   agent.talk("Run 'date'")
   class Agent
+    ##
+    # Returns a provider
+    # @return [LLM::Provider]
+    attr_reader :llm
+
     ##
     # Set or get the default model
     # @param [String, nil] model
@@ -77,11 +82,10 @@ module LLM
     # @option params [String] :model Defaults to the provider's default model
     # @option params [Array<LLM::Function>, nil] :tools Defaults to nil
     # @option params [#to_json, nil] :schema Defaults to nil
-    def initialize(provider, params = {})
+    def initialize(llm, params = {})
       defaults = {model: self.class.model, tools: self.class.tools, schema: self.class.schema}.compact
-      @provider = provider
-      @ses = LLM::Session.new(provider, defaults.merge(params))
-      @instructions_applied = false
+      @llm = llm
+      @ctx = LLM::Context.new(llm, defaults.merge(params))
     end
 
     ##
@@ -99,13 +103,12 @@ module LLM
     #   puts response.choices[0].content
     def talk(prompt, params = {})
       i, max = 0, Integer(params.delete(:max_tool_rounds) || 10)
-      res = @ses.talk(apply_instructions(prompt), params)
-      until @ses.functions.empty?
+      res = @ctx.talk(apply_instructions(prompt), params)
+      until @ctx.functions.empty?
         raise LLM::ToolLoopError, "pending tool calls remain" if i >= max
-        res = @ses.talk @ses.functions.map(&:call), params
+        res = @ctx.talk @ctx.functions.map(&:call), params
         i += 1
       end
-      @instructions_applied = true
       res
     end
     alias_method :chat, :talk
@@ -126,40 +129,39 @@ module LLM
     #   puts res.output_text
     def respond(prompt, params = {})
       i, max = 0, Integer(params.delete(:max_tool_rounds) || 10)
-      res = @ses.respond(apply_instructions(prompt), params)
-      until @ses.functions.empty?
+      res = @ctx.respond(apply_instructions(prompt), params)
+      until @ctx.functions.empty?
         raise LLM::ToolLoopError, "pending tool calls remain" if i >= max
-        res = @ses.respond @ses.functions.map(&:call), params
+        res = @ctx.respond @ctx.functions.map(&:call), params
         i += 1
       end
-      @instructions_applied = true
       res
     end
 
     ##
     # @return [LLM::Buffer<LLM::Message>]
     def messages
-      @ses.messages
+      @ctx.messages
     end
 
     ##
     # @return [Array<LLM::Function>]
     def functions
-      @ses.functions
+      @ctx.functions
     end
 
     ##
     # @return [LLM::Object]
     def usage
-      @ses.usage
+      @ctx.usage
     end
 
     ##
-    # @param (see LLM::Session#prompt)
-    # @return (see LLM::Session#prompt)
-    # @see LLM::Session#prompt
+    # @param (see LLM::Context#prompt)
+    # @return (see LLM::Context#prompt)
+    # @see LLM::Context#prompt
     def prompt(&b)
-      @ses.prompt(&b)
+      @ctx.prompt(&b)
     end
     alias_method :build_prompt, :prompt
 
@@ -169,7 +171,7 @@ module LLM
     # @return [LLM::Object]
     #  Returns a tagged object
     def image_url(url)
-      @ses.image_url(url)
+      @ctx.image_url(url)
     end
 
     ##
@@ -178,7 +180,7 @@ module LLM
     # @return [LLM::Object]
     #  Returns a tagged object
     def local_file(path)
-      @ses.local_file(path)
+      @ctx.local_file(path)
     end
 
     ##
@@ -187,54 +189,53 @@ module LLM
     # @return [LLM::Object]
     #  Returns a tagged object
     def remote_file(res)
-      @ses.remote_file(res)
+      @ctx.remote_file(res)
     end
 
     ##
     # @return [LLM::Tracer]
     #  Returns an LLM tracer
     def tracer
-      @ses.tracer
+      @ctx.tracer
     end
 
     ##
     # Returns the model an Agent is actively using
     # @return [String]
     def model
-      @ses.model
+      @ctx.model
     end
 
     ##
-    # @param (see LLM::Session#serialize)
-    # @return (see LLM::Session#serialize)
+    # @param (see LLM::Context#serialize)
+    # @return (see LLM::Context#serialize)
     def serialize(**kw)
-      @ses.serialize(**kw)
+      @ctx.serialize(**kw)
     end
     alias_method :save, :serialize
 
     ##
-    # @param (see LLM::Session#deserialize)
-    # @return (see LLM::Session#deserialize)
+    # @param (see LLM::Context#deserialize)
+    # @return (see LLM::Context#deserialize)
     def deserialize(**kw)
-      @ses.deserialize(**kw)
+      @ctx.deserialize(**kw)
     end
     alias_method :restore, :deserialize
 
     private
 
-    def apply_instructions(prompt)
+    ##
+    # @return [LLM::Prompt]
+    def apply_instructions(new_prompt)
       instr = self.class.instructions
-      return prompt unless instr
-      if LLM::Prompt === prompt
-        messages = prompt.to_a
-        prompt = LLM::Prompt.new(@provider)
-        prompt.system instr unless @instructions_applied
-        messages.each { |msg| prompt.talk(msg.content, role: msg.role) }
-        prompt
+      return new_prompt unless instr
+      if LLM::Prompt === new_prompt
+        @ctx.messages.empty? ? new_prompt.system(instr) : nil
+        new_prompt
       else
         prompt do
-          system instr unless @instructions_applied
-          user prompt
+          @ctx.messages.empty? ? _1.system(instr) : nil
+          _1.user(new_prompt)
         end
       end
     end