Class: LLM::Message
- Inherits:
-
Object
- Object
- LLM::Message
- Defined in:
- lib/llm/message.rb
Instance Attribute Summary collapse
-
#role ⇒ Symbol
readonly
Returns the role of the message.
-
#content ⇒ String
readonly
Returns the content of the message.
-
#extra ⇒ Hash
readonly
Returns extra context associated with the message.
-
#id ⇒ String
readonly
Returns a stable id for this message.
Instance Method Summary collapse
-
#inspect ⇒ String
Returns a string representation of the message.
-
#created_at ⇒ Time?
Returns the time this message was created, derived from the timestamp embedded in its UUIDv7 id, or nil when it is not.
-
#to_h ⇒ Hash
Returns a Hash representation of the message.
- #to_json ⇒ String
-
#==(other) ⇒ Boolean
(also: #eql?)
Returns true when two objects have the same role and content.
- #contents ⇒ Array<LLM::Function::Return, String>
-
#content! ⇒ Hash
Try to parse JSON content.
-
#reasoning_content ⇒ String?
Returns reasoning content associated with the message.
-
#compaction? ⇒ Boolean
Returns true when a message was created by context compaction.
-
#image_url? ⇒ Boolean
Returns true when a message contains an image URL.
-
#image_urls ⇒ Array<LLM::Object>
Returns image URL content items from the message.
-
#file? ⇒ Boolean
Returns true when a message contains a local or remote file.
-
#files ⇒ Array<LLM::Object>
Returns local and remote file content items from the message.
- #functions ⇒ Array<LLM::Function>
-
#assistant? ⇒ Boolean
Returns true when the message is an assistant message.
-
#system? ⇒ Boolean
Returns true when the message is a system message.
-
#user? ⇒ Boolean
Returns true when the message is a user message.
-
#tool_call? ⇒ Boolean
Returns true when the message requests a function call.
-
#tool_return? ⇒ Boolean
Returns true when the message represents a function return.
- #returns ⇒ Array<LLM::Function::Return>
-
#response ⇒ LLM::Response?
Returns the response associated with the message, or nil.
-
#annotations ⇒ Array<LLM::Object>
Returns annotations associated with the message.
-
#token_usage ⇒ LLM::Usage?
(also: #usage)
Returns token usage statistics.
-
#model ⇒ String?
Returns the model associated with a message.
-
#initialize(role, content, extra = {}) ⇒ LLM::Message
constructor
Returns a new message.
Constructor Details
#initialize(role, content, extra = {}) ⇒ LLM::Message
Returns a new message
32 33 34 35 36 37 |
# File 'lib/llm/message.rb', line 32 def initialize(role, content, extra = {}) @content = content @role = normalize_role(role) @extra = LLM::Object.from(extra) @id = extra[:id] || SecureRandom.uuid_v7 end |
Instance Attribute Details
#role ⇒ Symbol (readonly)
Returns the role of the message
8 9 10 |
# File 'lib/llm/message.rb', line 8 def role @role end |
#content ⇒ String (readonly)
Returns the content of the message
13 14 15 |
# File 'lib/llm/message.rb', line 13 def content @content end |
#extra ⇒ Hash (readonly)
Returns extra context associated with the message
18 19 20 |
# File 'lib/llm/message.rb', line 18 def extra @extra end |
#id ⇒ String (readonly)
Returns a stable id for this message. It is generated once on creation and restored with the runtime state on load.
24 25 26 |
# File 'lib/llm/message.rb', line 24 def id @id end |
Instance Method Details
#inspect ⇒ String
Returns a string representation of the message
252 253 254 255 256 |
# File 'lib/llm/message.rb', line 252 def inspect "#<#{LLM::Utils.object_id(self)} " \ "tool_call=#{tool_calls.any?} role=#{role.inspect} " \ "content=#{content.inspect} reasoning_content=#{reasoning_content.inspect}>" end |
#created_at ⇒ Time?
Returns the time this message was created, derived from the timestamp embedded in its UUIDv7 id, or nil when it is not.
43 44 45 |
# File 'lib/llm/message.rb', line 43 def created_at @created_at ||= LLM::Utils.(@id) end |
#to_h ⇒ Hash
Returns a Hash representation of the message.
This is the whole of a message as far as storage is concerned: Context#to_h maps its messages through here, and restore reads them back from the same list. A field that is not named here does not survive a save, so a field the runtime relies on after a restore has to be named.
Tool calls are stored in the abstract form the runtime reasons
about - {id, name, arguments} - and never in a provider's wire
form. Each request adapter builds its own blocks from that form,
which is what lets a conversation saved under one provider be
restored under another.
62 63 64 65 66 67 68 69 70 71 72 73 |
# File 'lib/llm/message.rb', line 62 def to_h { id: @id, role:, content:, reasoning_content:, compaction: extra.compaction, instructions: extra.instructions, tools: extra.tool_calls&.map { LLM::Object === _1 ? _1.to_h : _1 }, usage: }.compact.then { preserve_nil_content(_1) } end |
#to_json ⇒ String
77 78 79 |
# File 'lib/llm/message.rb', line 77 def to_json(...) LLM.json.dump(to_h, ...) end |
#==(other) ⇒ Boolean Also known as: eql?
Returns true when two objects have the same role and content. The id is ignored, because it identifies a message rather than describing it.
88 89 90 91 92 93 94 |
# File 'lib/llm/message.rb', line 88 def ==(other) if other.respond_to?(:to_h) to_h.except(:id) == other.to_h.except(:id) else false end end |
#contents ⇒ Array<LLM::Function::Return, String>
99 100 101 |
# File 'lib/llm/message.rb', line 99 def contents [*content].flatten end |
#content! ⇒ Hash
Try to parse JSON content
107 108 109 |
# File 'lib/llm/message.rb', line 107 def content! LLM.json.load(content) end |
#reasoning_content ⇒ String?
Returns reasoning content associated with the message
114 115 116 |
# File 'lib/llm/message.rb', line 114 def reasoning_content extra.reasoning_content end |
#compaction? ⇒ Boolean
Returns true when a message was created by context compaction
121 122 123 |
# File 'lib/llm/message.rb', line 121 def compaction? !!extra.compaction end |
#image_url? ⇒ Boolean
Returns true when a message contains an image URL
128 129 130 |
# File 'lib/llm/message.rb', line 128 def image_url? image_urls.any? end |
#image_urls ⇒ Array<LLM::Object>
Returns image URL content items from the message
135 136 137 |
# File 'lib/llm/message.rb', line 135 def image_urls contents.select { LLM::Object === it and it.kind == :image_url } end |
#file? ⇒ Boolean
Returns true when a message contains a local or remote file
142 143 144 |
# File 'lib/llm/message.rb', line 142 def file? files.any? end |
#files ⇒ Array<LLM::Object>
Returns local and remote file content items from the message
149 150 151 152 |
# File 'lib/llm/message.rb', line 149 def files kinds = [:local_file, :remote_file] contents.select { LLM::Object === it and kinds.include?(it.kind) } end |
#functions ⇒ Array<LLM::Function>
156 157 158 159 160 161 162 163 |
# File 'lib/llm/message.rb', line 156 def functions @functions ||= tool_calls.filter_map do |fn| function = available_tools.find { _1.name.to_s == fn.name } || LLM::Tool::NotFound(name: fn.name) function = function.dup function.tap { _1.id = fn.id } function.tap { _1.arguments = fn.arguments } end end |
#assistant? ⇒ Boolean
Returns true when the message is an assistant message
168 169 170 |
# File 'lib/llm/message.rb', line 168 def assistant? role == "assistant" || role == "model" end |
#system? ⇒ Boolean
Returns true when the message is a system message
175 176 177 |
# File 'lib/llm/message.rb', line 175 def system? role == "system" end |
#user? ⇒ Boolean
Returns true when the message is a user message
182 183 184 |
# File 'lib/llm/message.rb', line 182 def user? role == "user" end |
#tool_call? ⇒ Boolean
Returns true when the message requests a function call
189 190 191 |
# File 'lib/llm/message.rb', line 189 def tool_call? tool_calls.any? end |
#tool_return? ⇒ Boolean
Returns true when the message represents a function return
196 197 198 |
# File 'lib/llm/message.rb', line 196 def tool_return? returns.size > 0 end |
#returns ⇒ Array<LLM::Function::Return>
202 203 204 205 206 207 208 |
# File 'lib/llm/message.rb', line 202 def returns if LLM::Function::Return === content [content] else [*content].flatten.grep(LLM::Function::Return) end end |
#response ⇒ LLM::Response?
This method returns a response for assistant messages, and it returns nil for non-assistant messages
Returns the response associated with the message, or nil
216 217 218 |
# File 'lib/llm/message.rb', line 216 def response extra.response end |
#annotations ⇒ Array<LLM::Object>
This method might return annotations for assistant messages, and it returns an empty array for non-assistant messages
Returns annotations associated with the message
226 227 228 |
# File 'lib/llm/message.rb', line 226 def annotations @annotations ||= LLM::Object.from(extra.annotations || []) end |
#token_usage ⇒ LLM::Usage? Also known as: usage
This method returns token usage for assistant messages, and it returns nil for non-assistant messages
Returns token usage statistics
236 237 238 239 |
# File 'lib/llm/message.rb', line 236 def token_usage return nil unless assistant? @token_usage ||= LLM::Usage.from(extra.usage || response&.usage) end |
#model ⇒ String?
Returns the model associated with a message
245 246 247 |
# File 'lib/llm/message.rb', line 245 def model response&.model end |