Class: Omakase::Agent

Inherits:
Object
  • Object
show all
Defined in:
lib/omakase/agent.rb

Overview

Fields are state, methods are what the model can call, generates declares the methods the model implements.

Constant Summary collapse

RUNNING =

The generations this thread is inside, so one cannot re-enter itself.

:omakase_running

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(chat: nil) ⇒ Agent

chat: injects a prepared RubyLLM::Chat — the seam for tests.



131
132
133
 
# File 'lib/omakase/agent.rb', line 131

def initialize(chat: nil)
  @chat = chat
end

Class Method Details

.chat_options ⇒ Object 



86
 
# File 'lib/omakase/agent.rb', line 86

def chat_options = @chat_options ||= {}

.describe(text) ⇒ Object

Documents the method defined next — the docstring Ruby does not have.



53
54
55
 
# File 'lib/omakase/agent.rb', line 53

def describe(text)
  @pending_description = text
end

.descriptions ⇒ Object 



84
 
# File 'lib/omakase/agent.rb', line 84

def descriptions = @descriptions ||= {}

.generates(name, prompt = nil, takes: nil, returns: nil, strategy: nil, model: nil, &schema) ⇒ Object

Without a prompt, the method name is the prompt. A block instead of a string is a prompt read at call time, on the agent. takes: names the keyword arguments, and then Ruby checks them.



60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
 
# File 'lib/omakase/agent.rb', line 60

def generates(name, prompt = nil, takes: nil, returns: nil, strategy: nil, model: nil, &schema)
  # Redeclaring an inherited generation is how a subclass specialises one.
  # Landing on a method you wrote is not that, and would replace it unseen.
  if Capabilities.names(self).include?(name) && !generations.key?(name)
    raise Error, "#{self}##{name} is already a method — generates would replace it"
  end

  unless prompt.nil? || prompt.is_a?(String) || prompt.is_a?(Proc)
    raise Error, "#{self}##{name}: a prompt is a String or a block returning one, got #{prompt.class}"
  end

  generations[name] = Generation.new(
    name:,
    prompt: prompt || humanize(name),
    schema: Schema.define(returns:, &schema),
    strategy: Strategies.fetch(strategy || self.strategy),
    model:
  )
  define_generation_method(name, takes)
  define_singleton_method(name) { |**inputs| new.public_send(name, **inputs) }
end

.generations ⇒ Object 



82
 
# File 'lib/omakase/agent.rb', line 82

def generations = @generations ||= {}

.instructions(text = nil) ⇒ Object 



20
21
22
23
24
 
# File 'lib/omakase/agent.rb', line 20

def instructions(text = nil)
  return @instructions.to_s if text.nil?

  @instructions = text
end

.mcp(name, **options) ⇒ Object

An MCP server's tools, as methods on the agent. Options are passed to ruby_llm-mcp verbatim: mcp :files, transport_type: :stdio, config: {command: "npx", …}.



34
35
36
37
 
# File 'lib/omakase/agent.rb', line 34

def mcp(name, **options)
  require "ruby_llm/mcp"
  MCP.attach(self, RubyLLM::MCP.add_client(name: name.to_s, **options))
end

.memory ⇒ Object

Two more methods: one to save something, one to search it by meaning. The store is a field, so it marshals with the agent and outlives the run.



45
46
47
48
49
50
 
# File 'lib/omakase/agent.rb', line 45

def memory
  describe "Save something worth remembering after this run"
  define_method(:remember) { |text| (@memory ||= Memory.new).remember(text) }
  describe "Search what you remember, by meaning; the closest few come back"
  define_method(:recall) { |query, limit: 5| (@memory ||= Memory.new).recall(query, limit:) }
end

.model(id = nil, **options) ⇒ Object

The model and any RubyLLM chat option. Naming a provider takes the model id on trust, since providers like OpenRouter or Ollama serve ids that are not in RubyLLM's registry.



13
14
15
16
17
18
 
# File 'lib/omakase/agent.rb', line 13

def model(id = nil, **options)
  return chat_options if id.nil? && options.empty?

  options = {assume_model_exists: true, **options} if options[:provider]
  @chat_options = {model: id, **options}.compact
end

.skill(path) ⇒ Object

A skill directory — a SKILL.md with YAML front matter. Its description joins the agent's capabilities; its body arrives when the model asks.



41
 
# File 'lib/omakase/agent.rb', line 41

def skill(path) = Skills.attach(self, path)

.strategy(name = nil) ⇒ Object 



26
27
28
29
30
 
# File 'lib/omakase/agent.rb', line 26

def strategy(name = nil)
  return @strategy || :code_act if name.nil?

  @strategy = name
end

Instance Method Details

#chat(**overrides) ⇒ Object

A fresh conversation per call — two threads calling one agent must not share a mutable chat. What carries between calls is the object's own state. Overrides land on top of the class's options; an injected chat ignores them.



138
 
# File 'lib/omakase/agent.rb', line 138

def chat(**overrides) = @chat || Omakase.chat_factory.call(**self.class.chat_options.merge(overrides))

#context ⇒ Object

That state, as the model should read it: rebuilt on every call, and added to the class's instructions. Override it to remember anything.



142
 
# File 'lib/omakase/agent.rb', line 142

def context = nil

#doc(object) ⇒ Object

For generated code meeting an object whose type it does not know.



151
 
# File 'lib/omakase/agent.rb', line 151

def doc(object) = puts(Doc.of(object))

#finish(value) ⇒ Object

How generated code answers: with the value itself.



154
 
# File 'lib/omakase/agent.rb', line 154

def finish(value) = throw(Executor::RESULT, value)

#marshal_dump ⇒ Object

Resuming a run is loading the object back, so an agent marshals like any other Ruby object — minus the live chat, which is rebuilt on demand.



146
 
# File 'lib/omakase/agent.rb', line 146

def marshal_dump = (instance_variables - [:@chat]).to_h { |name| [name, instance_variable_get(name)] }

#marshal_load(state) ⇒ Object 



148
 
# File 'lib/omakase/agent.rb', line 148

def marshal_load(state) = state.each { |name, value| instance_variable_set(name, value) }

#p(*args) ⇒ Object Also known as: pp 



162
163
164
165
 
# File 'lib/omakase/agent.rb', line 162

def p(*args)
  args.each { |arg| omakase_output.puts(arg.inspect) }
  (args.size <= 1) ? args.first : args
end

#print(*args) ⇒ Object 



160
 
# File 'lib/omakase/agent.rb', line 160

def print(*args) = omakase_output.print(*args)

#puts(*args) ⇒ Object

Printing from generated code goes to the observation, not to the process's stdout — and the buffer is per thread, so concurrent agents stay separate.



158
 
# File 'lib/omakase/agent.rb', line 158

def puts(*args) = omakase_output.puts(*args)