step9

#!/usr/bin/env ruby

require "bundler"
Bundler.require(:default)
Dotenv.load

require_relative "lib/render"

renderer = Render.new

prompt = "Use the python-mnemonic library to generate a mnemonic code in English for the phrase 'Heroku AI'"

request = <<~JSON
{
  "model": "claude-3-5-sonnet",
  "messages": [
    {
      "role": "user",
      "content": "#{prompt}"
    }
  ],
  "tools": [
    {
      "type": "heroku_tool",
      "function": "code_exec_python"
    }
  ],
  "tool_choice": "auto"
}
JSON

renderer.heroku_print("Here is the request we are sending to Heroku for inferencing...")

renderer.print_markdown(
  <<~MARKDOWN
  ```json
  #{request}
  ```
  MARKDOWN
)

answer = renderer.confirm("Ready to send the inference request?")
exit(0) unless answer

response = renderer.inference_request(request:, print_last_message: false)
renderer.print_large_box(response.dig("choices", 0, "message", "content"))

# Let's print out the generated code
renderer.all_responses.each do |r|
  msg = r.dig("choices", 0, "message")
  tool_call = msg.dig("tool_calls", 0)
  if tool_call
    args = JSON.parse(tool_call.dig("function", "arguments"))
    if tool_call.dig("function", "name").match(/\Acode_exec_python/)
      renderer.heroku_print("Here is what was run on the Dyno's shell prior to executing the code:")
      renderer.print_markdown(args["shell_command"])
      renderer.heroku_print("Here is the generated code that was run:")
      python_code = <<~PYTHON
      ```python
      #{args["code"]}
      ```
      PYTHON
      renderer.print_markdown(python_code)
    end
  end
end

renderer.hr