Use case

Support triage that lives in your Rails app.

Route tickets, summarise conversations, and draft first replies with a single agent class. Deploys with `rails agent:deploy`.

The agent

class SupportTriage < RailsAgents::Base
  model "gpt-4o-mini"
  tool :fetch_ticket, using: Zendesk::FetchTicketTool
  tool :tag_ticket,   using: Zendesk::TagTicketTool
  instructions <<~P
    Summarise the ticket in 2 sentences, tag urgency
    (low, med, high), and suggest a first reply.
  P
end

Wire it into Sidekiq

class TriageJob < RailsAgent::AgentJob
  agent SupportTriage

  def on_success(result, ticket:)
    ticket.update!(ai_summary: result.summary, ai_tag: result.tag)
  end
end

TriageJob.perform_later(ticket_id: ticket.id)

What you get in the dashboard

  • Live trace of every triage run
  • Cost per ticket, aggregated by day
  • One-click rollback if a prompt regresses
  • Human-review queue for low-confidence tags