Admin Settings

Token usage, model & prompt configuration, maintenance mode, metadata refresh, and session/endpoint diagnostics for the ferry chatbot.

Token Usage

Aggregated OpenAI token consumption, with optional session/run/date filters. Read-only — viewing it has no effect on the bot.

Loading token usage…

Capabilities: loading…

Input Tokens

N/A

Output Tokens

N/A

Total Tokens

N/A

Cached Tokens

N/A

Cost

N/A

Model Selection

Choose the provider + model the chatbot runs on, and (optionally) the temperature. Saving applies to new chats only; active sessions finish on their current model.

Temperature controls randomness. Lower (e.g. 0.2) = more focused, consistent, repeatable answers — best for factual ferry/translation replies. Higher (e.g. 0.8–1.0) = more varied, creative wording. Leave blank to use the system default.

Loading model selection…

Applies to new conversations. Active sessions finish on their current model.

Prompt Management

Every bot has its own instructions — pick a bot above to read or edit them. Two shared pieces are filled in automatically each time the bot answers a customer, wherever their marker appears in the text: {{COMMON_PROMPT}} is replaced by the Common Prompt (the instructions every bot shares), and {{RULES}} is replaced by the Rules you manage in the section below. You only write this bot's own wording here; the shared parts are added for you, so you don't need to copy them in.

Loading prompts…

Rule Assistant

Paste a travel operator's rough instruction and let an AI turn it into a clean, ready-to-use rule — then check whether it clashes with the Common Prompt or an existing rule, or is already covered. Review the result and click Add as rule. Nothing is saved until you do.

Bot:
Provider:
Check with:

Prompt Rules

Short, extra instructions added on top of a bot's own instructions — handy for small tweaks you want to switch on or off without editing a bot's main text. Each rule can apply to every bot or just one, and to every AI provider or just one; set both from the dropdowns on the rule (a new rule applies everywhere by default). Untick Enabled to pause a rule without deleting it.

Loading prompt rules…

Bot:
Provider:

    Voice Bridge (ElevenLabs)

    The ElevenLabs voice-agent ID embedded on the Voice Demo page. The voice bot's prompt is managed under Prompt Management → Voice Bot.

    Loading agent ID…

    Maintenance Mode

    Gate the chat endpoint with a message. Turning ON makes every chat return the message below instead of answering — the LLM is not called.

    When ON, every chat request immediately returns the message below instead of running the bot. The LLM is not called, no tools fire, and no transcript is recorded for the gated turn.

    Loading maintenance state…

    Last updated: —

    Metadata CRON

    Schedule for the background GTFS metadata refresh (ports, vessels, accommodations, etc.). Updating the expression reschedules the job; Trigger Refresh runs it now.

    Loading metadata CRON status…

    Last Run

    N/A

    Schedule

    N/A

    Chat Sessions

    Browse chat sessions and download their transcripts. Read-only — no effect on the bot.

    Loading sessions…

    Session history

    Select a session to view history.

    
          
    
        
    
      
    

      
    
        
    
          
    Endpoint Health
    
          
    
            
            
            
          
    
        
    
        
    Live status of the metadata/data endpoints (and optional trace/transcript checks for a session). Read-only diagnostics.
    

        
    
          
    
            
              
    
                Endpoint
                URL
                Status
                Last checked
                Preview
                Action