AI features

Overview

Your app can use AI to chat with users, draft content, summarise long text, recommend things, classify input, and more. You don't pick a model, you don't manage API keys, and you don't write prompt engineering plumbing — describe the feature you want in chat and Pentoggle wires up everything: the model, the personality, the access rules, and the in-app UI.

What you can build

• Chatbots and assistants — support agents, sales bots, in-app tutors, anything conversational
• Content generation — blog drafts, product descriptions, email replies, marketing copy
• Summaries & analysis — TL;DRs of long documents, sentiment, key takeaways
• Smart suggestions — auto-tags, recommended next actions, related items
• Classification & routing — categorise a support ticket, score a lead, detect spam
• Multiple AI features per app — each one tuned independently (different personality, different audience, different output length)

Example prompts

"Add a support chatbot that answers questions using my
FAQ page. Keep replies short and friendly."

"Let admins click 'Auto-draft' on a blog post and generate
the body from the title and 3 bullet points."

"When a contact form is submitted, generate a one-sentence
summary of the message for the admin email."

"Make the chatbot's tone more formal and limit replies
to about 150 words."

Setup (none required)

AI features are fully managed by Pentoggle. No API keys to paste, no provider account, no model selection, no infrastructure. Ask the agent for the AI feature you want and it picks a high-quality model, drafts a sensible personality, and sets reasonable defaults.

Tweaking an AI feature

Each AI feature in your app shows up as a card in the Settings tab under AI features. The card shows, in plain language:

• What it does — a one-line summary (e.g., "Support chatbot for the FAQ page")
• Personality & instructions — the rules the AI follows (tone, format, what to refuse to answer, etc.)
• Reply length — roughly how long responses should be
• Who can use it — public visitors, signed-in users only, admins only, etc.

Each card has an Edit via chat button. Click it, describe what you want changed in plain language ("make the chatbot's tone more formal", "limit replies to two paragraphs", "only let signed-in users use this"), and the agent applies the change.

How it works

1. You describe an AI feature in chat (e.g., "add a chatbot for my FAQ")
2. The agent creates the in-app UI (a chat panel, a button, a function call) plus an AI configuration that holds the instructions, length cap, and access rules
3. When a user triggers the feature, Pentoggle checks who they are against the access rules
4. The configured instructions are prepended to the user's input and sent to the underlying model
5. The response streams back into your app, capped at the configured length
6. Usage is metered against your app's live capacity and shows up on your usage dashboard

Limitations

• No raw model picking — you describe what you want in chat; you can't paste a specific provider/model name. The platform chooses for you and upgrades over time
• No bring-your-own API key — AI runs on platform-managed inference. You can't route through your own OpenAI/Anthropic/etc. account
• Counts against live capacity — heavy AI usage shows up on your usage dashboard alongside database and backend calls
• Reply length is a soft cap — the platform aims for the configured length but a particularly chatty model may go slightly over before stopping

Troubleshooting

• The AI sounds off — open the card in Settings, click Edit via chat, and describe the tone or format you want in plain language
• Replies get cut off mid-sentence — the reply length is too short for what the AI is being asked. Ask the agent to "increase reply length on the support chatbot"
• "You don't have access" from a public page — the access rule is too strict. Ask the agent to "allow visitors to use the support chatbot without signing in"
• Suddenly slower than before — the platform may have rotated to a newer model. Check the card; if the new default doesn't fit, ask the agent to switch to a faster style ("prefer speed over depth on this chatbot")
• The AI keeps suggesting things you don't want — tighten the personality instructions via Edit via chat with a specific rule ("never recommend competitor products")