A step-by-step guide to every menu in the AI dashboard — from building agents, training NLP, and testing conversations to managing quota, plans, and API keys.
The first page after signing in: quota, usage trend, agents, and AI service status at a glance.
Overview — quota cards, token trend & distribution charts, agent list, and service status.
Step by step
1
Sign in with your Resmi.ID account via Open Dashboard. If you manage more than one business, pick it at the top right — all data follows that business.
2
The four top cards: this month's Token usage and AI replies against your plan limit, Requests today, and the number of Active agents.
3
The Usage Trend chart shows tokens per day — switch the range with 7 days, 30 days, or 90 days. Usage Distribution compares input and output tokens.
4
Below are Agents for This Business and Service Status (Gemini, local NLP, the AI API). The Current plan card in the left menu always shows remaining quota.
⚠
Quota running low? An amber warning appears at 75% and a red one when it runs out — after that the AI stops replying until the next period. Replies from trained NLP don't use token quota, so training your agents makes the quota last longer.
Quota cards & hero1 / 2
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Test how an agent answers before it goes live on WhatsApp.
Try AI — pick an agent, send a message, and see whether the reply came from NLP or Gemini.
Step by step
1
Choose an agent in Choose Agent (active agents only). The status beside it reads NLP trained, NLP not trained, or Gemini only when no agent is selected.
2
Write a message the way a customer would, then press Send or Enter (Shift+Enter for a new line). Up to 4,000 characters. Sample Questions on the right come from the agent's training data — click one to fill it in.
3
Under each reply you'll see its source: NLP with the intent and confidence, or Gemini when NLP doesn't recognise the question yet.
4
Wrong answer, or too many replies falling back to Gemini? Press Train this agent's NLP to add more sample questions.
ℹ
NLP is used when its confidence is at least 75% — below that, Gemini answers. Gemini replies count against your monthly token and reply quota; NLP replies don't. Conversation history is kept per agent; switching agents starts a new conversation.
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Create an AI agent for each need — sales, support, customer service — each with its own instructions and tone.
Agents — the agent list with status, channels, intent count, and the new agent form.
Step by step
1
Press New Agent. Fill in the Agent name (required) and Skill / Instructions — the agent's job and limits, up to 8,000 characters — then pick a Tone of voice: Formal, Friendly, Casual, or Direct. Press Save.
2
In the list, each card shows the tone, the channels using the agent (WhatsApp, Telegram, Direct chat), the number of intents, and whether its NLP is trained.
3
Use the switch to turn an agent on or off, Edit to change it, and the ⋯ menu to delete it. Search and filter by status above the list.
4
Press Train NLP on an agent card to teach it customer questions — see the next section.
ℹ
Up to 20 agents per business. A new agent isn't attached to any channel yet — connect it from that channel's own settings, for example the AI Agent tab in the WhatsApp dashboard. Agent instructions add to the base instructions in AI Settings; they don't replace them.
Agent list1 / 2
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Teach an agent to recognise customer questions with example phrases — its answers are sent without using Gemini quota.
Train NLP — the intent list, the sample question editor, and the Test Model tab.
Step by step
1
Press New Intent. Fill in the Intent name (lowercase letters, digits, dots, hyphens — e.g. harga.paket), the Answer sent to customers, and several Sample Questions. Paste many lines at once — each line becomes one example.
2
Press Train Now. Training runs in the background — you can leave the page. The status changes to Model ready when it's done.
3
Open the Test Model tab, write a question the way a customer would, then press Test. The result shows the answer, intent, and confidence — this test doesn't use Gemini quota.
💡
Lots of data? Use Download Excel, then Import Excel (.xlsx, max 5 MB). The columns are Nama Intent, Jawaban, and Contoh Kalimat — one row per example. Importing replaces every intent of that agent, so download a backup first. Each intent needs 1–100 examples (max 255 characters); inactive intents aren't trained.
Step 1 · Intents1 / 3
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Choose an AI plan, continue payment, and view your invoice history.
Plans & Billing — plan cards with token & reply quotas, plus payment history.
Step by step
1
On the Subscription Plans tab, compare each plan's token and reply quota, then press Choose Plan (or Switch to This Plan if you already subscribe).
2
The resmi.id payment page opens. Transfer the exact amount — its last three digits are a unique code so the payment is detected automatically. AI invoices are issued alongside your other resmi.id service invoices.
3
The plan becomes active once payment is received. Haven't finished paying? The button changes to Continue Payment. Every invoice is in the Payment History tab.
ℹ
Need a bigger quota or a custom integration? The Enterprise card → Contact Sales opens a WhatsApp chat with our team.
Subscription plans1 / 2
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Set AI behaviour for every agent in the business: on/off, tone, answer source, language, and limits.
AI Settings — the business settings form, sample reply, and base instructions (system prompt).
Step by step
1
Enable AI for this business — when it's off, every agent stops replying.
2
Choose the Default tone of voice, the Answer source (Hybrid — NLP first, then Gemini, the most economical; NLP only; or Gemini only), and the Main language (Indonesian, English, or match the customer).
3
Set the limits: NLP confidence threshold (0–1, default 0.6), Daily token limit (empty = follow the plan quota), and Maximum reply length (100–4,000 characters, Gemini replies only).
4
Write the Base Instructions (max 8,000 characters) or use an Instruction Template, then press Save Settings. Test the result in Sample Reply with Try it.
⚠
Save before testing — the Sample Reply uses the saved settings and consumes token quota. A tone set on an agent overrides the default tone here.
Settings form1 / 2
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Connect AI Resmi.ID to your apps, website, or internal systems with API keys.
Developer — the API key list, the new key form, and endpoint documentation with code samples.
Step by step
1
Press New Key. Fill in the Key name, choose the Environment (Production or Test), tick the Permissions — ai:chat to send messages, ai:read to read agents & usage — and the Expiry (1–365 days, empty = never).
2
Press Create Key and copy the key right away — it's shown only once. In the list, keys appear masked with their permissions, creation time, and last use.
3
Open the Documentation tab: send the key in the x-api-key header to https://ai.resmi.id. Endpoints: GET /api-dev/agents, GET /api-dev/usage, and POST /api-dev/chat with message & conversation_id — code samples in cURL, Node.js, PHP, and Python are included.
⚠
Keep keys secret — never put them in shared code or browser apps. Up to 10 active keys per business, 120 requests per minute per key. Key leaked? Open the ⋯ menu → Revoke; systems using it stop immediately.
Step 1 · API key list1 / 3
All screenshots use fictional sample data (a business called "Toko Sejahtera"), not real customer data.
Still have questions?
Our team is ready to help you set up agents, training data, or API integrations.
