Blog UzzAIGetting Started
Initial Setup — Configure Your AI Assistant in 10 Minutes
After connecting WhatsApp, configure the assistant prompt, choose the AI model and run your first conversation test.
05/03/2026•6 min•Por Equipe Uzz.Ai
Role para descobrir a jornada completa
SUMÁRIO12 seções
With WhatsApp connected, the next step is to configure your AI assistant's behavior. UzzApp uses an "Agents" system — each agent has a system prompt (the instructions that define how it should behave), an AI model and response settings.
What you will configure in this setup
- System prompt: the assistant's instructions (name, tone, what it can and cannot answer)
- AI model: which AI engine will be used (Groq Llama or OpenAI GPT-4o)
- Temperature and tokens: control over creativity and response length
- Language and time zone
- Functionality test before going live
Step 1 — Access the Agents Manager
- 1In the UzzApp dashboard, click "Agents" in the side menu.
- 2You will see the default agent already created. Click "Edit".
- 3If you want to create a new agent from scratch, click "New Agent".
Step 2 — Write the System Prompt
- 1In the agent editor, find the "System Prompt" field.
- 2Write the assistant's instructions. Define: assistant name, company name, tone of voice, what it can answer and what should be escalated to a human.
- 3Be specific: the more context you provide, the better the service will be.
- 4Save the prompt by clicking "Save Draft" before activating.
Example System Prompt
You are Sofia, the virtual assistant for Company XYZ. Respond in English with a professional and friendly tone. You can answer about: business hours (Mon–Fri 8am–6pm), products from the catalogue, exchange and return policies (7 days), and scheduling. If the customer wants to talk to a human or the topic is not covered above, use the transfer-to-agent tool.
Step 3 — Choose the AI Model
- 1In the "AI Provider" field, select Groq (Llama 3.3 70B) or OpenAI (GPT-4o).
- 2Groq is recommended for most cases: ultra-fast and lower cost.
- 3OpenAI GPT-4o is recommended if you need image analysis (vision) or more elaborate responses.
- 4Make sure the API key of the chosen provider is configured in Settings → Credentials.
Seção não suportada (índice 8).
Step 4 — Configure Advanced Parameters (Optional)
- Temperature (0.0 to 1.0): controls creativity. 0.3–0.7 is ideal for service. High values make the bot more "creative" but less predictable.
- Maximum Tokens: response size limit. 500–1000 tokens is adequate for most cases.
- Message History: how many previous messages the AI considers (default: 15).
Step 5 — Activate the Agent and Test
- 1Click "Publish" to activate the agent.
- 2Send a test message to the connected WhatsApp number.
- 3Check the response in Dashboard → Conversations.
- 4If the response was not adequate, return to the editor and adjust the prompt.
- 5Repeat the test until the behavior is correct.
Agent not responding?
Check: (1) Is the agent status "Active"? (2) Are the AI provider API credentials configured and correct? (3) Is the WhatsApp webhook active? Access Dashboard → Debug to see the processing logs for each message.