[Extension] How to use the Google Gemini extension in Blip

1. Introduction

2. Features

3. Installation and setup

4. Using Gemini in the Builder flow

5. Usage examples

6. Support

Brief description: Connect Google Gemini artificial intelligence models to your chatbot in Blip and generate automatic responses based on your instructions and your own knowledge base.

Introduction

The Google Gemini extension brings Google's AI models into your chatbot flows in Blip. You configure the API key, choose the model, and define how the assistant should behave, and Gemini then generates responses within the Builder flow.

For example, a customer asks on WhatsApp about the delivery time for a product. The Gemini block receives the message, consults the documents you added to the knowledge base, and returns a response in natural language, which the flow sends back to the customer.

Features

• Generate responses with Google's AI (Gemini) within the Builder flow.

• Choose from the Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash-Lite, Gemini 3 Flash, and Gemini 3 Pro models.

• Knowledge base: upload documents (PDF, TXT, CSV, Markdown, HTML, XML, RTF, JavaScript, Python, and CSS) for Gemini to answer based on them.

• Assistant instructions to define the role, scope, and tone of responses.

• Control over temperature and the token limit per call.

• Security configuration by category (harassment, hate speech, sexually explicit content, and dangerous material), with four blocking levels.

• Store the response in a variable for use in the rest of the flow.

Supported on the following channels: all channels available on the Blip platform.

Prerequisites

• Access to the bot in Blip with permission to install extensions.

• A Google account with access to Google AI Studio.

• A valid Gemini API key.

Installation and setup

Install the extension on the bot

Install the extension on the bot that contains the flow in which Gemini will generate responses (the service bot), not on the Router.

1. Access the desired bot in Blip.

2. Open the Blip Store and find the Google Gemini extension, or search for Google Gemini in the store search.

3. Select the extension and click Install.

4. Choose the bot where the extension will be used.

After installing, open the extension page to start the setup. In the bot menu, click the three dots and find the Google Gemini extension.

Create an API key in Google AI Studio

The extension uses the Gemini API, so you need a Google API key. On the extension page, below the API key field, the link under Available here opens the Google AI Studio key page.

1. Access Google AI Studio.

2. Open the API keys area and click Create API Key.

3. Select or import a Google Cloud project and confirm. The key is generated.

4. Copy the key and store it in a safe place.

If you use a Google Workspace account (business account), you may need to ask your organization's administrator for permission to access Google AI Studio and create the key. Google maintains a specific guide for enabling Gemini in the enterprise.

Warning: do not share the key in documents, screenshots, tickets, or public channels. If a key is exposed, generate a new one in Google AI Studio and replace the old one.

Configure the model and security

Return to Blip and open the Google Gemini extension on the AI settings tab.

1. Under Model, select the Gemini model that the assistant will use. Each model has its own costs and limits, defined by your Google account plan.

2. Paste the key generated in Google AI Studio into the API key field.

3. In the Security section, adjust the blocking level for each category. Higher levels block more responses, including some harmless ones. Keep the default values if there is no specific rule for your operation.

4. Click Save. The extension tests the key before saving. If the key is invalid, it displays an error message instead of completing the process.

Add a knowledge base

The knowledge base provides specific information for Gemini to respond accurately. Open the Knowledge base tab.

1. Under Documents, drag or select the file you want to upload. The accepted formats are PDF, TXT, CSV, Markdown, HTML, XML, RTF, JavaScript, Python, and CSS.

2. Click Upload new file. The file will appear in the Files list.

3. To remove a document, click the trash can button next to it.

Use clear, up-to-date documents. Avoid duplicate versions or contradictory information because this confuses the responses. The knowledge base complements the assistant's instructions but does not replace a clear definition of its behavior.

Using Gemini in the Builder flow

Install the Builder extension in Chrome

To add Gemini to the flow, you need the Builder extension in Google Chrome. Open the Integrating with Builder tab on the extension page and click Install extension to complete the installation in the Chrome Web Store.

Add the integration block to the flow

1. In Builder, open the point in the flow where Gemini should generate the response.

2. Click the add block icon. With the Chrome extension installed, the option to add an integration block appears.

3. Open the integration tab and select the Google Gemini integration.

Configure the block fields:

• Input variable: the flow variable containing the user's message. The default is input.content.

• Instructions: the assistant's behavior (see the next section).

• Temperature: a value from 0 to 1 that controls the creativity of the response.

• Token limit: the maximum number of tokens in the response, from 1 to 8192 (1 token is equivalent to approximately 3 letters). The default is low, so increase this value if the responses need to be longer.

• Result variable (under Save response): the variable where the Gemini response will be stored.

The model and security settings come from the AI settings tab on the extension page, not from the block.

Define the assistant's instructions

The instructions tell Gemini how to behave. Clearly describe the assistant's role, which topics it can answer about, which ones it should refuse, what tone to use, and what to do when it cannot find the information. Avoid generic instructions such as "answer everything" and instruct the model not to make up data. If the Gemini API returns an error, the extension returns a standard response ("I don't know how to talk about that subject.") and logs the error.

Save the response in a variable

Define a name in the block's Result variable, for example geminiResponse. After running the integration, use this variable in a content block:

{{geminiResponse}}

Make sure that the name used in the content block is exactly the same as the one defined in the integration.

Test the flow

Before publishing, run the bot in test mode and validate its behavior:

1. Send a question within the scope of the instructions and confirm that the response was stored in the variable.

2. Send questions outside the scope and check whether the assistant refuses as you defined.

3. Ask questions covered by the knowledge base files and check the responses.

4. Test the case where the information does not exist.

Usage examples

A clear instruction defines the assistant. For example:

You are a virtual assistant responsible for answering questions about the company's products and services. Respond only based on the information in the knowledge base. Use clear, polite, and concise language. If you cannot find the answer, state that you do not have enough data and direct the user to speak with an agent. Do not make up information, prices, timelines, or rules.

Based on this foundation, you can create a product Q&A assistant, an assistant for consulting internal manuals, or a customer service support tool that responds before involving a human agent.

Warning: usage limits and charges depend on the model and the plan subscribed to on your Google account, and may be measured by requests and tokens, per minute, or per day. Experimental models typically have lower limits and are less stable. Validate costs and limits before using the integration in production, and avoid sending personal or sensitive data without prior assessment.

Support

If you have a question, problem, or suggestion, contact us by email at support@wiv.com.br.

Need more help? Explore our content on Blip Academy or Blip Community, watch tutorials on our YouTube channel or get your questions answered through our support channel 😃