Bot Bender: Google Workspace Group Management Agent

This is Bot Bender, a resilient chat tool designed to handle daily Google Workspace Group administration tasks (like managing members and roles) and act as a technical assistant by providing instant answers to FAQs.

The project combines Google Cloud's power: Gemini AI for understanding natural language and the precise execution capabilities of the Admin SDK.

---

Core Capabilities

This bot is engineered for secure, efficient management across your domain:

• Conversational Administration: Users can simply type what they need (e.g., "add user@ to group.test@"). The Gemini AI interprets the intent behind the text.
• Security First: The sensitive Service Account keys are loaded securely at runtime via Google Secret Manager (GSM), ensuring they are never exposed in code or variables.
• Flexible Resolution: The bot can find users and groups by accepting full emails, partial emails, or simply their display names (e.g., "Juan Pérez").
• Quick Support (FAQs): It answers support questions using a two-part strategy: first, a fast, direct lookup against the knowledge base, and then a semantic fallback with Gemini for complex queries.
• Permission Control: It enforces organizational rules, allowing group management only to approved Managers, Owners, or the Global Administrator.

---

📷 Bot Interface Preview

Here is a look at the command menu and the interaction workflow within Google Chat.

Quick Command Menu

Command Dropdown Interaction

---

Technology Stack

The bot is designed on a modern, cloud-native architecture ready for serverless deployment:

Technology	Role in the Project
Node.js / Express	Provides the execution environment and manages receiving chat events.
Google Admin SDK	The core library for performing the actual user and group management in Workspace.
Google Secret Manager	Secure handling of the Admin key used for domain delegation.
Google Firestore	Dynamically stores the AI prompts and the structured FAQ database.
Gemini API	The AI engine for natural language understanding and semantic search.

---

💾 Firestore Configuration

FAQ Data Structure

The bot relies on a specific structure where the content is stored as a JSON string within a document.

• Collection: faq

Configuration Process

1. Write your FAQ content following the JSON array format below.
2. In the Firestore console, create a collection named faq.
3. Create a new document and add the field faq_documentation. Paste the complete JSON content as a String value.

[
  {
    "id": 1,
    "category": "Access and Passwords",
    "questions": ["How can I reset my password?", "I forgot my key."],
    "keywords": ["reset", "password", "key"],
    "standard_answer": "To reset your password, you must go to the self-service portal...",
    "detailed_steps": ["Visit the company's web portal.", "Click on 'Forgot my password'."]
  }
]