	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Gemini API for FAQ Generation from Documents

Gemini API for FAQ Generation from Documents

Author: Venkata Sudhakar

FAQ generation is a high-leverage automation: a well-written FAQ page can deflect 30-40% of support tickets before they are ever created. The Gemini API reads source documents - product manuals, return policies, shipping guides, warranty terms - and generates natural-sounding question-and-answer pairs that match how customers actually ask questions, not how documents are written.

ShopMax India generates FAQs automatically for every product category and policy document. When a new return policy was introduced, Gemini generated 24 FAQ entries in under 30 seconds from the policy document. The support team reviewed and published them within an hour. Return-related support tickets dropped 31% in the following two weeks as customers found answers in the FAQ before contacting support.

The below example shows how to generate a product FAQ from a manual or specification document using Gemini.

import vertexai
from vertexai.generative_models import GenerativeModel
import json

vertexai.init(project="shopmax-india", location="asia-south1")
model = GenerativeModel("gemini-2.0-flash")

def generate_faq(source_document: str, topic: str,
                  num_questions: int = 10) -> list:
    prompt = f"""Generate a FAQ for ShopMax India customers based on this {topic} document.
Create exactly {num_questions} question-answer pairs.
Return ONLY a JSON array where each item has:
- question: the customer question (natural language, as a customer would ask it)
- answer: clear, helpful answer in 2-4 sentences
- category: one of [shipping, returns, product, payment, warranty, general]

Focus on questions customers actually ask - practical, specific, actionable.

Document:
{source_document[:10000]}

JSON array:
"""

response = model.generate_content(
        prompt,
        generation_config={"temperature": 0.4, "response_mime_type": "application/json"}
    )
    return json.loads(response.text)

returns_policy = """
ShopMax India Returns Policy (Effective April 2024)

Return Window: Customers may return products within 10 days of delivery.
Electronics must be in original packaging with all accessories.
Defective items qualify for replacement or refund within 30 days.

Non-returnable items: Software, personal care items, customised products.
Opened software licenses cannot be returned under any circumstances.

Refund Timeline: Refunds to original payment method within 5-7 business days.
UPI and wallet refunds process within 24 hours.
Credit/debit card refunds take 5-7 business days depending on your bank.

Pickup: Free pickup from registered delivery address.
Customers must package items securely before pickup agent arrival.
"""

faqs = generate_faq(returns_policy, "returns policy", num_questions=5)
for faq in faqs:
    print(f"Q: {faq['question']}")
    print(f"A: {faq['answer']}")
    print(f"Category: {faq['category']}")
    print()

It gives the following output,

Q: How many days do I have to return a product?
A: You can return most products within 10 days of delivery. Electronics must
   be in original packaging with all accessories included. Defective items
   qualify for replacement or refund within 30 days.
Category: returns

Q: How long does a refund take to reach my account?
A: Refunds to UPI or wallet are processed within 24 hours. Credit and debit
   card refunds take 5-7 business days depending on your bank. All refunds
   are credited to your original payment method.
Category: payment

The below example shows how to generate FAQs for multiple product categories and organise them into a structured knowledge base.

It gives the following output,

Generated 5 FAQs for: Returns Policy
Generated 5 FAQs for: Shipping Guide

Total FAQ entries generated: 10

Returns Policy: 5 FAQs, categories: ['returns', 'payment', 'general']
Shipping Guide: 5 FAQs, categories: ['shipping', 'general']

The question phrasing is intentionally natural - Gemini generates questions the way customers actually type them into search bars, not the formal language used in policy documents. This dramatically improves search relevance when the FAQ knowledge base is searched. ShopMax India also feeds these FAQ pairs directly into its RAG pipeline, meaning the customer service agent can retrieve and cite specific FAQ answers when customers ask matching questions rather than generating new responses from scratch each time.

Send your comments, suggestions or queries regarding this site to [email protected].