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

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > Product Review Analysis Agent

Product Review Analysis Agent

Author: Venkata Sudhakar

ShopMax India receives hundreds of customer reviews each day across its electronics catalogue. Reading every review manually is not practical. A review analysis agent that extracts sentiment, clusters recurring complaints, and surfaces actionable insights helps the product and merchandising teams act faster on customer feedback.

This tutorial builds a Product Review Analysis Agent using ADK and Gemini. The agent analyses a batch of reviews for any product, computes a sentiment breakdown, extracts top themes, and produces a structured insight report that the team can act on immediately.

The below example shows the review pipeline applied to ShopMax India product feedback.

from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService

# --- Sample review database (simulates Firestore or BigQuery) ---
REVIEWS = {
    "SKU-1001": [
        {"id": "R001", "rating": 2, "text": "Picture quality is great but the remote stopped working after 2 weeks. Very disappointed."},
        {"id": "R002", "rating": 5, "text": "Excellent TV! Setup was easy and colours are stunning. Worth every rupee."},
        {"id": "R003", "rating": 3, "text": "Good picture but delivery took 10 days. Remote is cheap quality."},
        {"id": "R004", "rating": 1, "text": "Dead pixels on arrival. ShopMax refused exchange. Worst experience ever."},
        {"id": "R005", "rating": 4, "text": "Great value. Sound could be better but picture is superb for the price."},
        {"id": "R006", "rating": 2, "text": "Remote broke in month 1. Customer support was unhelpful. Not buying again."},
    ],
    "SKU-1002": [
        {"id": "R007", "rating": 5, "text": "Best headphones I have owned. Noise cancellation is outstanding."},
        {"id": "R008", "rating": 4, "text": "Sound is excellent. A bit heavy for long sessions but overall great."},
        {"id": "R009", "rating": 5, "text": "Battery lasts all day. Connecting to phone is instant. Highly recommended."},
        {"id": "R010", "rating": 3, "text": "Good quality but slightly overpriced compared to Amazon."},
    ],
}

# --- Tool: fetch reviews for analysis ---
def get_reviews(sku: str) -> dict:
    """Fetch all customer reviews for a product SKU.
    Args:
        sku: Product SKU code
    Returns:
        List of reviews with rating and text
    """
    if sku not in REVIEWS:
        return {"error": f"No reviews found for SKU: {sku}"}
    reviews = REVIEWS[sku]
    avg_rating = round(sum(r["rating"] for r in reviews) / len(reviews), 1)
    return {
        "sku": sku,
        "total_reviews": len(reviews),
        "average_rating": avg_rating,
        "reviews": reviews,
    }

# --- Tool: compute sentiment breakdown ---
def compute_sentiment(sku: str) -> dict:
    """Compute positive, neutral, and negative sentiment counts from ratings.
    Args:
        sku: Product SKU code
    Returns:
        Sentiment distribution and alert flag
    """
    if sku not in REVIEWS:
        return {"error": f"No reviews for SKU: {sku}"}
    reviews = REVIEWS[sku]
    positive = sum(1 for r in reviews if r["rating"] >= 4)
    neutral = sum(1 for r in reviews if r["rating"] == 3)
    negative = sum(1 for r in reviews if r["rating"] <= 2)
    total = len(reviews)
    neg_pct = round(negative / total * 100)
    return {
        "positive": positive,
        "neutral": neutral,
        "negative": negative,
        "negative_percent": neg_pct,
        "alert": "HIGH_NEGATIVE" if neg_pct >= 30 else "OK",
    }

# --- Agent setup ---
review_tool = FunctionTool(func=get_reviews)
sentiment_tool = FunctionTool(func=compute_sentiment)

agent = LlmAgent(
    name="review_analysis_agent",
    model="gemini-2.0-flash",
    instruction="""You are a product review analyst for ShopMax India.
When asked to analyse a product:
1. Call get_reviews to fetch the full review set.
2. Call compute_sentiment for the sentiment breakdown.
3. Read through the review texts and identify the top 3 recurring themes (e.g. remote quality, delivery time, picture quality).
4. Produce a structured report with: overall rating, sentiment breakdown, top themes, and 2-3 recommended actions.
Keep the tone professional and focused on what the product team should do next.""",
    tools=[review_tool, sentiment_tool],
)

session_service = InMemorySessionService()
runner = Runner(agent=agent, app_name="review_agent", session_service=session_service)

import asyncio

async def analyse(query: str) -> str:
    session = await session_service.create_session(app_name="review_agent", user_id="product_team")
    from google.genai.types import Content, Part
    msg = Content(role="user", parts=[Part(text=query)])
    result = ""
    async for event in runner.run_async(user_id="product_team", session_id=session.id, new_message=msg):
        if event.is_final_response() and event.content:
            result = event.content.parts[0].text
    return result

response = asyncio.run(analyse("Analyse reviews for SKU-1001 and tell me what we need to fix urgently."))
print(response)

It gives the following output,

Product Review Report - SKU-1001 (Samsung 55-inch 4K TV)

Overall Rating: 2.8 / 5  (6 reviews)

Sentiment Breakdown:
  Positive (4-5 stars): 2  (33%)
  Neutral  (3 stars):   1  (17%)
  Negative (1-2 stars): 3  (50%)  ** HIGH_NEGATIVE ALERT **

Top Recurring Themes:
  1. Remote control quality  - mentioned in 3 of 6 reviews (defective/cheap)
  2. Dead pixels / hardware defects  - 1 DOA unit reported
  3. Delivery delay  - 10-day delivery mentioned as concern

Recommended Actions:
  1. Escalate remote control defect to Samsung supplier - 50% complaint rate
  2. Review DOA exchange policy - current refusal is damaging brand trust
  3. Audit logistics partner for SKU-1001 - target delivery under 5 days

Extend this agent by connecting it to a real Firestore reviews collection and scheduling daily reports via Cloud Scheduler. Add a Looker Studio dashboard to visualise sentiment trends across the entire ShopMax India catalogue over time.

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