🚀 Build with AI: Brand Monitoring Challenge

Welcome to the Brand Monitoring Challenge for our Build with AI workshop! In this exercise you’ll use the ADK (and optionally MCP) to build an end-to-end agent that:

1. Ingests mentions of a brand from multiple sources (Reddit, Twitter, News APIs, general web).
2. Analyzes what people are saying—sentiment, key topics, issue detection.
3. Generates a consolidated report highlighting the brand’s current public perception and areas of concern.

---

🛠 Workshop Tools & Prerequisites

• Agent Development Kit (ADK) Tool for defining your agent, tools, and function calls.

• MCP (Model Context Protocol) (optional) You’ll be given MCP credentials to connect. If your quota runs out, feel free to roll your own connectors or use plain function-calling.

• Gemini The LLM you’ll use for analysis and report generation.

• Dependencies Install all required packages from the provided requirements.txt:

  pip install -r requirements.txt
  

---

🚦 Challenge Statement

Given the name of a brand (user input), Build an AI agent that:

1. Queries at least three data sources (e.g., Reddit, Twitter, News API, Web Scraper).
2. Performs sentiment analysis, topic extraction, and issue detection on the collected mentions.
3. Outputs a structured Brand Monitoring Report (see “Expected Output” below).

---

📥 Inputs

• company_name (string) e.g. "Acme Co."

---

📈 Expected Output

Your agent should return a or Markdown report containing:

1. Executive Summary A 2–3 sentence overview of overall sentiment and top concerns.

2. Sentiment Breakdown Percentage of positive / neutral / negative mentions, by source.

3. Top 5 Topics & Issues Most frequently discussed themes (e.g., “shipping delays,” “customer support”).

4. Trend Analysis Simple time-series summary (mentions per day), highlighting peaks.

5. Sample Mentions 2–3 representative quotes (with source name and link) for each sentiment category.

6. Recommendations Based on detected issues, suggest 2–3 actionable next steps.

---

🏗 Your Tasks

1. Configure Connectors

   • Use MCP with provided credentials, or implement your own via function calls/web scraping.

2. Define Agent & Tools in ADK

   • Declare functions for data ingestion, analysis, and report generation.

3. Implement Analysis Pipeline

   • Ingest raw mentions → clean/filter/summarize → generate report.

4. Generate Report

   • Get the results into the Markdown structure above.

5. Run Your Solution

   • Run your solution with two different brand names.

---

📝 Submission (Suggested)

• GitHub Repo containing:

  • README.md
  • Source code for your agent
  • requirements.txt
  • Example outputs for at least two brands

• Evaluation Criteria (for your reference):

  1. Data Coverage: ≥3 sources ingested.
  2. Report Quality: Clarity, completeness, structure.
  3. Tooling: Correct use of ADK functions (and MCP if used).
  4. Code Quality: Readable, modular, documented.

---

Setup

Setup Instructions

• Gemini API key Access google: http://aistudio.google.com/

  Login with your google account.

  Create an api key.

  Create a file named .env.

  Add the variable GOOGLE_API_KEY with the value of the api key from ai studio

ADK

Open UI:

adk web

Run the agent as a service:

adk api_server

Local MCP

to run the local MCP, make sure you have the correct credentials in your .env file

cd 06_challenge/mcp/
uvicorn src.api:app --host 0.0.0.0 --port 8001 --reload

Credentials Websites

🛠️ Reddit Credentials

Website: reddit.com/prefs/apps What you’ll need:

• Client ID (aka “App ID”)
• Client Secret
• User Agent (e.g. my-app/0.1 by your_username)

---

🐦 Twitter Credentials

Website: developer.twitter.com/en/portal/dashboard What you’ll need:

• Bearer Token

---

🗞️ News API (Tavily)

Website: tavily.com What you’ll need:

• API Key

---

Just head to each link, follow their “Create new app” or “Get API key” flows, and then stash these values safely—your code will thank you! 🎉

Good luck, and happy building! 🌟