SiteMate Pro: The AI OS for African Construction

This is a submission for the Algolia Agent Studio Challenge: Consumer-Facing Conversational Experiences

What I Built

SiteMate Pro is an "AI Operating System" for the African construction industry. It solves a critical problem: Opacity.

In Nigeria, building material prices change daily, data is fragmented across WhatsApp chats, and fraud is rampant. Engineers often waste days calculating costs manually, and projects fail because of inaccurate estimates.

SiteMate Pro is a conversational agent that acts as a Super-Intelligent Quantity Surveyor.

• Conversational Planning: You can tell it, "Build a 4-bedroom duplex on swampy soil in Lekki."
• Instant Estimation: It doesn't just "guess" prices like standard LLMs. It retrieves real-time market data to generate a precise, location-specific Bill of Quantities (BOQ).
• Fraud Detection: It filters supplier bids in a marketplace, flagging prices that are suspiciously high or low compared to the indexed market rate.

Demo

try the Live App: Launch SiteMate Pro

🎥 Video Walkthrough (7 Mins):

Source Code:
https://github.com/AdMub/Sitemate-Construction-Agent

How I Used Algolia Agent Studio

I leveraged Algolia Agent Studio to solve the biggest problem with Generative AI in engineering: Hallucination.

A standard LLM does not know the price of a bag of cement in Lagos today. If asked, it will hallucinate a number, which is dangerous for a construction budget.

My Architecture (RAG Pipeline):

1. Index: I modeled construction material data (Cement, Granite, Iron Rods) and supplier profiles as records in Algolia.
2. Retrieval: When a user prompts the agent (e.g., "Estimate a foundation for swampy soil"), the system first queries Algolia to fetch the current unit prices for that specific location (e.g., Lekki vs. Ibadan).
3. Targeted Prompting: I inject this retrieved data into the Gemini prompt context. The system instruction is strictly engineered: "You are a Senior Engineer. Use ONLY the retrieved Algolia prices to calculate the total. Do not use outside training data for costs."

This turns the AI from a creative writer into a grounded calculator, ensuring every Naira quoted is backed by real data.

Why Fast Retrieval Matters

In construction, speed is trust.

• Contextual Accuracy: By retrieving data based on specific facets (Location: "Lekki", Soil: "Swampy"), the agent avoids giving generic advice that causes building collapse.
• Fraud Prevention: Fast retrieval allows us to cross-check incoming supplier bids against thousands of historical data points in milliseconds. If a supplier quotes 30% above the indexed average, the user is warned instantly.
• User Experience: The conversation feels natural because the latency is near-zero. The user asks a complex question, and the Bill of Quantities appears instantly, creating a "magical" experience that replaces days of spreadsheet work.