Skip to content

Commit 1885d55

Browse files
tarunchytarunchy
authored
Create flameguardai.mdx
1 parent 36b663d commit 1885d55

File tree

1 file changed

+163
-0
lines changed

1 file changed

+163
-0
lines changed

‎docs/showcase/flameguardai.mdx

Lines changed: 163 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,163 @@
1+
## 🔥 Inspiration
2+
3+
The devastating LA wildfires were more than a tragic event — they were a turning point. While watching homes burn to the ground, I saw a story about one house that miraculously survived. That story stuck with me.
4+
5+
What protected that home? It wasn’t luck. It was planning — fire-resistant materials, defensible space, and design choices made long before disaster struck. This realization — combined with the rising difficulty of securing home insurance — inspired us to build **FlameGuard AI™**.
6+
7+
We wanted to create a tool that empowers **homeowners** to take fire risk prevention into their own hands — not just react after it’s too late.
8+
9+
## 🧠 What it does
10+
11+
**FlameGuard AI™** helps homeowners, buyers, and property professionals detect and act on **external fire vulnerabilities** like wildfires or neighboring structure fires. It’s more than a scan — it’s a personalized research assistant for your home.
12+
13+
### Demo
14+
15+
[![Watch Live video](https://img.youtube.com/vi/EI5yT7_aD6U/0.jpg)](https://www.youtube.com/watch?v=EI5yT7_aD6U)
16+
17+
### Try it out
18+
19+
- [https://flameguardai.dlyog.com](https://flameguardai.dlyog.com)
20+
- [https://flameguardai-mcp.dlyog.com](https://flameguardai-mcp.dlyog.com)
21+
22+
23+
### Key Features:
24+
- 📸 Upload a home photo
25+
- 👁️ Analyze visible fire risks via **OpenAI Vision API**
26+
- 📚 Trigger deep research using the **Perplexity Sonar API**
27+
- 📄 Get a detailed, AI-generated report with:
28+
- Risk summary
29+
- Prevention strategies
30+
- Regional best practices
31+
- 🛠️ Optional contractor referrals for mitigation
32+
- 💬 Claude (MCP) chatbot integration for conversational analysis
33+
- 🧾 GDPR-compliant data controls
34+
35+
Whether you’re protecting your home, buying a new one, or just want peace of mind — **FlameGuard AI™ turns a photo into a plan**.
36+
37+
38+
## ⚙️ How we built it
39+
40+
FlameGuard AI™ is powered by a modern GenAI stack and built to scale.
41+
42+
- **Frontend**: Lightweight HTML dashboard with user account control, photo upload, and report access
43+
- **Backend**: Python (Flask) with RESTful APIs
44+
- **Database**: PostgreSQL (local) with **Azure SQL-ready** schema
45+
- **Containerization**: Dockerized for flexible local or cloud deployment
46+
- **Cloud-ready**: Built for **Azure App Service** with dedicated database support
47+
48+
### 🔍 Deep Research with Perplexity Sonar API
49+
50+
The real innovation is how we use the **Perplexity Sonar API**:
51+
52+
1. After analyzing the uploaded image, we identify specific vulnerabilities (e.g., flammable roof, dry brush nearby).
53+
2. For each risk, we generate a **custom research plan** — covering severity, mitigation strategies, and localized insights.
54+
3. We run **iterative agentic-style calls** to Perplexity Sonar — treating it like a research assistant gathering the best available information.
55+
4. The research is **aggregated, organized, and formatted** into an actionable HTML report — complete with citations, links, and visual guidance.
56+
57+
This kind of **structured, trustworthy, AI-powered research would not be possible without Perplexity**.
58+
59+
60+
## 🚧 Challenges we ran into
61+
62+
- 🔁 Designing **multi-step research prompts** for Perplexity Sonar that maintain context across multiple fire vulnerabilities
63+
- 🏠 Building **robust image analysis prompts** that generalize across diverse home types, lighting, and angles
64+
- ⏳ Coordinating **asynchronous flows** between user input, vision analysis, deep research, and report generation
65+
- ❌ One major challenge: we attempted to **automatically retrieve contractor email addresses** via Perplexity API to pre-fill outreach requests — with user consent, we wanted to connect homeowners directly to local professionals and even **submit quote requests** on their behalf.
66+
- While the idea worked conceptually, the email results from Perplexity were not reliably accurate or complete.
67+
- For now, we've **disabled this feature** in the UI, but it's a high-priority enhancement for our roadmap.
68+
- Future plan: we'll integrate a **dedicated MCP server for contractors**, allowing us to directly submit quote requests from homeowners in a structured, secure workflow.
69+
- ⚖️ Balancing speed and depth: AI-generated insights must feel fast, but research quality takes time — designing a UX that handles that tension gracefully was not easy.
70+
71+
72+
## 🏆 Accomplishments that we're proud of
73+
74+
- Successfully used **OpenAI Vision + Perplexity Sonar API** together in a meaningful, real-world workflow
75+
- Built a functioning **MCP server** that integrates seamlessly with Claude for desktop users
76+
- Created a product that is **genuinely useful for homeowners today** — not just a demo
77+
- Kept the experience simple, affordable, and scalable from the ground up
78+
- Made structured deep research feel accessible and trustworthy
79+
80+
81+
## 📚 What we learned
82+
83+
- The **Perplexity Sonar API** is incredibly powerful when used agentically — not just for answers, but for reasoning.
84+
- Combining **multimodal AI (image + research)** opens up powerful decision-support tools.
85+
- Users want **actionable insights**, not just data — pairing research with guidance makes all the difference.
86+
- Trust and clarity are key: our design had to communicate complex information simply and helpfully.
87+
88+
89+
## 💳 Admin Credit System
90+
91+
To help manage resource usage and scale responsibly, we built a **credit-based system** for users. This allows us to give out free trial credits, support premium usage tiers, and track consumption transparently.
92+
93+
### What the Credit System Does:
94+
- 🧾 **Tracks credits** per user (free and purchased)
95+
- ➕ Admins can **grant credits**
96+
- ➖ Admins can **revoke credits** safely (no negative balances)
97+
- 🔄 **Usage is logged** with full transaction history
98+
- 👁️ Credits update in real-time within the admin panel
99+
100+
### Admin Panel Features:
101+
- View current user balances
102+
- Grant or subtract credits via secure modals
103+
- Automatically updates the user interface after each action
104+
- View recent credit transaction history
105+
106+
### Credit Transaction Types:
107+
- `admin_grant`: credits added by admin
108+
- `usage`: credits used by the user for assessments
109+
- `admin_revoke`: manual deduction by admins
110+
- `initial_free`: signup bonus for new users
111+
112+
### Backend Highlights:
113+
- 🔐 Safe SQL logic with balance checks
114+
- 📦 Modular logic in `credit_utils.py`
115+
- 🧠 Clean REST APIs for credit updates
116+
- 📝 Full logging of every transaction
117+
118+
We designed this credit system to be **simple for admins**, **transparent for users**, and **ready for future billing integration**.
119+
120+
---
121+
### 📊 Market Opportunity (U.S.)
122+
123+
- 🏠 **Home Services & Improvement Market**: ~$740B TAM
124+
Covers home maintenance, upgrades, and safety systems
125+
Sources: Verified Market Research, GlobeNewswire
126+
127+
- 🔥 **Wildfire & Fire Risk Mitigation**: $8.6B → $19.9B
128+
Rising demand due to climate change & insurance premiums
129+
Source: Verified Market Research
130+
131+
- 🧾 **Home Inspection Market**: ~$4B annually
132+
Home inspections are required for ~90% of real estate sales
133+
Source: IBISWorld
134+
135+
- 💸 **Estimated SAM**: $100–$150B
136+
Focused on wildfire-prone homeowners, real estate firms, insurers, and contractors
137+
(California, Colorado, Oregon, etc.)
138+
139+
- 🚀 Huge gap in end-to-end solutions that go beyond assessment to action
140+
FlameGuard AI fills this gap by identifying risks **and** connecting to local pros
141+
142+
143+
## 🚀 What's next for FlameGuard AI™ - Prevention is Better Than Cure
144+
145+
We're just getting started.
146+
147+
### Next Steps:
148+
- 🌐 Deploy to **Azure App Services** with production-ready database
149+
- 📱 Launch mobile version with location-based scanning
150+
- 🏡 Partner with **home inspection services** and **homeowners associations**
151+
- 💬 Enhance Claude/MCP integration with voice-activated AI reporting
152+
- 💸 Introduce B2B plans for real estate firms and home safety consultants
153+
- 🛡️ Expand database of **local contractor networks** and regional fire codes
154+
155+
156+
We’re proud to stand with homeowners — not just to raise awareness, but to enable action.
157+
158+
**FlameGuard AI™ – Because some homes survive when others don’t.**
159+
160+
### 🙏 Thank You
161+
162+
Big thanks to **DevPost** and **Perplexity** for making this hackathon possible. Without the **Perplexity Sonar API**, this level of structured, trusted, AI-driven research would simply not be possible.
163+

0 commit comments

Comments
 (0)