Inspiration
Passion doesn't save businesses — intelligence does. 90% of startups collapse within their first two years, and the pattern is almost always the same: founders pour their life savings into a market they never rigorously analyzed. We asked a simple question — what if every entrepreneur could access a brutally honest, data-backed survival assessment before committing a single dollar? That question became Vettura.
What it does
Vettura is a real-time business intelligence platform that answers the question every founder is afraid to ask: will my business actually survive here? Provide your business type, target location, and starting budget. Within seconds, Vettura delivers:
A Survival Probability Score (0–100) calibrated against live competitor density and market conditions An interactive competitor heatmap built on OpenStreetMap, showing exactly who you're up against and where they operate
5-year financial projections spanning revenue trajectories, expense curves, and profit forecasts — grounded in local economic reality, not wishful thinking
A narrative market briefing structured in chapters — written as though a seasoned mentor is walking you through the battlefield
Risk factor analysis paired with concrete, actionable mitigation strategies
AI voice narration powered by ElevenLabs — your entire report delivered aloud in a natural human voice, turning a data dump into an immersive strategic experience
How we built it
- Frontend: React + Vite for rapid rendering, Recharts for financial visualizations, Leaflet for dynamic competitor mapping
- Backend: A zero-framework pure Python HTTP server handling /api/analyze and /api/narrate endpoints
- AI Analysis: OpenRouter API (GPT-4o-mini) for market intelligence, financial modeling, and competitive kill strategies
- Competitor Data: OpenStreetMap + Overpass API for real-time local business extraction — entirely free, entirely global
- Voice: ElevenLabs Text-to-Speech API (eleven_turbo_v2) for narrating full reports, individual chapters, and risk breakdowns
The core scoring formula weighs five critical dimensions: Survival Score=0.30C+0.25M+0.25F+0.15L+0.05D\text{Survival Score} = 0.30C + 0.25M + 0.25F + 0.15L + 0.05DSurvival Score=0.30C+0.25M+0.25F+0.15L+0.05D Where CC C = Competitive Intensity, MM M = Market Opportunity, FF F = Financial Viability, LL L = Location Quality, and DD D = Differentiation Potential.
Challenges we ran into
Zero-budget data pipeline: Building a complete competitor intelligence system on free OpenStreetMap and Overpass API data — instead of Google Places — meant engineering custom tag mappings across 18+ business categories and handling wildly inconsistent geographic coverage
Audio streaming through a proxy chain: Getting ElevenLabs audio/mpeg byte streams to flow correctly from the Python backend through the Vite dev proxy to the browser was a multi-layered debugging battle — CORS, content types, and chunked transfer all broke at different stages
Framework-free backend: Serving CORS preflight responses, binary audio streams, and structured JSON from a raw BaseHTTPRequestHandler simultaneously — with no Flask, no FastAPI, nothing — was an exercise in controlled suffering
Grounding the LLM in reality: Language models love to hallucinate optimistic financials. Forcing GPT-4o-mini to produce locally calibrated, defensible projections required injecting real competitor context directly into the prompt and enforcing strict output schemas
Accomplishments that we're proud of
- Engineered a survival scoring engine built on a weighted multi-factor model fed by real-time data — not static assumptions
- Delivered end-to-end voice narration via ElevenLabs, transforming a static business report into something you can listen to while driving, walking, or thinking — an experience no competing tool offers today
- Achieved zero reliance on paid data sources — every API in the stack is free, making Vettura accessible to bootstrapped founders anywhere on earth
- Created a storytelling layer that converts raw market data into a seven-chapter narrative arc — making the difference between a forgettable spreadsheet and a report founders actually read and remember
What we learned
- Real-world competitor data is inherently noisy — OpenStreetMap coverage ranges from excellent to nearly empty depending on geography, demanding robust fallback strategies
- ElevenLabs eleven_turbo_v2 at stability 0.5 and style 0.3 hits a sweet spot — authoritative enough for business narration, human enough to hold attention
- Structured JSON generation from LLMs requires aggressive schema enforcement — even capable models drift toward vague or inflated outputs without hard constraints
- The distance between "works on localhost" and "works end-to-end through a proxy chain" is precisely the graveyard where most hackathon projects are buried
What's next for Vettura
- Voice Q&A: Let users speak questions ("What's my biggest risk?") and get spoken answers back using ElevenLabs + Whisper — a full voice assistant loop
- Real foot traffic data via city open data APIs
- Competitor kill strategy audio — ElevenLabs narrates your personalized domination plan like a strategic briefing
- PDF export of the full report with charts
- Multi-city comparison — analyze the same business across 3 cities simultaneously and hear the comparison spoken aloud
Built With
- axios
- claude
- css
- elevenlabs
- elevenlabs-text-to-speech-api
- gpt-4o-mini
- javascript
- leaflet.js
- nominatim
- openrouter-api
- openstreetmap
- overpass-api
- python
- react
- react-router
- recharts
- vite
Log in or sign up for Devpost to join the conversation.