VodarAI

Landing Page
Pipeline Visualization
Sentimental Analysis of Community Voices
Users Quotes
References

Inspiration

As AI tool learners and builders, we kept hitting the same wall: the feedback that actually matters is scattered everywhere: Reddit threads, Product Hunt comments, Hacker News discussions, Dev.to posts, Hashnode articles. Reading through it all manually is slow, noisy, and honestly pretty biased toward whatever you stumble across first. VodarAI started as our own fix for this. We wanted something that could turn all that fragmented community chatter into clear, actionable product signals, without the hours of tab-switching.

What it does

You give VodarAI a query - say, "Cursor" - and it goes out and pulls relevant discussions from across platforms automatically. From there, it synthesizes everything into:

a plain-English overview of what people are actually saying
sentiment and controversy breakdown
recurring pain points and themes
notable quotes worth reading
a clean, structured markdown report You also get full source transparency through card and list views, with export options for Notion and Markdown so it fits into however your team already works.

How we built it

Frontend: Next.js + TypeScript + Tailwind CSS
Backend: FastAPI + Pydantic + asyncio
Data ingestion: Native API fetchers and Browser Use agents
Analysis & generation: Gemini API
Output layer: SSE-based live pipeline progress, Notion export, Markdown preview and copy

Challenges we ran into

Bot protections and Cloudflare made browser-based scraping unreliable on some platforms.
Cross-platform normalization turned out to be genuinely hard. Every source has its own schema, its own quirks, its own signal-to-noise ratio.
LLM robustness. early on, report generation would silently fail on serialization edge cases or model fallbacks.

Accomplishments that we're proud of

We shipped an end-to-end product people can actually use: type any AI tool name, collect live community signals from multiple sources, and get a structured, export-ready report in minutes.

What we learned

We learned that reliability is the real feature. Browser automation is much more fragile than it looks, and small prompt/timing changes can break extraction quickly.