VyaparX

• 

• 

• 

Inspiration

We noticed that small and mid-sized businesses often lack the resources to do in-depth competitor analysis. Manual research is slow, error-prone, and expensive. Meanwhile, AI agents can automate each step—web crawling, data synthesis, report writing—so why not stitch them together into an on-demand Competitive Intelligence “crew”? That idea became VyaparX, your personal business insider.

What it does

VyaparX is a Streamlit-powered web app where you enter any company name or URL and instantly get:

Website scraping & data extraction (WebCrawlAgent via Browserbase)

Insight synthesis (InsightSynthesizerAgent)

Polished PDF/Markdown report (ReportWriterAgent)

Marketing copy snippets (CopywriterAgent)

Quality checks & reasoning trace (FeedbackEvaluatorAgent via W&B Weave)

In under 2 minutes, VyaparX delivers SWOT tables, market-position graphs, opportunity heatmaps, and ready-to-use executive summaries.

How we built it Frontend: Streamlit for rapid prototyping & polished UI (custom CSS, gradient logo)

Agents & Orchestration: CrewAI multi-agent framework tying together:

Exa for company/category lookup

Browserbase (Stagehand) for scraping

Weave for introspection & feedback loops

AI Models: OpenAI GPT-4 for analysis generation, GPT-3.5-turbo for fast background tasks

Cloud & Deployment:

Google Cloud Storage for intermediate data

Vertex AI for memory persistence

Hosted on Fly.io with Docker

CI/CD: GitHub Actions auto-deploy on merge

Challenges we ran into Rate-limits & scraping delays: We built exponential back-off and parallelized crawling.

Maintaining agent context: Chaining prompts across 5+ agents required careful prompt-design and prompt-templates stored in Notion.

Styling Streamlit under time pressure: Rapid CSS overrides to get the logo, gradient headers, and responsive layout working in under an hour.

Accomplishments that we’re proud of Multi-agent orchestration on a single page with live progress spinners.

End-to-end automation, from company lookup to downloadable report.

Built & deployed a robust MVP in under 24 hours, complete with CI/CD pipeline.

Traceable reasoning via W&B Weave, so users can inspect intermediate agent decisions.

What we learned Designing prompt-templates as first-class components significantly improved agent handoffs.

Streamlit’s CSS theming can be bent to our will for gradients, clipped‐text, and full-width call-to-action buttons.

Multi-agent systems shine for complex workflows when each agent has a clear, isolated responsibility.

What’s next for VyaparX Dark mode & custom color themes for brand alignment

User accounts & history dashboards powered by Firestore

Custom competitor sets (upload your own URLs)

Interactive charts with Recharts for deeper drill-down

Enterprise integration: Slack, Teams, and Zapier connectors

Built With

• browserbase/stagehand-storage-&-memory:-google-cloud-storage
• fly.io
• github
• html/css-ai-&-orchestration:-crewai-multi-agent-framework
• languages-&-frameworks:-python
• openai-gpt-4-&-3.5-turbo-data-&-scraping:-exa
• streamlit
• vertex-ai-memory-observability:-w&b-weave-deployment:-docker