What We Improved in PolySignal
We significantly upgraded PolySignal's performance, infrastructure, and user experience across multiple fronts.
Performance: The Events API went from 11–16 seconds to under 1 second using a PostgreSQL materialized view that pre-computes the latest snapshot per market. We also added composite database indexes and reduced connection pool sizes to stay within the managed PostgreSQL connection limit.
Infrastructure: We moved the backend from DigitalOcean App Platform (which blocks WebSocket upgrades) to a dedicated Droplet with Nginx as a reverse proxy. A DigitalOcean Load Balancer sits in front for health checks and traffic routing. This unlocked true WebSocket support for live dashboard updates.
CI/CD: We set up a dual GitHub Actions pipeline — one workflow auto-deploys the frontend via App Platform, another SSH-deploys the backend to the Droplet on every push to main.
UX: Resolved markets are hidden by default with a toggle. Newly extracted events are pinned to the top of the dashboard with a glow animation. Labels now say "Last synced" instead of "Updated." The Liquidity Heatmap was rebuilt with YES/NO book toggle and judge-friendly empty states.
AI: All AI endpoints (trading signal, sentiment, prediction) are now fully operational at under 1 second response time.
Log in or sign up for Devpost to join the conversation.