Optimizing Sourcegraph Cody-Generated Code for Performance

Cody may choose to implement a feature by calling multiple existing functions sequentially when a more efficient combined query or batch operation would be better. For large codebases, it may not know that a specific caching layer exists and generates a new cache rather than using the established one

Provide Cody with context about performance-critical patterns in your codebase. Document caching and batching abstractions in a way that makes them discoverable. Review generated code on hot paths to ensure it uses established performance infrastructure

Before optimizing, measure. Use your browser's Lighthouse audit to get a baseline score. Check Core Web Vitals (LCP, FID, CLS) — these directly affect your Google ranking. Use the Performance tab in DevTools to identify which components take the longest to render. For server-side performance, add timing logs to API routes to find slow endpoints.

The fastest way to improve performance: optimize and lazy-load images, enable compression on your server, remove unused CSS and JavaScript, implement code splitting so users only download code for the page they're viewing, and add proper caching headers for static assets. These changes alone can cut load times significantly.

If your Sourcegraph Cody app uses a database, slow queries are often the biggest performance bottleneck. Add indexes on columns you filter or sort by. Use SELECT to fetch only the columns you need instead of SELECT *. Implement pagination for lists — loading 10,000 rows at once will always be slow. Consider adding a caching layer for data that doesn't change frequently.