My 2024 review and high-level report. Let's start with some stats for the nostr.build: - 720 Million media requests served - 650 TiB of bandwidth consumed - 8 Million unique visitors - 85% average cache hit rate - Improved median CPU time per request from 200 down to 2 milliseconds per request - 20 requests per second are served on average - 99.994% uptime for the hosted media - 98% uptime for the website and uploading servers - 2.5K daily free uploads on average Some improvements for the free uploads: - Increased free upload size from 10MiB to 50MiB - Increased image resolution from 1980px to 3840px - Increased per server RAM from 1GiB to 4GiB - Added full mirror of all uploaded media to the additional vendor (backup) - Implemented moderation headers (can be used to hide NSFW content if desired) Now some of the stats of my dev work for the nostr.build: - 1,333 git commits - 42 repositories with my contributions - 10k+ lines of code written, or changed Some features and completed projects: - Rewritten nostr.build account page (AI, stats, nostr share, drag-n-drop, variaty of sizes, poster for video, etc.) - Added AI Studio, onboarded multiple models from SD and Flux. - Added media statistics - Rewritten media serving front-end (dropping CPU utilization by 100x) and improving TTFB as well as cache hit rate (from 50%+ to 85%+ on average) - Implemented variaty of the spam, and other illegal media upload prevention mechanisms (automated AI media scanner, PhotoDNA scanner, etc.) - Implemented AI Moderation API that is shareable with other nostr media hosts (e.g., nostrcheck.me) - Implemented upload of pdf, zip, etc., with full virus scanning. - Various improvements to the PHP code, bug fixes and optimizations. - Implemented e.nostr.build to allow easy embedding and sharing of the video/audio files for the account holders - Referal links and account credits - Onboarded more nostr clients for uploads (e.g., YakiHonne) Some of the prototyping I've done for nostr.build: - Semantic search of the media (ML based), multilingual - Basic search for media using AI Vision - Video transcoding using serverless and server-based setups - Video transcoding to DASH/HLS (CMAF) - Multi-party audio/video chat Some of the in-flight projects that I'll have to complete in 2025: - Seamless implementation of the Blossom protocol (70% done) - Rewrite of the post-upload scanning to utilize workflows (50% done) - New video upload and transcoding experience, HLS, DASH, VBR video/audio (10% done) - New NoSQL based data-plane backend with the new API (50% done) - Full rewrite of the Accounts page (5% done, blocked by new API and data-plane) - Additional models for the AI Studio (90% done) - <Secret projects with new experiences and features> (25% done) I am sure I am forgetting a bunch, and this is only done from my memory.