Prism
2026Can a RAG pipeline feel instant if the expensive half happens off the request?
A chat-with-video workspace. Paste a YouTube URL and it queues a job immediately rather than making you wait — a background worker fetches the transcript, chunks it, embeds it and stores the vectors. Questions get embedded too, matched by similarity search over an HNSW index, and answered against the chunks that actually came back.
- NestJS
- PostgreSQL
- pgvector
- Redis
- BullMQ
- Prisma
- Gemini