K
KioRAG Pipelines with Postgres + pgvector: A Production Guide
Why we chose Postgres over Pinecone for our vector store and how we optimized hybrid search for 10M+ documents.
📄
A deep dive into the architecture behind busky.ai’s agent orchestration engine — how we moved from simple completions to autonomous multi-step reasoning pipelines.
Why we chose Postgres over Pinecone for our vector store and how we optimized hybrid search for 10M+ documents.
A complete walkthrough of everything new in busky.ai v2 — Claude 4, Gemini 2.5 Flash, custom agents, and project-level context.
Lessons learned from fine-tuning Llama 4 and DeepSeek on legal and financial corpora — data prep, LoRA configs, and evaluation.
As models get smarter, the craft shifts from writing clever one-shot prompts to designing reusable, composable prompt systems.