Implementation notes and engineering decisions from production systems — M-Pesa integrations, AI features, database architecture, and cloud infrastructure.
RSS feedopens in new tabHonest lessons from building a Claude-powered chatbot, a pgvector semantic search pipeline, and an AI assistant into real products — covering prompt architecture, cost decisions, rate limiting, and where AI actually earns its place.
How to implement semantic search in PostgreSQL using the pgvector extension and OpenAI embeddings — covering schema design, embedding generation, cosine similarity queries, IVFFlat vs HNSW indexing, and fallback strategies.
A complete guide to integrating Safaricom's M-Pesa STK Push using the Daraja API in a Next.js 14 App Router project — covering authentication, STK Push request, callback handling, idempotency, and error states.