Build notes
Notes on streaming AI into real products: clients, models, costs and protocols.
-
Proxy SSE Through nginx Without Buffering
Configure nginx for unbuffered SSE, set streaming headers and timeouts, and verify that AI responses arrive incrementally.
-
Server-Sent Events for Chat Developers: Step by Step
Build streaming AI chat with server-sent events: choose a client, parse messages, render replies, and handle cancellation and reconnects.
-
How to Read Usage Logs and Find Expensive Requests
Follow a practical walkthrough to rank AI requests by cost, investigate token usage, trace retries, and verify savings.
-
Caching Repeated Context in Support Bots: 7 Steps
Build a layered caching strategy for support bots with stable prompt prefixes, safe cache keys, policy invalidation, and streaming benchmarks.
-
Trial Credits and Abuse Prevention in Chat Apps
Build a safer AI chat trial with credit ledgers, streaming spend controls, signup checks, and abuse monitoring in six practical steps.
-
-
Compare AI Models for Customer-Support Chat Widgets
A step-by-step guide to comparing support chatbot models on accuracy, streaming speed, safety, cost, and real customer outcomes.
-
Cursor with an OpenAI-Compatible Relay: Key Settings
Connect Cursor to an OpenAI-compatible relay, configure the base URL and model ID, and troubleshoot streaming, authentication, and agent tools.
-