Lightweight OpenTelemetry for Chatbots
How to add distributed tracing, custom metrics, and cost tracking to three different chatbot architectures with only 25MB of overhead. Cloud-native observability without Prometheus or Grafana.
Read More →Insights and learnings from building real-world software.
How to add distributed tracing, custom metrics, and cost tracking to three different chatbot architectures with only 25MB of overhead. Cloud-native observability without Prometheus or Grafana.
Read More →When building chatbots, one critical decision is whether and how to persist conversation history. Lessons learned from deploying three different implementations with different persistence strategies.
Read More →Most chatbots are black boxes. I built a step-by-step agent that shows every query and result — and the difference in trust, debugging, and usefulness was dramatic.
Read More →Three chatbot implementations: Kafka + HTTP polling, WebSocket streaming, and WebSocket + SQL. Same Django backend, completely different flows and PostgreSQL usage patterns.
Read More →Design principles extracted from Claude Code's multi-step agent pattern: structured output, guardrails in code, error recovery, and more.
Read More →