Back to Logbook
May 10, 2026 Project: #Portfolio

The AI layer goes in.

energised

Big day. Built the entire AI subsystem in one push. Five migrations for the AI tables: ai_generations, ai_prompts, ai_logs, ai_templates, ai_providers. Then nine AI skill classes — blog drafting, work entries, logbook entries, social posts, SEO metadata, lead detection, NLP classification, emotion analysis, translation. Each is a self-contained class with its own system prompt.

Wrote AIManager to abstract OpenAI and Gemini behind one interface with provider fallback. Also got the first cut of the RAG service running — document ingestion, basic semantic search. It is keyword-only for now, vectors come later.

The platform is no longer just a CMS. It can think.

Lessons Learned

A provider-agnostic interface beats hard-coding any single LLM API.