Zero-Downtime Schema and pgvector Index Migrations on Live PostgreSQL
How to evolve PostgreSQL tables and pgvector HNSW indexes in production without tripping lock queues, starving connection pools, or spilling graph builds to disk.
Closing the Reader Feedback Loop: Search Console, Cookieless Umami, and JakeMon
How Wikantik, Pressroom, and JakeMon combine Google Search Console diagnostics with cookieless Umami telemetry to prioritize content improvements.
The Wikantik Reader's Field Guide: How Storage, Hybrid Search, and Dual MCP Fit Together
A guided tour through Wikantik's 19 engineering posts—connecting POSIX Markdown storage, BM25 + pgvector hybrid search, and dual MCP servers.
The Anatomy of Agent-Grade Content: Semantic Triples and RAG Invariants
Explore how structural spines, deterministic triples, and RAG invariants transform messy wiki prose into high-precision agentic retrieval context.
The Agentic Dogfood Loop: How Requirements Enter Through Wikantik and Exit in Code Across the Ecosystem
How Wikantik serves as the living requirements and governance engine for Pressroom, Stockroom, and Jakemon through agent briefings and architectural ADRs.
Why We Clobbered the Knowledge Graph: Search Ranking Realities and Architecture Simplification in Wikantik
Why Wikantik deleted 70,500 lines of Knowledge Graph code: empirical search ranking failure, operational latency, and the power of dense vector retrieval.
Designing an Agent-First Publishing Engine: Pictures by Default, Scoped API Tokens, and Native MCP in Pressroom
For more than two decades, content management systems and blogging engines have been designed around a singular assumption: a human operator sits in front of a rich-text WYSIWYG editor, manually uploading photographs from a local filesystem, cropping thumbnails, and fiddling with taxonomy…
The Evolution of Wikantik 2.5: Tiered Agent Briefings, Sandboxed Visual Architecture, and Markdown Ingestion
From storage lifecycle invariants in v2.4.63 to tiered project-context briefings (ADR-0011), sandboxed PlantUML/Mermaid visual modeling, and rich Markdown ingestion.
The Unfinished Roadmap: Horizontal Scale, PgBouncer, and What Wikantik Still Needs
Architectural analysis of Wikantik's single-host operating boundaries, the prerequisites for horizontal scaling with PgBouncer, and section-level Knowledge Graph reranking.
The Operational Reality of Agents in Production: MCP Protocol Edge Cases
Production realities of Model Context Protocol: resolving non-daemon thread shutdown deadlocks in the MCP SDK, AST parsing overhead in citation generation, and frontmatter curation.
Silent Data Loss and Destructive Admin Tools: Fixing Storage Invariants
Postmortem on storage durability invariants: fixing in-place file truncation, PrintWriter exception suppression, and destructive index rebuild cascades in Wikantik 2.4.62.
Surviving the N+1 Storm: Database Profiling and Statement Pruning
Tracing N+1 statement amplification in Wikantik's JDBC layer and the refactoring that collapsed thousands of round-trips into constant-time queries.
The 79 GB Disk Leak: Jena TDB2 B+Trees and Storage Subsystem Hardening
Postmortem of an Apache Jena TDB2 disk leak: B+tree copy-on-write allocation, scheduler timer limitations, and storage subsystem repairs in Wikantik 2.4.62.
Decomposing the God Module: Package Extraction and the Complexity Ratchet
Decomposing the monolithic WikiEngine core: reducing CBO coupling from 143 to 86, extracting sub-modules, and enforcing the CI complexity ratchet.
Beyond the Fork: Settling the Architecture of an Agent-Grade Knowledge Base
An engineering analysis of Wikantik's tri-tier storage architecture, dual MCP protocol gateways, and post-fork stabilization objectives.