Tech
Memory Architecture
How CollectHive stores and retrieves knowledge — three storage types, extraction pipeline, and freshness mechanisms.
Platform Stack
Five services, four networks, three environments, and the product stack — how CollectHive's infrastructure fits together.
Recent
What We Capture Per User
A full inventory of the telemetry, activity logs, and behavioural data Amber collects on production — and why each category exists.
Web Research
How Amber reads the web — Firecrawl, smart extraction, and what happens when sites say no.
Cloudflare
How we use Cloudflare Pages, Access, and DNS to host and protect the CollectHive knowledge hub.
Dev Containers: What's New
Live code previews, fixed sign-in redirects, and automatic idle pausing — what changed and what it means for your container.
Mining Your Agent's Mistakes
How a one-line incident log and transcript analysis revealed six recurring failure patterns in our AI agents, and why the fix wasn't what we expected.
Amber Monitor
A local DevOps dashboard that watches Agent Amber across all environments, with AI-powered triage and feature drift tracking.
How This Site Gets Updated
A content agent, markdown files, and a git push — how site.collecthive.ai goes from draft to live in 30 seconds.
CI/CD Pipeline
From push to deploy — TypeScript checks, tests, Docker builds, auto-deploy to staging, manual gate for production.
Decision: Hybrid Storage Over Pure Vectors
Why we chose structured facts + vector conversations + active decisions instead of a single vector database.
Heartbeat & Automation
How agents work autonomously — heartbeats, cron jobs, and the constraints that keep them safe.
Scope & Promotion
How knowledge flows between users — private by default, promoted only with explicit approval.