Run the research cache and the decision council before building anything

Ready now

Every research query runs /research check first against 312+ cached entries; a miss runs Perplexity through the cache-aware wrapper and then a NotebookLM synthesis; entries carry staleness windows by domain; a hook blocks direct calls.

Kit

workspace-os

You need

Account	Why	Free or paid	Plan
Perplexity API (via the llm CLI)	Web research with citations (sonar-pro model). Every query goes through /research so the cache is checked first.	Paid	Pay as you go API credits (a few dollars covers dozens of queries). Perplexity Pro subscribers get monthly API credit.
Google NotebookLM	Synthesis over cached research: every Perplexity miss is followed by a NotebookLM pass (both, always, on a cache miss).	Free	Free with a Google account; NotebookLM Plus comes with Google One AI Premium but is not required.
Python 3.12 + uv	Runs every pipeline script (video gates, library export, research cache, Brain export, PDF generator, MIDI tools) inside per-project virtual environments.	Free	Free.
Claude Code (Anthropic Claude Pro or Max)	The AI operator every system runs through. Skills, hooks, memory, MCP servers and the Brain plugin all load inside Claude Code.	Paid	Claude Pro (20 USD/mo) works for light use; Max (100 USD or 200 USD/mo) for daily multi-hour sessions. Anthropic terms require each user to hold their own subscription; never share a login.

What you bring

Your research account credit
A Google account
Your question, with the context around it

The steps

Install both CLIs and sign in
Create the cache and its scripts
Create the /research skill
Install the cache-check hook
Put the gate in CLAUDE.md
Run a real research question end to end
Weekly staleness pass
Build the council for decisions

Goes further if you have already built Install hooks and gates so your rules enforce themselves. You can build this without it.

Get the full playbook

Plug in your AI Read it in the Library

Part of: Get your database sorted