Skip to content

CATCHME-light.md

Latest commit

 

History

History
77 lines (50 loc) · 2.13 KB

CATCHME-light.md

File metadata and controls

77 lines (50 loc) · 2.13 KB
name catchme-light
description Query the user's always-on screen recording via the catchme CLI. Use when the user mentions catchme, activity recording, screen capture, digital footprint, or asks about what they were doing/seeing earlier.

CatchMe — Screen Memory CLI

Query the user's always-on screen recording via the catchme CLI.

Prerequisite: catchme awake must already be running in the background.

Find & Activate the Environment

The conda env name may not be catchme. Discover it first:

find "$(conda info --base)/envs" -path "*/bin/catchme" 2>/dev/null | head -1

This prints something like /opt/miniconda3/envs/ENVNAME/bin/catchme. Extract the env name from the path and activate it:

conda activate ENVNAME

Verify awake is running:

catchme ram

If no awake process appears, inform the user that catchme awake needs to be started first.

Commands

catchme ask -- "<question>"

Query activity history in natural language. The retrieval pipeline navigates the activity tree, inspects screenshots, and returns a natural language answer.

catchme ask -- "What was I working on this morning?"
catchme ask -- "When did I last open Figma?"
catchme ask -- "Summarize my afternoon session"

catchme cost

Show LLM token usage (last 10 min / today / all time).

catchme disk

Show storage breakdown (database, screenshots, trees) and total event count.

catchme ram

Show RAM usage of all running catchme processes.

catchme web [-p PORT]

Launch web dashboard (default: http://127.0.0.1:8765).

Troubleshooting

Symptom Fix
find returns nothing catchme is not installed in any conda env
command not found: catchme Wrong env activated — re-run the find command above
ask returns no results catchme awake not running — check catchme ram
LLM errors Check data/config.json in the catchme project root

Notes

  • catchme awake must be running for ask to have recent data
  • All data is stored 100% locally — nothing leaves your machine unless a cloud LLM API is configured