Ingeniería
Deep Recall archive: start bots fresh without losing chat history
Bots can't delete chat, so every wake re-reads a growing history and burns tokens.
Qué dolía
@32Bridger explains a word-for-word Deep Recall archive (day files, topic files, INDEX.md), lookup via index → Bloodhound search → Librarian helper, and a 5-step Renew handoff so a fresh bot keeps history without loading it all. Kit path and hand steps are in the Article. Not affiliated with xAI.
Qué instalar o crear
Archive word-for-word → index-only on wake → Bloodhound → Librarian → Renew with handoff.
Bots
Main bot
Reads only INDEX.md and handoff; uses Bloodhound then Librarian for old detail.
Librarian
Reads one topic file in a short-lived chat and returns a short cited answer.
Plugins
Ninguno el primer día. Sigue los pasos del primer uso.
Rutinas
Ninguna en el primer uso. Mira primero una pasada manual.
El prompt
Esto es un consejo, no un prompt de sistema para pegar. El autor no publicó uno. Sigue los pasos del primer uso.
Primer uso
- 1Read https://x.com/32Bridger/status/2113429530182295647 and the Article (credit @32Bridger).
- 2Either drop the Deep Recall kit into chat and say "Set this up," or build day/topic files + INDEX.md by hand from a real chat export.
- 3Install Bloodhound (read-only search) and test known word / nonsense / no-arg.
- 4Add Librarian (skill+helper, separate bot, or small slice reads) and standing rules: index-first, never whole-archive.
- 5When chat is long, Renew: old bot writes handoff + updates archive, you create a fresh bot that reads only handoff + INDEX.md.
Valla de aprobación
Bot asks before anything irreversible; you delete the old bot yourself after Renew.
Qué ajustar
Steal the rule: never read the whole archive into chat; search, then ask a short-lived helper.
Fuente
₿ridger @32Bridger · Publicación original
Sin afiliación con xAI. No escribimos este prompt. Siguen aplicando las condiciones de las plantillas de terceros.
Paquete Pro
¿Quieres la configuración completa, no solo la idea?
Los paquetes Pro te dan el texto de la tarea, los prompts, los horarios y los tropiezos de los casos de uso más populares.