# Hermes Persistent OS

**Building an autonomous, self-improving agent system on Hermes.**

## Status
- **Phase:** Active
- **Last updated:** 2026-06-20
- **Priority:** High

## Goals
- [x] Establish personal-agent-os repo as the external brain
- [x] Deploy daily brief with context-from chaining (free data sources)
- [x] Document masterclass patterns (wakeAgent gates, no-agent polling, silent delivery)
- [ ] Build email triage gate (wakeAgent pattern, Himalaya-backed)
- [ ] Build compliance monitoring gate (context-from, regulatory RSS)
- [ ] Extend daily brief to save entries into knowledge-base/ automatically
- [ ] Wire a weekly index cron that rebuilds tag maps across the repo

## Architecture Notes
- **Primary interface**: Telegram gateway (leonard profile)
- **Brain**: personal-agent-os repo on Gitea
- **Patterns**: cron script → context injection → agent curation → delivery
- **Cost model**: Free data sources (wttr.in, HN Firebase, GitHub API) keep the expensive LLM pass minimal

## Resources
- `knowledge-base/automation-blueprints/hermes/` — documented patterns
- `agent-config/` — Hermes profile configs
- `_meta/decision-log.md` — key decisions and why

## Related KB Entries
- automation-blueprints/hermes/ — all patterns live here
- hermes-integrations/ — platform-specific setup notes
- ai-agent-patterns/ — broader agent design patterns
- tool-evaluations/ — evaluations of tools used in this project