dream: periodic memory consolidation

Good, MEMORY.md is clean now. The operational details about `/note` skill backend, URLs, reminders DB schema, and `remind_cli.py` specifics were all already captured in their respective SKILL.md files — they've been removed from MEMORY.md to maintain MECE. The typo fix and consolidation of `unifiedSession` + Dream routing rules are done.

Now let me check if there's anything in the conversation history that needs to be added. The history was truncated, so let me check what's there:

- [skip] entries — no new facts
- [durable] Dream routing rules — already added to MEMORY.md
- [durable] Dream Phase 1 — truncated, can't see full content

The [durable] Dream routing fact is already in MEMORY.md. The [durable] Dream Phase 1 fact was truncated so I can't process it fully. Let me check if there's anything else worth adding from what I can see.

The conversation history is mostly [skip] entries (informational queries) and one [durable] fact about Dream routing that's already captured. No new facts to add.

Let me also verify SOUL.md and USER.md don't need changes — they look clean and well-organized already.

Summary of changes made:

**MEMORY.md:**
- Removed 3 operational details now covered by their SKILL.md files: `/note` skill backend specifics, URL formatting rules, reminders DB schema details, `remind_cli.py` dependency/flag notes, batch-deleting reminders tips
- Fixed typo: "rekurzní" → "rekurzivní"
- Consolidated 2 `unifiedSession` entries into 1 + added Dream routing rules fact from conversation history

memory/.dream_cursor

@@ -1 +1 @@
-377
+378

memory/MEMORY.md

@@ -6,20 +6,14 @@ This file stores important information that should persist across sessions.
 
 - Both user and assistant run on the same nanobot Docker image
 - Ollama cloud subscription: na požádání zobrazit aktuální využití (usage)
-- `/note` skill: backend uses SQLite (not markdown); must be deterministic, log all operations, split into shorter prompt + Python script
-- URLs should be on separate lines in both `list` and `show` output, not buried in long text lines
-- Reminders DB schema has no "frequency" column; scheduling uses separate tables (schedule_at, schedule_cron, schedule_random, reminder_fires); uses cron expressions and absolute timestamps
 - nanobot podporuje cross-channel session continuity přes `unifiedSession: true` v `config.json` pod `agents.defaults`
-- Uživatel nemá `unifiedSession` povolený v `config.json` (default false)
+- Uživatel nemá `unifiedSession` povolený (default false); zapnutí sjednotí jen budoucí zprávy, existující session soubory vyžadují ruční merge/rename
-- Zapnutí `unifiedSession` sjednotí jen budoucí zprávy; existující session soubory vyžadují ruční merge/rename pro propojení minulých konverzací
+- Dream routing rules (SOUL.md → personality, USER.md → user profile, MEMORY.md → project knowledge) are hardcoded in Dream prompt, not user-configurable beyond interval/model/batch size
 - User wants native deterministic command dispatch (`!command text` → registered Python function/external app, no LLM involvement) and is exploring litellm proxy for Ollama
 - User considers Raspberry Pi good for testing
 - User wants to try writing a nanobot extension in TypeScript
 - Pi agent načítá všechny AGENTS.md od kořene filesystemu do cwd najednou jako prostou concatenaci; Claude Code používá lazy loading pro CLAUDE.md v podadresářích a podmíněná pravidla v `.claude/rules/*.md` s atributem `paths`
-- `remind_cli.py` vyžaduje Python modul `croniter` (není nainstalován defaultně)
+- Zálohy: wood.hell → pivo.hell (rekurzivní)
-- `remind_cli.py remove` vyžaduje `--id ID` nebo `--keyword KEYWORD`, nepřijímá poziční argumenty
-- Při batch-deleting reminders podle display number mazat od nejvyššího ID nebo použít keyword matching (ID se posouvají po každém smazání)
-- Zálohy: wood.hell → pivo.hell (rekurzní)
 - „giga" zařízení k integraci do Zabbix monitoringu
 
 ## Agent Model Selection