dream: periodic memory consolidation

This commit is contained in:
nanobot
2026-07-15 07:02:01 +02:00
parent 9d08290b4f
commit edfd675662

View File

@@ -8,10 +8,9 @@ This file stores important information that should persist across sessions.
## Runtime / Deployment ## Runtime / Deployment
- Installed as a `uv` tool: package `nanobot-ai`, update via `uv tool upgrade nanobot-ai`
- Runs as a systemd user service `nanobot.service` - Runs as a systemd user service `nanobot.service`
- Model switching via `my` tool requires `tools.my.allow_set = true` in config - Model switching via `my` tool requires `tools.my.allow_set = true` in config
- `compact-memory-auto-daily` cron job runs in Telegram session (id: 69484aa6) so reports deliver to Telegram instead of WebUI - `compact-memory-auto-daily` runs as a system crontab script (not a nanobot cron/jobs.json job) with a fresh session per run; delivers its report straight to Telegram via Bot API
## Integrations ## Integrations
@@ -20,13 +19,12 @@ This file stores important information that should persist across sessions.
## Architecture ## Architecture
- nanobot podporuje cross-channel session continuity přes `unifiedSession: true` v `config.json` pod `agents.defaults` - nanobot podporuje cross-channel session continuity přes `unifiedSession: true` v `config.json` pod `agents.defaults`; uživatel nemá `unifiedSession` povolený (default false) — zapnutí sjednotí jen budoucí zprávy, existující session soubory vyžadují ruční merge/rename
- Uživatel nemá `unifiedSession` povolený (default false); zapnutí sjednotí jen budoucí zprávy, existující session soubory vyžadují ruční merge/rename - Dream routing (SOUL.md → personality, USER.md → user profile, MEMORY.md → project knowledge, AGENTS.md → work instructions) and core logic are hardcoded in nanobot Python source; adjustable settings: intervalH, modelOverride, maxBatchSize, maxIterations
- Dream routing rules (SOUL.md → personality, USER.md → user profile, MEMORY.md → project knowledge, AGENTS.md → work instructions & decision-making rules) and core logic (Phase 1/2 instructions, fact/decision/ephemeral classification, deduplication) are hardcoded in nanobot Python source; only adjustable settings are intervalH, modelOverride, maxBatchSize, maxIterations
- User wants native deterministic command dispatch (`!command text` → registered Python function/external app, no LLM involvement) and is exploring litellm proxy for Ollama - User wants native deterministic command dispatch (`!command text` → registered Python function/external app, no LLM involvement) and is exploring litellm proxy for Ollama
- 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` - 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`
- Zálohy: wood.hell → pivo.hell (rekurzivní) - Zálohy: wood.hell → pivo.hell (rekurzivní)
--- ---
*This file is automatically updated by nanobot when important information should be remembered.* *This file is automatically updated by nanobot when important information should be remembered.*