# Long-term Memory

This file stores important information that should persist across sessions.

## Project Context

- Both user and assistant run on the same nanobot Docker image
- Ollama cloud subscription: na požádání zobrazit aktuální využití (usage)
- 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
- 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`
- Zálohy: wood.hell → pivo.hell (rekurzivní)
- „giga" zařízení k integraci do Zabbix monitoringu

## Agent Model Selection
- Prioritizes agentic performance, correct tool calling, and overall result quality
- Conservative fallback nanobot agent model: DeepSeek V3.2:cloud
- OpenRouter is pay-per-token alternative to Ollama subscription for model access
- Gemini Flash via Google AI Studio Free Tier: 15 RPM limit — unusable for agent work; only viable for simple prompts without tool calls
- Gemini Flash via Google AI Studio Tier 1 (paid): 360+ RPM
- Haiku (Claude) via OpenRouter: high rate limits, viable agent model alternative
- Gemini Flash Lite via OpenRouter: high rate limits, viable agent model alternative
- Kimi K2.6 has known bug with random switching to Chinese output (reported by Cursor and Reddit users) — critical risk for Czech use
- minimax-m3:cloud is blocked for nanobot agent deployment due to empty tool result responses (ollama/ollama #16389)
- deepseek-v4-pro:cloud is blocked for interactive nanobot agent use due to 15.4 tok/s and 57s TTFT
- deepseek-v4-flash:cloud is a viable nanobot agent alternative with 1M ctx, MIT license, and ~30-50 tok/s speed
- qwen3.5:397b-cloud is a viable nanobot agent alternative with multimodal support, 1M ctx, 201 languages including Czech, but has speed and accuracy tradeoffs
- devstral-2:123b-cloud is a viable nanobot agent alternative with Terminal-Bench 77.3%, coding-only focus, and 128K ctx limit
- User is interested in switching to GLM-5.2:cloud as primary nanobot agent model if/when it becomes available on Ollama Cloud
- User prefers Qwen model for deep research tasks (not currently in presets)
- Available model presets: gemini-flash, gemini-flash-lite, glm, haiku, kimi, minimax, nemotron-3-super, sonnet
- Model specs: `nemotron-3-super` ctx 262144 / max output 16384; `gemini-flash` ctx 256000 / max output 16384

### NuGet package caching
- Evaluating NuGet package hosting/caching on Linux
- Preferred solution: BaGetter (bagetter/BaGetter) — active BaGet fork with multiple upstream mirror support (PR #269), solves original BaGet's single-upstream limitation

## 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`
- Model switching via `my` tool requires `tools.my.allow_set = true` in config

---

*This file is automatically updated by nanobot when important information should be remembered.*