# Current Context (Tony / Local PC)

> Update this when major changes occur (new hardware, model switches, vault restructuring).

- Primary machine: Windows 11 desktop with RTX 3060 12GB
- Inference: LFM2.5-2.6B via llama.cpp at :1920 (primary, 65K ctx, ~145 tok/s), FreeToken gpt-oss-20b at :1919 (manual, ~120 tok/s, can't coexist on 12GB GPU), brother nodes on tailnet (5060: Qwen3.8 27B 88K, 5090: Qwen3.6 27B MTP 200K)
- Models available: Gemma-4-26B, gpt-oss-20b, various Qwen, LFM2.5, Qwen3.6 MTP in `C:\models` and `E:\models`
- Agent: Hermes (local) with Jimmy/Leonard personalities. Jimmy = main agent, Leonard = operator/consolidator
- Memory: Obsidian vault (Agent OS layout) = external brain; Hermes native memory also enabled
- Key projects: VeriPath (core), Hermes local setup, Thornhill Construction site, vault-kb RAG
- Model preference: LFM2.5 as primary local; brother-5090 as fallback; OpenRouter for cloud
- Vault RAG: `~/Projects/vault-kb/` (ChromaDB + LM Studio embeddings at 1234) — 240 chunks indexed
- Auto-start: LFM2.5 via `start-models.bat` in Windows Startup; watchdog cron auto-relaunches
- Voice/TTS: wake word "hey Jimmy" (sherpa), Edge TTS AriaNeural, auto_tts=true, barge-in enabled
- Git sync: `vault-sync` cron pushes vault to Gitea every 6h