🦾 ESMGU LOKÁL — Best Local Full Agent

Backend: Ollama | Default: qwen3:8b | Sessions: 1

Komplement: Ritch Bitch (Spark sovereign)

Endpoints

• POST /api/local/chat {message, session_id?, model?, use_tools?}
• GET /api/local/sessions
• GET /api/local/tools
• GET /health

Available models (Ollama)

qwq:latest (default reasoning), gpt-oss:20b (general), codestral (code), qwen3:8b (fast), deepseek-r1:8b (reasoning)

Tools

fs_read, fs_list, shell_exec (denied list aktivní), http_get (jen localhost+Tailscale), esmgu_query, tresor_request