# Session: 2026-05-05 12:20:05 UTC

- **Session Key**: agent:socrates:telegram:direct:8386527252
- **Session ID**: 1a65759c-fbc5-4287-b629-488244df126e
- **Source**: telegram

## Conversation Summary

user: Can you give me the overview of default models for our agents and the fallback chains?
assistant: Here's your agent/fallback overview:

**Current Session (Socrates)**
- Running model: `kimi-k2.6:cloud`

**From MEMORY.md — The Board's default models:**

| Agent | Default Model | Role |
|-------|--------------|------|
| **Socrates** 🧠 | `ollama/kimi-k2.5:cloud` → now `kimi-k2.6:cloud` | Backend, pipelines, AI architecture |
| **Daedalus** 🎨 | `ollama/gemma4:26b` | Frontend, UI/UX, design system |
| **Wadsworth** 📋 | `ollama/glm-5.1:cloud` | Chief of Staff, routing, coordination |

**Gateway Fallback Chain** (from OpenClaw config audit, 2026-04-18):
```
glm-5.1:cloud → kimi-k2.5:cloud → gemma4:latest → qwen2.5-coder:7b
```

**Family Assistant Pipeline** (v1.0, local on Gaming PC):
- Default LLM parse: `qwen2.5-coder:7b`
- Vision/OCR: `qwen3-vl:8b` (keep_alive: 0)
- Embeddings: `nomic-embed-text`

**Permanent Model Rules** (family assistant):
- ✅ `qwen2.5-coder:7b` — single model for ALL pipeline tasks
- ✅ `qwen3-vl:8b` — vision/OCR only
- ⛔ No `home-assist` model — ever
- ⛔ No `deepseek-r1` — burns thinking tokens, fails JSON
- ⛔ No `qwen3:8b` — thinking can't be disabled, 3-13x slower

**Note:** Your current session shows `kimi-k2.6:cloud` running — looks like the model got bumped from k2.5 to k2.6 since the last config audit. Auto-failover or manual update?