omp: drop static model lists, fix stale model refs (miche)

- Remove static model lists from llama-swap.miche and llama-swap.byte
  (discovery: type: proxy now fetches live models from /v1/models)
- Update byte model IDs: qwen3.6-35b-heretic-mtp/a3b-mtp -> qwen36-35b-genesis,
  gemma-4-12b-heretic -> gemma4-12b-hauhau
- Update miche model IDs: gemma4-12b -> gemma4-12b-hauhau,
  qwen3.6-27b-mtp-rocmfp4-turbo -> qwen3.6-27b-mtp,
  qwen3.6-35b-a3b-mtp-rocmfp4 -> qwen36-35b-genesis

dot_omp/private_agent/config.yml

@@ -8,12 +8,12 @@ theme:
   light: light
 setupVersion: 1
 modelRoles: 
-  default: deepseek/deepseek-v4-flash:high
+  default: zai-coding/glm-5.2:high
   task: llama-swap.miche/qwen3.6-27b-mtp
   smol: llama-swap.byte/qwen36-35b-genesis
   plan: zai-coding/glm-5.2:xhigh
   slow: minimax-code/MiniMax-M3:high
-  vision: llama-swap.miche/gemma4-12b
+  vision: llama-swap.miche/gemma4-12b-hauhau
   tiny: llama-swap.byte/qwen36-35b-genesis
 retry: 
   fallbackChains: 
@@ -21,17 +21,17 @@ retry:
       - zai-coding/glm-5.2
       - minimax-code/MiniMax-M3
       - deepseek/deepseek-v4-pro
-      - llama-swap.miche/qwen3.6-27b-mtp-rocmfp4-turbo
+      - llama-swap.miche/qwen3.6-27b-mtp
       - llama-swap.byte/qwen3.6-27b-mtp
     task: 
-      - llama-swap.miche/qwen3.6-27b-mtp-rocmfp4-turbo
+      - llama-swap.miche/qwen3.6-27b-mtp
       - llama-swap.byte/qwen3.6-27b-mtp
       - zai-coding/glm-5.2
       - minimax-code/MiniMax-M3
     smol: 
-      - llama-swap.miche/qwen3.6-35b-a3b-mtp-rocmfp4
+      - llama-swap.miche/qwen36-35b-genesis
       - llama-swap.byte/qwen36-35b-genesis
-      - llama-swap.miche/qwen3.6-27b-mtp-rocmfp4-turbo
+      - llama-swap.miche/qwen3.6-27b-mtp
       - zai-coding/glm-5.2
       - minimax-code/MiniMax-M3
     plan: 
@@ -45,9 +45,9 @@ retry:
       - deepseek/deepseek-v4-pro
       - llama-swap.miche/step-3.7-flash
     vision: 
-      - llama-swap.miche/gemma4-12b
+      - llama-swap.miche/gemma4-12b-hauhau
       - llama-swap.byte/gemma4-12b-hauhau
-      - llama-swap.miche/qwen3.6-27b-mtp-rocmfp4-turbo
+      - llama-swap.miche/qwen3.6-27b-mtp
 tools: 
   approvalMode: yolo
 memory: