refactor and support for multi algs fusion

.claude/skills/adapt-new-diffusion-model/SKILL.md

@@ -52,15 +52,12 @@ def is_diffusion_model(model_or_path):
 ```
 
 If your model doesn't have `model_index.json`, either create one in the model
-directory or pass diffusion-specific options through new-architecture
-`ExtraConfig` / `AutoRound` kwargs:
+directory or pass diffusion-specific options through `AutoRound` kwargs:
 
 ```python
-from auto_round.compressors.config import ExtraConfig
-
 ar = AutoRound(
     model,
-    extra_config=ExtraConfig(num_inference_steps=5),
+    num_inference_steps=5,
 )
 ```
 

.claude/skills/add-vlm-model/SKILL.md

@@ -104,7 +104,7 @@ The new architecture routes multimodal calibration through:
 
 If your model works with an existing template/processor, prefer passing
 `template=...`, `processor=...`, or `image_processor=...` through `AutoRound` /
-`ExtraConfig` instead of adding compressor code.
+kwargs instead of adding compressor code.
 
 ## Step 3: Add Calibration Template
 

.gitignore

@@ -12,3 +12,4 @@ tmp_autoround/
 ut_log_dir/
 CLAUDE.local.md
 docs/plan/
+.codegraph/

AGENTS.md

@@ -71,3 +71,10 @@ Test fixtures create tiny models (OPT-125M, Qwen-0.6B) at session scope — firs
 - `setup.py` forces `CC=CXX=g++` at import time
 - Version is computed dynamically from git tags — untagged commits produce dev versions
 - Some test dependencies (AutoAWQ, GPTQModel, llama-cpp) require manual git installs — see comments in `test/test_cuda/requirements.txt`
+
+## AI 行为准则
+
+1. **Think Before Coding** — 改动前先复述目标、找到相关代码位置、说出预期影响，再写。
+2. **Simplicity First** — 优先用最简方案。能改 3 行不改 30 行；能复用不新建。
+3. **Surgical Changes** — 只动跟当前任务相关的代码。看到不顺眼但无关的代码，记下来不改。
+4. **Goal-Driven Execution** — 完成后必须用可观测方式验证（运行测试 / 打印结果 / 跑示例），不靠"我觉得对了"。