Let Ollama own model residency

web/README.md

@@ -38,7 +38,7 @@ https://lab.k1412.top/
 - Search Agent 查询理解：`ChatGPT-5.6:light`
 - Search Agent 结果简报、搜索追问、论文对话：`ChatGPT-5.6:fast`
 - Ollama 请求单并发执行，并缓存结果到 `web/cache/ai/`
-- 应用侧不下发 `num_ctx`，上下文长度由 Ollama 服务端 runner 统一控制
+- 应用侧不下发 `num_ctx` / `keep_alive`，上下文长度和模型常驻由 Ollama 服务端统一控制
 - arXiv 摘要缓存到 `web/cache/arxiv/`
 
 ## Knowledge Design

web/app.py

@@ -1236,7 +1236,6 @@ def call_ollama(model: str, prompt: str, mode: str) -> dict[str, Any]:
         "model": model,
         "prompt": prompt,
         "stream": False,
-        "keep_alive": "2m",
         "options": options,
     }
     data = json.dumps(payload, ensure_ascii=False).encode("utf-8")