Stop overriding Ollama context size

web/README.md

@@ -38,6 +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
 - arXiv 摘要缓存到 `web/cache/arxiv/`
 
 ## Knowledge Design

web/app.py

@@ -1219,19 +1219,18 @@ def clean_model_response(value: str) -> str:
 def call_ollama(model: str, prompt: str, mode: str) -> dict[str, Any]:
     options = {
         "temperature": 0.2,
-        "num_ctx": 4096,
         "num_predict": 520,
     }
     if mode == "translate":
-        options.update({"num_ctx": 3072, "num_predict": 700, "temperature": 0.1})
+        options.update({"num_predict": 700, "temperature": 0.1})
     elif mode == "deep":
-        options.update({"num_ctx": 8192, "num_predict": 900, "temperature": 0.25})
+        options.update({"num_predict": 900, "temperature": 0.25})
     elif mode in {"atlas", "topic", "path", "compare"}:
-        options.update({"num_ctx": 8192, "num_predict": 950, "temperature": 0.22})
+        options.update({"num_predict": 950, "temperature": 0.22})
     elif mode == "search_plan":
-        options.update({"num_ctx": 4096, "num_predict": 420, "temperature": 0.1})
+        options.update({"num_predict": 420, "temperature": 0.1})
     elif mode in {"search_summary", "search_followup", "paper_chat"}:
-        options.update({"num_ctx": 8192, "num_predict": 780, "temperature": 0.18})
+        options.update({"num_predict": 780, "temperature": 0.18})
 
     payload = {
         "model": model,