fix: improve AI coach chat context, 128k window, and output limits

Co-authored-by: Cursor <cursoragent@cursor.com>

ai_client.py

@@ -104,7 +104,13 @@ def _openai_chat_url() -> str:
     return f"{base}/chat/completions"
 
 
-def _generate_openai(prompt: str, images: List[tuple], temperature: float) -> str:
+def _generate_openai(
+    prompt: str,
+    images: List[tuple],
+    temperature: float,
+    *,
+    max_tokens: int | None = None,
+) -> str:
     api_key = _openai_api_key()
     if not api_key:
         return "AI 调用失败：未配置 OPENAI_API_KEY（请在当前实例目录 .env 中设置，修改后需重启服务）"
@@ -124,12 +130,14 @@ def _generate_openai(prompt: str, images: List[tuple], temperature: float) -> st
         messages = [{"role": "user", "content": content}]
     else:
         messages = [{"role": "user", "content": prompt}]
-    body = {
+    body: dict = {
         "model": _openai_model(),
         "messages": messages,
         "temperature": temperature,
         "stream": False,
     }
+    if max_tokens is not None and max_tokens > 0:
+        body["max_tokens"] = int(max_tokens)
     r = requests.post(
         _openai_chat_url(),
         headers=headers,
@@ -145,12 +153,21 @@ def _generate_openai(prompt: str, images: List[tuple], temperature: float) -> st
     return (msg.get("content") or "").strip() or "AI 生成失败：空内容"
 
 
-def _generate_ollama(prompt: str, images: List[tuple], temperature: float) -> str:
+def _generate_ollama(
+    prompt: str,
+    images: List[tuple],
+    temperature: float,
+    *,
+    max_tokens: int | None = None,
+) -> str:
+    options: dict = {"temperature": temperature}
+    if max_tokens is not None and max_tokens > 0:
+        options["num_predict"] = int(max_tokens)
     payload = {
         "model": _ollama_model(),
         "prompt": prompt,
         "stream": False,
-        "options": {"temperature": temperature},
+        "options": options,
     }
     if images:
         payload["images"] = [b64 for b64, _mime in images]
@@ -165,13 +182,14 @@ def ai_generate(
     image_paths: Optional[Sequence[str]] = None,
     images_b64: Optional[Sequence[str]] = None,
     temperature: float = 0.2,
+    max_tokens: int | None = None,
 ) -> str:
     """统一文本生成；失败时返回以「AI 调用失败」开头的说明。"""
     images = _collect_images(image_paths, images_b64)
     try:
         if _use_openai():
-            return _generate_openai(prompt, images, temperature)
-        return _generate_ollama(prompt, images, temperature)
+            return _generate_openai(prompt, images, temperature, max_tokens=max_tokens)
+        return _generate_ollama(prompt, images, temperature, max_tokens=max_tokens)
     except requests.HTTPError as e:
         detail = ""
         try: