From 789610d01f1a45c139d5f6fbc3563be1f762a0a8 Mon Sep 17 00:00:00 2001
From: lacerate551 <128470311+lacerate551@users.noreply.github.com>
Date: Sun, 21 Jun 2026 23:30:31 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20INTENT=5FMAX=5FTOKENS=20=E4=BB=8E=201024?=
 =?UTF-8?q?=20=E6=8F=90=E5=8D=87=E8=87=B3=202048=EF=BC=8C=E9=80=82?=
 =?UTF-8?q?=E9=85=8D=E6=8E=A8=E7=90=86=E6=A8=A1=E5=9E=8B=E6=80=9D=E8=80=83?=
 =?UTF-8?q?=E9=93=BE=E9=A2=84=E7=AE=97?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

mimo-v2.5 推理模型思考链消耗 ~800-1000 tokens，原 1024 不够导致
content 为空、全部输出进入 reasoning_content，引发 JSON 数组格式
解析错误。2048 可确保思考链 + JSON 输出均有足够空间。
---
 config.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/config.py b/config.py
index c172567..c6d12ab 100644
--- a/config.py
+++ b/config.py
@@ -105,7 +105,7 @@ LLM_MAX_TOKENS = 3000            # 最大输出 token 数（推理模型思考
 # ----- 意图分析（轻量、确定性高）-----
 # INTENT_MODEL 在顶部「一、API 密钥与模型」中统一配置
 INTENT_TEMPERATURE = 0.1
-INTENT_MAX_TOKENS = 1024        # 非推理模型无需思考链预算，JSON 输出 1024 足够
+INTENT_MAX_TOKENS = 2048        # 推理模型需思考链预算（~1000 tokens），JSON 输出 ~200 tokens
 INTENT_HISTORY_WINDOW = 6        # 分析时取最近几条历史消息
 
 # ==============================================================================