fix(gateway): 修复 Qwen3.8 推理参数并增加上游自愈

统一 Qwen3.8 固定思考、推理档位、预算互斥与温度下限规则，并覆盖 Chat Completions 和原生 Responses 请求。新增安全参数纠正、有限重试和并发安全的进程内 LRU 学习；补充共享行为向量、httptest 回归与真实模型验收用例。验证：go test ./... -count=1 通过，gofmt 检查通过；真实 Qwen3.8 流式、非流式及缓存重学习验收通过。
2026-08-01 22:50:45 +08:00
parent 50a13de70b
commit 9c093974a1
9 changed files with 1547 additions and 33 deletions
@@ -0,0 +1,10 @@
+[
+  { "name": "none-to-low", "input": { "reasoning_effort": "none" }, "expected": { "enable_thinking": true, "reasoning_effort": "low" } },
+  { "name": "minimal-to-low", "input": { "reasoning_effort": "minimal" }, "expected": { "enable_thinking": true, "reasoning_effort": "low" } },
+  { "name": "medium", "input": { "reasoning_effort": "medium" }, "expected": { "enable_thinking": true, "reasoning_effort": "medium" } },
+  { "name": "high-to-xhigh", "input": { "reasoning_effort": "high" }, "expected": { "enable_thinking": true, "reasoning_effort": "xhigh" } },
+  { "name": "max-to-xhigh", "input": { "reasoning_effort": "max" }, "expected": { "enable_thinking": true, "reasoning_effort": "xhigh" } },
+  { "name": "force-thinking", "input": { "enable_thinking": false }, "expected": { "enable_thinking": true, "reasoning_effort": "low" } },
+  { "name": "budget-wins", "input": { "reasoning_effort": "high", "thinking_budget": 300000 }, "expected": { "enable_thinking": true, "thinking_budget": 262144 } },
+  { "name": "temperature-floor", "input": { "temperature": 0.1 }, "expected": { "enable_thinking": true, "temperature": 0.6 } }
+]