feat(llm): update model and reasoning parameters

api/main.py

@@ -149,8 +149,10 @@ async def chat_stream_endpoint(request: Request):
     payload["messages"] = messages
     payload.pop("user_id", None)
     payload.pop("include_audio", None)
-    if not payload.get("model"):
-        payload["model"] = OPENAI_LLM_MODEL
+    payload["model"] = OPENAI_LLM_MODEL
+    payload["thinking"] = {"type": "disabled"}
+    payload["reasoning"] = {"effort": "minimal"}
+    payload["max_tokens"] = 200
     payload["stream"] = True
 
     base_url = OPENAI_API_BASE_URL_CONFIG or ""

config/config.py

@@ -30,7 +30,7 @@ OPENAI_EMBEDDING_MODEL = "doubao-embedding-large-text-250515"
 OPENAI_EMBEDDING_DIM = 2048
 OPENAI_EMBEDDING_KEY = os.getenv("DOUBAO_API_KEY", "YOUR_API_KEY_PLACEHOLDER_IF_NOT_IN_ENV")
 OPENAI_EMBEDDING_BASE = "https://ark.cn-beijing.volces.com/api/v3"
-OPENAI_LLM_MODEL = "doubao-seed-1-6-251015"
+OPENAI_LLM_MODEL = "doubao-seed-1-8-251228"
 
 # ---huggingface模型---
 HUGGINGFACE_KEY = "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"
@@ -63,7 +63,7 @@ MEM0_CONFIG = {
         "provider": "openai",
         "config": {
             "api_key": OPENAI_API_KEY_FROM_CONFIG,
-            "model": "doubao-seed-1-6-250615",
+            "model": OPENAI_LLM_MODEL,
             "openai_base_url": OPENAI_API_BASE_URL_CONFIG
         }
     },

memory_module/README.md

@@ -54,7 +54,7 @@ MEM0_CONFIG = {
         "provider": "openai",
         "config": {
             "api_key": OPENAI_API_KEY_FROM_CONFIG,
-            "model": "doubao-seed-1-6-250615",
+            "model": "doubao-seed-1-8-251228",
             "openai_base_url": OPENAI_API_BASE_URL_CONFIG
         }
     },

memory_module/memory_integration.py

@@ -417,7 +417,9 @@ class LocalMemoryIntegration:
                     {"role": "system", "content": system_prompt},
                     {"role": "user", "content": user_input}
                 ],
-                # reasoning_effort="minimal", # 你原始代码里的，如果豆包不支持就删掉
+                thinking={"type": "disabled"},
+                reasoning={"effort": "minimal"},
+                max_tokens=200,
                 # 推荐开启 JSON 模式，让豆包强制输出 JSON
                 response_format={"type": "json_object"} 
             )