feat: LLM 切换为 DeepSeek deepseek-flash

This commit is contained in:
2026-10-08 12:29:41 +08:00
parent 098b119655
commit 7c751bfc02
2 changed files with 6 additions and 6 deletions
+3 -3
View File
@@ -11,7 +11,7 @@
- 运行时:Node.js 20+
- 框架:Fastify 5
- LLM:SiliconFlow OpenAI 兼容接口(Qwen/Qwen2.5-32B-Instruct)
- LLM:DeepSeek OpenAI 兼容接口(deepseek-flash / DeepSeek-V4.1-Flash)
- 运行时依赖:除 fastify 外无其他外部依赖
## 项目结构
@@ -60,9 +60,9 @@ lpt-ai/src/
## LLM 配置
```env
LLM_API_URL=https://api.siliconflow.cn/v1/chat/completions
LLM_API_URL=https://api.deepseek.com/v1/chat/completions
LLM_API_KEY= # 必填,未配置时 /health 返回 llmAvailable: false
LLM_MODEL=Qwen/Qwen2.5-32B-Instruct
LLM_MODEL=deepseek-flash
LLM_TIMEOUT_MS=60000 # 单次 LLM 调用超时
PORT=5199
```