cnphpbb/deploy.stack
1
0
Fork 0
forked from DevOps/deploy.stack
Code Pull Requests Activity

fix(honcho): override deduction/induction specialist model to deepseek/siliconflow

Browse Source 
honcho-deriver dreamer 模块的 deduction/induction specialist 有独立的
model config（src/config.py:1189-1208），默认值 hardcode gpt-5.4-mini +
openai SDK 默认 endpoint（api.openai.com/v1）。原 stack.yml 只覆盖了
DERIVER_MODEL_CONFIG 和 EMBEDDING_MODEL_CONFIG，导致 dream cycle 在
国内持续 timeout。

- stack.yml deriver 段显式注入 DEDUCTION_MODEL_CONFIG__* /
  INDUCTION_MODEL_CONFIG__*（model=deepseek-ai/DeepSeek-V4-Flash，
  base_url=https://api.siliconflow.cn/v1）
- DERIVER_MODEL_CONFIG__OVERRIDES__BASE_URL 同步加上
- env.cfg.example: 同步 HONCHO_DEDUCTION_MODEL /
  HONCHO_INDUCTION_MODEL 变量；HONCHO_DERIVER_MODEL 默认值从 gpt-5.4-mini
  改为 deepseek-ai/DeepSeek-V4-Flash；HONCHO_LLM_BASE_URL 默认从 openai.com
  改为 siliconflow.cn

诊断依据：vm61 docker inspect honcho-deriver + logs 显示
src.dreamer.orchestrator 仍在请求 openai/gpt-5.4-mini。
.env.*
This commit is contained in:
cnphpbb
2026-06-19 21:21:46 +08:00
parent 7309ec1261
commit 8b07e76d67
2 changed files with 15 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
honcho/stack.yml
+8 -1
View File
@@ -93,7 +93,14 @@ services:
- CACHE_ENABLED=true
- NAMESPACE=${HONCHO_NAMESPACE:-honcho}
- DERIVER_MODEL_CONFIG__TRANSPORT=openai
- DERIVER_MODEL_CONFIG__MODEL=${HONCHO_DERIVER_MODEL:-gpt-5.4-mini}
- DERIVER_MODEL_CONFIG__MODEL=${HONCHO_DERIVER_MODEL:-deepseek-ai/DeepSeek-V4-Flash}
- DERIVER_MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_LLM_BASE_URL:-https://api.siliconflow.cn/v1}
- DEDUCTION_MODEL_CONFIG__TRANSPORT=openai
- DEDUCTION_MODEL_CONFIG__MODEL=${HONCHO_DEDUCTION_MODEL:-deepseek-ai/DeepSeek-V4-Flash}
- DEDUCTION_MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_LLM_BASE_URL:-https://api.siliconflow.cn/v1}
- INDUCTION_MODEL_CONFIG__TRANSPORT=openai
- INDUCTION_MODEL_CONFIG__MODEL=${HONCHO_INDUCTION_MODEL:-deepseek-ai/DeepSeek-V4-Flash}
- INDUCTION_MODEL_CONFIG__OVERRIDES__BASE_URL=${HONCHO_LLM_BASE_URL:-https://api.siliconflow.cn/v1}
- DERIVER_WORKERS=${HONCHO_DERIVER_WORKERS:-1}
- EMBEDDING_MODEL_CONFIG__TRANSPORT=openai
- EMBEDDING_MODEL_CONFIG__MODEL=${HONCHO_EMBEDDING_MODEL:-Qwen/Qwen3-Embedding-8B}