Skip to content

Conversation

@WainWong
Copy link

Summary

  • 支持多轮对话上下文用于表 RAG 选择
  • 在进行表嵌入匹配时,结合历史问题上下文提高表选择准确性
  • 添加配置项控制多轮上下文功能开关和历史记录数量

Changes

  • backend/apps/chat/curd/chat.py: 添加获取历史问题的函数
  • backend/apps/chat/task/llm.py: 传递历史问题到表选择逻辑
  • backend/apps/datasource/crud/datasource.py: 在表嵌入时使用历史问题
  • backend/common/core/config.py: 添加多轮上下文相关配置

- Add MULTI_TURN_EMBEDDING_ENABLED and MULTI_TURN_HISTORY_COUNT config
- Add get_chat_history_questions() to retrieve recent questions from same chat
- Add build_context_query() to concatenate history questions with current question
- Update calc_table_embedding() to use context query for better table matching
- Pass history_questions through get_table_schema() to LLMService

This improves table structure retrieval accuracy by considering the full
conversation context instead of just the latest question.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant