refactor(backend): update services and register OCR router

- router.py: register ocr_router with OCR tag
- ontology.py: update ontology service logic
- orchestrator.py: update orchestrator service logic
- user_agent.py: update user agent schema and service
- schemas/user_agent.py: update user agent data schemas
This commit is contained in:
caoxiaozhu
2026-05-12 03:03:15 +00:00
parent 33826929ba
commit ca29025063
5 changed files with 56 additions and 14 deletions

View File

@@ -666,6 +666,8 @@ class SemanticOntologyService:
"entry_source": payload.context_json.get("entry_source"),
"attachment_names": payload.context_json.get("attachment_names", []),
"attachment_count": payload.context_json.get("attachment_count", 0),
"ocr_summary": payload.context_json.get("ocr_summary", ""),
"ocr_documents": payload.context_json.get("ocr_documents", []),
"request_context": payload.context_json.get("request_context"),
"role_codes": payload.context_json.get("role_codes", []),
},
@@ -690,6 +692,7 @@ class SemanticOntologyService:
"即使没有明确说“生成草稿”,也优先使用 expense + draft。"
"出现“客户”不等于应收,出现“供应商”不等于应付,必须结合动作词和业务目标判断。"
"只有明确查询、统计、列出、多少、明细、对比时才优先使用 query 或 compare。"
"附件名称和 OCR 摘要只作为辅助证据,不能编造未出现的事实。"
"信息不足时 clarification_required=true并给出一句简短中文追问。"
"missing_slots 使用简短 snake_case例如 expense_type, amount, "
"customer_name, participants, attachments。"