fix(rag): recall complete course homework lists

ai_knowledge_base_v2/apps/backend/app/services/knowledge_agent_service.py

@@ -32,7 +32,8 @@ from app.services.rag_service import PromptService, RagResult, RetrievedChunk
 SAFETY_RULE_VERSION = "minimum-safety-v1"
 MAX_LEXICAL_CANDIDATES = 12
 MAX_SELECTED_SECTIONS = 4
-MAX_OVERVIEW_SELECTED_SECTIONS = 8
+MAX_OVERVIEW_LEXICAL_CANDIDATES = 48
+MAX_OVERVIEW_SELECTED_SECTIONS = 24
 BUSINESS_MARKERS = {"课程", "大本营", "训练营", "老师", "卢慧", "功课", "学员", "课堂", "练习", "觉察", "内在"}
 
 
@@ -105,10 +106,17 @@ class KnowledgeAgentService:
         try:
             if need_knowledge and selected_ids:
                 terms = cls._query_terms(retrieval_question)
-                candidates = cls.search_knowledge(db, terms, selected_ids, catalog)
-                trace.append(cls._trace("search_knowledge", len(trace) + 1, {"queryTerms": terms, "knowledgeIds": selected_ids}, {"candidateCount": len(candidates), "candidates": [cls._candidate_trace(x) for x in candidates]}, started))
-                await cls._rerank(db, retrieval_question, candidates, trace, started)
                 practice_overview = cls._is_practice_overview(retrieval_question)
+                candidate_limit = cls._candidate_limit(retrieval_question)
+                candidates = cls.search_knowledge(
+                    db,
+                    terms,
+                    selected_ids,
+                    catalog,
+                    candidate_limit=candidate_limit,
+                )
+                trace.append(cls._trace("search_knowledge", len(trace) + 1, {"queryTerms": terms, "knowledgeIds": selected_ids, "candidateLimit": candidate_limit}, {"candidateCount": len(candidates), "candidates": [cls._candidate_trace(x) for x in candidates]}, started))
+                await cls._rerank(db, retrieval_question, candidates, trace, started)
                 selected = cls._select_sections(
                     candidates,
                     limit=cls._selection_limit(retrieval_question),
@@ -323,7 +331,15 @@ class KnowledgeAgentService:
         return {"knowledgeId": item.knowledge.id, "knowledgeName": item.knowledge.name, "versionId": item.version.id, "chunkId": item.chunk.id, "sectionId": item.section.id, "title": item.chunk.title, "lexicalScore": item.lexical_score, "rerankScore": item.rerank_score, "selected": item.selected, "discardReason": item.discard_reason}
 
     @classmethod
-    def search_knowledge(cls, db: Session, terms: list[str], selected_ids: list[int], catalog: list[dict]) -> list[Candidate]:
+    def search_knowledge(
+        cls,
+        db: Session,
+        terms: list[str],
+        selected_ids: list[int],
+        catalog: list[dict],
+        *,
+        candidate_limit: int = MAX_LEXICAL_CANDIDATES,
+    ) -> list[Candidate]:
         versions_by_kb = {item["knowledgeId"]: item["versionId"] for item in catalog}
         version_ids = [versions_by_kb[item] for item in selected_ids if item in versions_by_kb]
         if not version_ids:
@@ -349,7 +365,7 @@ class KnowledgeAgentService:
             if score > 0:
                 candidates.append(Candidate(chunk, section, knowledge, version, score))
         candidates.sort(key=lambda item: (item.lexical_score, item.version.published_at or item.version.created_at), reverse=True)
-        return candidates[:MAX_LEXICAL_CANDIDATES]
+        return candidates[:candidate_limit]
 
     @classmethod
     async def _rerank(cls, db: Session, question: str, candidates: list[Candidate], trace: list[dict], started: float) -> None:
@@ -385,6 +401,15 @@ class KnowledgeAgentService:
         prefer_numbered_practice: bool = False,
     ) -> list[Candidate]:
         if prefer_numbered_practice:
+            numbered_candidates = [
+                item for item in candidates
+                if KnowledgeAgentService._is_numbered_practice_title(item.chunk.title)
+            ]
+            if numbered_candidates:
+                for item in candidates:
+                    if not KnowledgeAgentService._is_numbered_practice_title(item.chunk.title):
+                        item.discard_reason = "作业清单优先采用编号练习章节"
+                candidates = numbered_candidates
             candidates = sorted(
                 candidates,
                 key=lambda item: (
@@ -415,6 +440,12 @@ class KnowledgeAgentService:
             return MAX_OVERVIEW_SELECTED_SECTIONS
         return MAX_SELECTED_SECTIONS
 
+    @staticmethod
+    def _candidate_limit(question: str) -> int:
+        if KnowledgeAgentService._is_practice_overview(question):
+            return MAX_OVERVIEW_LEXICAL_CANDIDATES
+        return MAX_LEXICAL_CANDIDATES
+
     @staticmethod
     def _is_practice_overview(question: str) -> bool:
         overview_markers = ("有哪些", "是什么", "包括什么", "都有什么", "列出", "汇总", "总结")