[Question]: Clarification Needed on Top N and Top K Parameters in Chat->Prompt Engine Configuration

[maqy1995]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

confused with Top N and Top K parameter.
Based on the configuration of Chat Configuration-> Prompt Engine, Top N means number of chunks from the retriever, Top K which associated with Rerank model means K chunk will be sent into the rerank model.
However, from the code below, it appears that retrieved chunks numbers is min(RERANK_LIMIT = 64 , Top K), and all retrieved chunks will be fed into rerank model?

ragflow/api/db/services/dialog_service.py

Lines 212 to 226 in b77ce4e

	kbinfos = retriever.retrieval(
	" ".join(questions),
	embd_mdl,
	tenant_ids,
	dialog.kb_ids,
	1,
	dialog.top_n,
	dialog.similarity_threshold,
	dialog.vector_similarity_weight,
	doc_ids=attachments,
	top=dialog.top_k,
	aggs=False,
	rerank_mdl=rerank_mdl,
	rank_feature=label_question(" ".join(questions), kbs),
	)

and here TOP_N was assigned to page_size, :

ragflow/rag/nlp/search.py

Lines 342 to 367 in 200b6f5

	def retrieval(self, question, embd_mdl, tenant_ids, kb_ids, page, page_size, similarity_threshold=0.2,
	vector_similarity_weight=0.3, top=1024, doc_ids=None, aggs=True,
	rerank_mdl=None, highlight=False,
	rank_feature: dict | None = {PAGERANK_FLD: 10}):
	ranks = {"total": 0, "chunks": [], "doc_aggs": {}}
	if not question:
	return ranks
	RERANK_LIMIT = 64
	req = {"kb_ids": kb_ids, "doc_ids": doc_ids, "page": page, "size": RERANK_LIMIT,
	"question": question, "vector": True, "topk": top,
	"similarity": similarity_threshold,
	"available_int": 1}
	if isinstance(tenant_ids, str):
	tenant_ids = tenant_ids.split(",")
	sres = self.search(req, [index_name(tid) for tid in tenant_ids],
	kb_ids, embd_mdl, highlight, rank_feature=rank_feature)
	ranks["total"] = sres.total
	if rerank_mdl and sres.total > 0:
	sim, tsim, vsim = self.rerank_by_model(rerank_mdl,
	sres, question, 1 - vector_similarity_weight,
	vector_similarity_weight,
	rank_feature=rank_feature)

similar question about retrieve_testing module:#4797

[KevinHuSh]

Top K is refered by ES KNN ranking procedure.