From abc588daac6053ec2edbdcde3f5a2dc5eb7d50b8 Mon Sep 17 00:00:00 2001 From: sigoden Date: Fri, 21 Jun 2024 06:00:26 +0800 Subject: feat: support rerank (#620) --- config.example.yaml | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'config.example.yaml') diff --git a/config.example.yaml b/config.example.yaml index ead96f4..15877b3 100644 --- a/config.example.yaml +++ b/config.example.yaml @@ -35,16 +35,20 @@ bots: # Specifies the embedding model to use rag_embedding_model: null +# Specifies the rerank model to use +rag_rerank_model: null # Specifies the chunk size rag_chunk_size: null # Specifies the chunk overlap rag_chunk_overlap: null # Specifies the number of documents to retrieve rag_top_k: 4 -# Specifies the minimum relevance score for vector search -rag_min_score_vector: 0 -# Specifies the minimum relevance score for full-text search -rag_min_score_text: 0 +# Specifies the minimum relevance score for vector searching +rag_min_score_vector_search: 0 +# Specifies the minimum relevance score for full-text searching +rag_min_score_fulltext_search: 0 +# Specifies the minimum relevance score for reranking +rag_min_score_rerank: 0 # Defines the query structure using variables like __CONTEXT__ and __INPUT__ to tailor searches to specific needs rag_template: | @@ -88,6 +92,9 @@ clients: # max_input_tokens: 2048 # default_chunk_size: 2000 # max_concurrent_chunks: 100 + # - name: xxxx + # mode: rerank # Rerank model + # max_input_tokens: 2048 # patches: # : # The regex to match model names, e.g. '.*' 'gpt-4o' 'gpt-4o|gpt-4-.*' # chat_completions_body: # The JSON to be merged with the chat completions request body. -- cgit v1.2.3