Model Runtime (#1858)

Co-authored-by: StyleZhang <jasonapring2015@outlook.com> Co-authored-by: Garfield Dai <dai.hai@foxmail.com> Co-authored-by: chenhe <guchenhe@gmail.com> Co-authored-by: jyong <jyong@dify.ai> Co-authored-by: Joel <iamjoel007@gmail.com> Co-authored-by: Yeuoly <admin@srmxy.cn>
2024-01-02 23:42:00 +08:00
parent e91dd28a76
commit d069c668f8
807 changed files with 171310 additions and 23806 deletions
--- a/api/core/rerank/rerank.py
+++ b/api/core/rerank/rerank.py
@@ -0,0 +1,58 @@
+from typing import List, Optional
+
+from langchain.schema import Document
+
+from core.model_manager import ModelInstance
+
+
+class RerankRunner:
+    def __init__(self, rerank_model_instance: ModelInstance) -> None:
+        self.rerank_model_instance = rerank_model_instance
+
+    def run(self, query: str, documents: List[Document], score_threshold: Optional[float] = None,
+            top_n: Optional[int] = None, user: Optional[str] = None) -> List[Document]:
+        """
+        Run rerank model
+        :param query: search query
+        :param documents: documents for reranking
+        :param score_threshold: score threshold
+        :param top_n: top n
+        :param user: unique user id if needed
+        :return:
+        """
+        docs = []
+        doc_id = []
+        unique_documents = []
+        for document in documents:
+            if document.metadata['doc_id'] not in doc_id:
+                doc_id.append(document.metadata['doc_id'])
+                docs.append(document.page_content)
+                unique_documents.append(document)
+
+        documents = unique_documents
+
+        rerank_result = self.rerank_model_instance.invoke_rerank(
+            query=query,
+            docs=docs,
+            score_threshold=score_threshold,
+            top_n=top_n,
+            user=user
+        )
+
+        rerank_documents = []
+
+        for result in rerank_result.docs:
+            # format document
+            rerank_document = Document(
+                page_content=result.text,
+                metadata={
+                    "doc_id": documents[result.index].metadata['doc_id'],
+                    "doc_hash": documents[result.index].metadata['doc_hash'],
+                    "document_id": documents[result.index].metadata['document_id'],
+                    "dataset_id": documents[result.index].metadata['dataset_id'],
+                    'score': result.score
+                }
+            )
+            rerank_documents.append(rerank_document)
+
+        return rerank_documents