From 4e992de91fdf7f587ddcfb779ff257ad9b66cdc2 Mon Sep 17 00:00:00 2001
From: "rhinoceros.xn" <rhinoceros.xn@gmail.com>
Date: Mon, 20 Apr 2026 11:39:17 +0800
Subject: [PATCH] Add tongyi gte-rerank-v2 (#14215)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

https://bailian.console.aliyun.com/cn-beijing?tab=api#/api/?type=model&url=2780056

### What problem does this PR solve?

_Briefly describe what this PR aims to solve. Include background context
that will help reviewers understand the purpose of the PR._

### Type of change
- [x] Other (please describe): add gte-rerank-v2、qwen3-rerank
---
 conf/llm_factories.json | 12 ++++++++++++
 rag/llm/rerank_model.py | 16 ++++++++++++++--
 2 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/conf/llm_factories.json b/conf/llm_factories.json
index b6d2b7ff18..0cadfe3679 100644
--- a/conf/llm_factories.json
+++ b/conf/llm_factories.json
@@ -790,6 +790,18 @@
                     "max_tokens": 1000000,
                     "model_type": "chat",
                     "is_tools": true
+                },
+                {
+                    "llm_name": "gte-rerank-v2",
+                    "tags": "RE-RANK,4k",
+                    "max_tokens": 4000,
+                    "model_type": "rerank"
+                },
+                {
+                    "llm_name": "qwen3-rerank",
+                    "tags": "RE-RANK,4k",
+                    "max_tokens": 4000,
+                    "model_type": "rerank"
                 }
             ]
         },
diff --git a/rag/llm/rerank_model.py b/rag/llm/rerank_model.py
index 5002fe7651..6730261ea7 100644
--- a/rag/llm/rerank_model.py
+++ b/rag/llm/rerank_model.py
@@ -375,7 +375,19 @@ class QWenRerank(Base):
 
         import dashscope
 
-        resp = dashscope.TextReRank.call(api_key=self.api_key, model=self.model_name, query=query, documents=texts, top_n=len(texts), return_documents=False)
+        # qwen3-rerank does not support return_documents parameter  
+        if self.model_name.startswith("qwen3-rerank"):  
+            resp = dashscope.TextReRank.call(  
+                api_key=self.api_key, model=self.model_name,  
+                query=query, documents=texts, top_n=len(texts)  
+            )  
+        else:  
+            resp = dashscope.TextReRank.call(  
+                api_key=self.api_key, model=self.model_name,  
+                query=query, documents=texts,  
+                top_n=len(texts), return_documents=False  
+            )  
+
         rank = np.zeros(len(texts), dtype=float)
         if resp.status_code == HTTPStatus.OK:
             try:
@@ -549,4 +561,4 @@ class RAGconRerank(Base):
 
         rank = Base._normalize_rank(rank)
 
-        return rank, token_count
\ No newline at end of file
+        return rank, token_count