Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

fix(model_meta): discover TEI reranker models (#19210)

dawn70601910ef24
diff --git a/rag/llm/model_meta.py b/rag/llm/model_meta.py
index f4eb955122..aafa5dad9b 100644
--- a/rag/llm/model_meta.py
+++ b/rag/llm/model_meta.py
@@ -1063,7 +1063,7 @@ class HuggingFace(Base):
if isinstance(model_type, dict):
if "embedding" in model_type:
return [LLMType.EMBEDDING.value]
- if "rerank" in model_type:
+ if "reranker" in model_type:
return [LLMType.RERANK.value]
return []
# TGI format: "text-generation" / "text2text-generation"
diff --git a/test/unit_test/rag/llm/test_model_meta_huggingface.py b/test/unit_test/rag/llm/test_model_meta_huggingface.py
new file mode 100644
index 0000000000..7d3101b3e7
--- /dev/null
+++ b/test/unit_test/rag/llm/test_model_meta_huggingface.py
@@ -0,0 +1,77 @@
+#
+# Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import pytest
+
+from common.constants import LLMType
+from rag.llm.model_meta import HuggingFace
+
+pytestmark = pytest.mark.p2
+
+
+ ("model_info", "expected_type", "expected_max_tokens"),
+ [
+ pytest.param(
+ {
+ "model_id": "/data/gte-rerank-int8",
+ "model_type": {
+ "reranker": {
+ "id2label": {"0": "LABEL_0"},
+ "label2id": {"LABEL_0": 0},
+ }
+ },
+ "max_input_length": 2048,
+ "version": "1.9.3",
+ },
+ LLMType.RERANK.value,
+ 2048,
+ id="tei-reranker",
+ ),
+ pytest.param(
+ {
+ "model_id": "/data/qwen3-embed-int8",
+ "model_type": {"embedding": {"pooling": "last_token"}},
+ "max_input_length": 2048,
+ "version": "1.9.3",
+ },
+ LLMType.EMBEDDING.value,
+ 2048,
+ id="tei-embedding",
+ ),
+ pytest.param(
+ {
+ "model_id": "meta-llama/Llama-2-7b-chat-hf",
+ "model_type": "text-generation",
+ "max_total_tokens": 4096,
+ },
+ LLMType.CHAT.value,
+ 4096,
+ id="tgi-text-generation",
+ ),
+ ],
+)
+def test_huggingface_formats_discovered_model(model_info, expected_type, expected_max_tokens):
+ provider = HuggingFace(api_key="", base_url="http://inference-server")
+
+ assert provider._format_model_list(model_info) == [
+ {
+ "name": model_info["model_id"],
+ "model_types": [expected_type],
+ "features": [],
+ "max_tokens": expected_max_tokens,
+ }
+ ]