Skip to content

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

fix(embedder): add code model dimensions and actionable 422 error for Jina (#928)

Deepak Dev Panwardf8ba978461a
diff --git a/openviking/models/embedder/jina_embedders.py b/openviking/models/embedder/jina_embedders.py
index 49e00b2b3..84713ba21 100644
--- a/openviking/models/embedder/jina_embedders.py
+++ b/openviking/models/embedder/jina_embedders.py
@@ -15,6 +15,8 @@ from openviking.models.embedder.base import (
JINA_MODEL_DIMENSIONS = {
"jina-embeddings-v5-text-small": 1024, # 677M params, max seq 32768
"jina-embeddings-v5-text-nano": 768, # 239M params, max seq 8192
+ "jina-code-embeddings-1.5b": 1024, # code model, max seq 8192
+ "jina-code-embeddings-0.5b": 768, # code model, max seq 8192
}
DEFAULT_JINA_QUERY_TASK = "retrieval.query"
@@ -140,6 +142,16 @@ class JinaDenseEmbedder(DenseEmbedderBase):
extra_body["late_chunking"] = self.late_chunking
return extra_body if extra_body else None
+ def _raise_task_error(self, error: openai.APIError) -> None:
+ """Raise an actionable error if a 422 indicates an invalid task type."""
+ if getattr(error, "status_code", None) == 422 and "task" in str(error.body):
+ raise RuntimeError(
+ f"Jina API rejected task type for model '{self.model_name}'. "
+ f"This usually means the model requires a different task prefix. "
+ f"Set 'query_param' and 'document_param' in your embedding config "
+ f"to a valid task type for this model. API details: {error.message}"
+ ) from error
+
def embed(self, text: str, is_query: bool = False) -> EmbedResult:
"""Perform dense embedding on text
@@ -167,6 +179,7 @@ class JinaDenseEmbedder(DenseEmbedderBase):
return EmbedResult(dense_vector=vector)
except openai.APIError as e:
+ self._raise_task_error(e)
raise RuntimeError(f"Jina API error: {e.message}") from e
except Exception as e:
raise RuntimeError(f"Embedding failed: {str(e)}") from e
@@ -200,6 +213,7 @@ class JinaDenseEmbedder(DenseEmbedderBase):
return [EmbedResult(dense_vector=item.embedding) for item in response.data]
except openai.APIError as e:
+ self._raise_task_error(e)
raise RuntimeError(f"Jina API error: {e.message}") from e
except Exception as e:
raise RuntimeError(f"Batch embedding failed: {str(e)}") from e
diff --git a/tests/unit/test_jina_embedder.py b/tests/unit/test_jina_embedder.py
index d34547193..e631062f6 100644
--- a/tests/unit/test_jina_embedder.py
+++ b/tests/unit/test_jina_embedder.py
@@ -324,3 +324,63 @@ class TestJinaDenseEmbedder:
dimension=256,
)
assert embedder.get_dimension() == 256
+
+ @patch("openviking.models.embedder.jina_embedders.openai.OpenAI")
+ def test_422_task_error_actionable_message(self, mock_openai_class):
+ """422 error mentioning 'task' should produce actionable RuntimeError."""
+ import openai
+
+ mock_client = MagicMock()
+ mock_openai_class.return_value = mock_client
+
+ error = openai.BadRequestError(
+ message="Validation error",
+ response=MagicMock(status_code=422, headers={}),
+ body={"detail": "Input should be 'nl2code.query' ... task"},
+ )
+ mock_client.embeddings.create.side_effect = error
+
+ embedder = JinaDenseEmbedder(
+ model_name="jina-code-embeddings-1.5b",
+ api_key="test-key",
+ )
+
+ with pytest.raises(RuntimeError, match="query_param.*document_param"):
+ embedder.embed("hello")
+
+ @patch("openviking.models.embedder.jina_embedders.openai.OpenAI")
+ def test_non_422_error_passthrough(self, mock_openai_class):
+ """Non-422 API errors should use the generic 'Jina API error' message."""
+ import openai
+
+ mock_client = MagicMock()
+ mock_openai_class.return_value = mock_client
+
+ error = openai.APIError(
+ message="Internal server error",
+ request=MagicMock(),
+ body=None,
+ )
+ mock_client.embeddings.create.side_effect = error
+
+ embedder = JinaDenseEmbedder(
+ model_name="jina-embeddings-v5-text-small",
+ api_key="test-key",
+ )
+
+ with pytest.raises(RuntimeError, match="Jina API error"):
+ embedder.embed("hello")
+
+ def test_code_model_dimensions(self):
+ """Code models should have correct default dimensions."""
+ embedder_1_5b = JinaDenseEmbedder(
+ model_name="jina-code-embeddings-1.5b",
+ api_key="test-key",
+ )
+ assert embedder_1_5b.get_dimension() == 1024
+
+ embedder_0_5b = JinaDenseEmbedder(
+ model_name="jina-code-embeddings-0.5b",
+ api_key="test-key",
+ )
+ assert embedder_0_5b.get_dimension() == 768