Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-maintaining Wiki.
Go agentagenticaichatbotdsh-pluginembeddingsevaluationgenerative-aigolangknowledge-basellmmulti-tenantollamaopenaiquestion-answeringragrerankingsemantic-searchvector-searchwiki
fix: report embedding dimension mismatches
十三分之五六七2544fedcd356
diff --git a/internal/application/service/knowledgebase_search.go b/internal/application/service/knowledgebase_search.goindex c4f7510d6..2be71595a 100644--- a/internal/application/service/knowledgebase_search.go+++ b/internal/application/service/knowledgebase_search.go@@ -2,6 +2,7 @@ package serviceimport ("context"+ "fmt""github.com/Tencent/WeKnora/internal/application/service/retriever"apperrors "github.com/Tencent/WeKnora/internal/errors"@@ -41,7 +42,38 @@ func (s *knowledgeBaseService) GetQueryEmbedding(ctx context.Context, kbID strinreturn nil, err}- return embeddingModel.Embed(ctx, queryText)+ vector, err := embeddingModel.Embed(ctx, queryText)+ if err != nil {+ return nil, err+ }+ if err := validateQueryEmbeddingDimension(embeddingModel, len(vector)); err != nil {+ logger.Errorf(ctx, "GetQueryEmbedding: %v", err)+ return nil, err+ }+ return vector, nil+}++// validateQueryEmbeddingDimension catches provider/model configuration drift+// before a malformed query reaches the vector store. Without this check, a+// provider returning a different vector size is reported as the misleading+// generic 2201 "bound store unavailable" error (or, for dimension-partitioned+// stores, silently produces no matches).+func validateQueryEmbeddingDimension(model embedding.Embedder, actual int) error {+ if model == nil || actual == 0 {+ return nil+ }+ expected := model.GetDimensions()+ if expected <= 0 || expected == actual {+ return nil+ }+ return apperrors.NewVectorStoreUnavailableError(+ "embedding vector dimension does not match the configured model",+ ).WithDetails(map[string]any{+ "model": model.GetModelName(),+ "expected_dimension": expected,+ "actual_dimension": actual,+ "hint": fmt.Sprintf("check the embedding model configuration and rebuild the affected index (%d dimensions)", expected),+ })}// ResolveEmbeddingModelKeys resolves embedding model IDs to their actual model@@ -479,6 +511,10 @@ func (s *knowledgeBaseService) resolveQueryEmbedding(logger.Errorf(ctx, "Failed to embed query text, query text: %s, error: %v", params.QueryText, err)return nil, err}+ if err := validateQueryEmbeddingDimension(embeddingModel, len(queryEmbedding)); err != nil {+ logger.Errorf(ctx, "resolveQueryEmbedding: %v", err)+ return nil, err+ }logger.Infof(ctx, "Query embedding generated successfully, embedding vector length: %d", len(queryEmbedding))return queryEmbedding, nil}diff --git a/internal/application/service/knowledgebase_search_dimension_test.go b/internal/application/service/knowledgebase_search_dimension_test.gonew file mode 100644index 000000000..d9c30e1e4--- /dev/null+++ b/internal/application/service/knowledgebase_search_dimension_test.go@@ -0,0 +1,43 @@+package service++import (+ "context"+ "testing"++ apperrors "github.com/Tencent/WeKnora/internal/errors"+ "github.com/Tencent/WeKnora/internal/models/embedding"+ "github.com/stretchr/testify/assert"+ "github.com/stretchr/testify/require"+)++type dimensionTestEmbedder struct {+ embedding.Embedder+ name string+ dimensions int+}++func (e dimensionTestEmbedder) GetModelName() string { return e.name }+func (e dimensionTestEmbedder) GetDimensions() int { return e.dimensions }+func (e dimensionTestEmbedder) Embed(context.Context, string) ([]float32, error) {+ return make([]float32, 768), nil+}++func TestValidateQueryEmbeddingDimension(t *testing.T) {+ err := validateQueryEmbeddingDimension(dimensionTestEmbedder{+ name: "configured-embedding",+ dimensions: 1536,+ }, 768)+ require.Error(t, err)+ appErr, ok := err.(*apperrors.AppError)+ require.True(t, ok)+ assert.Equal(t, apperrors.ErrVectorStoreUnavailable, appErr.Code)+ assert.Equal(t, "configured-embedding", appErr.Details.(map[string]any)["model"])+ assert.Equal(t, 1536, appErr.Details.(map[string]any)["expected_dimension"])+ assert.Equal(t, 768, appErr.Details.(map[string]any)["actual_dimension"])+}++func TestValidateQueryEmbeddingDimensionAllowsUnknownOrMatchingSize(t *testing.T) {+ assert.NoError(t, validateQueryEmbeddingDimension(dimensionTestEmbedder{dimensions: 768}, 768))+ assert.NoError(t, validateQueryEmbeddingDimension(dimensionTestEmbedder{dimensions: 0}, 768))+ assert.NoError(t, validateQueryEmbeddingDimension(dimensionTestEmbedder{dimensions: 1536}, 0))+}