Skip to content

Open Source alternative to Algolia + Pinecone and an Easier-to-Use alternative to ElasticSearch ⚡ 🔍 ✨ Fast, typo tolerant, in-memory fuzzy Search Engine for building delightful search experiences

Review changes II

ozanarmagane1d452b713d2
diff --git a/include/field.h b/include/field.h
index 9aae16c8..eca39f1e 100644
--- a/include/field.h
+++ b/include/field.h
@@ -50,7 +50,7 @@ namespace fields {
static const std::string vec_dist = "vec_dist";
static const std::string reference = "reference";
static const std::string create_from = "create_from";
- static const std::string model_path = "model_path";
+ static const std::string model_name = "model_name";
}
enum vector_distance_type_t {
@@ -77,7 +77,7 @@ struct field {
size_t num_dim;
std::vector<std::string> create_from;
- std::string model_path;
+ std::string model_name;
vector_distance_type_t vec_dist;
static constexpr int VAL_UNKNOWN = 2;
@@ -88,9 +88,9 @@ struct field {
field(const std::string &name, const std::string &type, const bool facet, const bool optional = false,
bool index = true, std::string locale = "", int sort = -1, int infix = -1, bool nested = false,
- int nested_array = 0, size_t num_dim = 0, vector_distance_type_t vec_dist = cosine, std::string reference = "", const std::vector<std::string> &create_from = {}, const std::string& model_path = "") :
+ int nested_array = 0, size_t num_dim = 0, vector_distance_type_t vec_dist = cosine, std::string reference = "", const std::vector<std::string> &create_from = {}, const std::string& model_name = "") :
name(name), type(type), facet(facet), optional(optional), index(index), locale(locale),
- nested(nested), nested_array(nested_array), num_dim(num_dim), vec_dist(vec_dist), reference(reference), create_from(create_from), model_path(model_path) {
+ nested(nested), nested_array(nested_array), num_dim(num_dim), vec_dist(vec_dist), reference(reference), create_from(create_from), model_name(model_name) {
set_computed_defaults(sort, infix);
}
@@ -320,8 +320,8 @@ struct field {
}
if(!field.create_from.empty()) {
field_val[fields::create_from] = field.create_from;
- if(!field.model_path.empty()) {
- field_val[fields::model_path] = field.model_path;
+ if(!field.model_name.empty()) {
+ field_val[fields::model_name] = field.model_name;
}
}
fields_json.push_back(field_val);
diff --git a/include/text_embedder.h b/include/text_embedder.h
index d65b7345..c1cb33fd 100644
--- a/include/text_embedder.h
+++ b/include/text_embedder.h
@@ -24,6 +24,5 @@ class TextEmbedder {
encoded_input_t Encode(const std::string& text);
BertTokenizer* tokenizer_;
static std::vector<float> mean_pooling(const std::vector<std::vector<float>>& input);
- static std::string get_absolute_model_path(const std::string& model_path);
std::string output_tensor_name;
};
diff --git a/include/text_embedder_manager.h b/include/text_embedder_manager.h
index 4dd033e5..d0c3ea0d 100644
--- a/include/text_embedder_manager.h
+++ b/include/text_embedder_manager.h
@@ -52,10 +52,32 @@ public:
}
static constexpr char* DEFAULT_MODEL_URL = "https://huggingface.co/ozanarmagan/e5-small-onnx/resolve/main/model.onnx";
- static constexpr char* DEFAULT_MODEL_NAME = "model.onnx";
+ static constexpr char* DEFAULT_MODEL_NAME = "model";
static constexpr char* DEFAULT_VOCAB_URL = "https://huggingface.co/ozanarmagan/e5-small-onnx/resolve/main/vocab.txt";
static constexpr char* DEFAULT_VOCAB_NAME = "vocab.txt";
inline static std::string model_dir = "";
+ inline static const std::string get_absolute_model_path(const std::string& model_name) {
+ if(model_dir.back() != '/') {
+ if(model_name.front() != '/') {
+ return model_dir + "/" + model_name + ".onnx";
+ } else {
+ return model_dir + model_name + ".onnx";
+ }
+ } else {
+ if(model_name.front() != '/') {
+ return model_dir + model_name + ".onnx";
+ } else {
+ return model_dir + "/" + model_name + ".onnx";
+ }
+ }
+ };
+ inline static const std::string get_absolute_vocab_path() {
+ if(model_dir.back() != '/') {
+ return model_dir + "/" + TextEmbedderManager::DEFAULT_VOCAB_NAME;
+ } else {
+ return model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME;
+ }
+ }
private:
TextEmbedderManager() = default;
std::unordered_map<std::string, TextEmbedder*> text_embedders;
diff --git a/src/collection.cpp b/src/collection.cpp
index f111eb34..54dc6b87 100644
--- a/src/collection.cpp
+++ b/src/collection.cpp
@@ -234,8 +234,8 @@ nlohmann::json Collection::get_summary_json() const {
field_json[fields::create_from] = coll_field.create_from;
}
- if(coll_field.model_path.size() > 0) {
- field_json[fields::model_path] = coll_field.model_path;
+ if(coll_field.model_name.size() > 0) {
+ field_json[fields::model_name] = coll_field.model_name;
}
if(coll_field.num_dim > 0) {
@@ -1163,7 +1163,7 @@ Option<nlohmann::json> Collection::search(std::string raw_query,
}
TextEmbedderManager& embedder_manager = TextEmbedderManager::get_instance();
- auto embedder = embedder_manager.get_text_embedder(search_field.model_path.size() > 0 ? search_field.model_path : TextEmbedderManager::DEFAULT_MODEL_NAME);
+ auto embedder = embedder_manager.get_text_embedder(search_field.model_name.size() > 0 ? search_field.model_name : TextEmbedderManager::DEFAULT_MODEL_NAME);
std::vector<float> embedding = embedder->Embed(raw_query);
vector_query._reset();
@@ -4654,7 +4654,7 @@ Option<bool> Collection::embed_fields(nlohmann::json& document) {
}
TextEmbedderManager& embedder_manager = TextEmbedderManager::get_instance();
- auto embedder = embedder_manager.get_text_embedder(field.model_path.size() > 0 ? field.model_path : "model.onnx");
+ auto embedder = embedder_manager.get_text_embedder(field.model_name.size() > 0 ? field.model_name : TextEmbedderManager::DEFAULT_MODEL_NAME);
std::vector<float> embedding = embedder->Embed(text_to_embed);
document[field.name] = embedding;
}
diff --git a/src/collection_manager.cpp b/src/collection_manager.cpp
index 59ee80ff..30721aa0 100644
--- a/src/collection_manager.cpp
+++ b/src/collection_manager.cpp
@@ -62,8 +62,8 @@ Collection* CollectionManager::init_collection(const nlohmann::json & collection
field_obj[fields::create_from] = std::vector<std::string>();
}
- if(field_obj.count(fields::model_path) == 0) {
- field_obj[fields::model_path] = "";
+ if(field_obj.count(fields::model_name) == 0) {
+ field_obj[fields::model_name] = "";
}
vector_distance_type_t vec_dist_type = vector_distance_type_t::cosine;
@@ -79,7 +79,7 @@ Collection* CollectionManager::init_collection(const nlohmann::json & collection
field_obj[fields::optional], field_obj[fields::index], field_obj[fields::locale],
-1, field_obj[fields::infix], field_obj[fields::nested], field_obj[fields::nested_array],
field_obj[fields::num_dim], vec_dist_type, field_obj[fields::reference], field_obj[fields::create_from],
- field_obj[fields::model_path]);
+ field_obj[fields::model_name]);
// value of `sort` depends on field type
if(field_obj.count(fields::sort) == 0) {
diff --git a/src/field.cpp b/src/field.cpp
index 2971e257..0e4b5193 100644
--- a/src/field.cpp
+++ b/src/field.cpp
@@ -663,26 +663,26 @@ Option<bool> field::json_field_to_field(bool enable_nested_fields, nlohmann::jso
}
}
- if(!field_json[fields::model_path].is_null() && field_json[fields::create_from].is_null()) {
- return Option<bool>(400, "Property `" + fields::model_path + "` can only be used with `" + fields::create_from + "`.");
+ if(field_json.count(fields::model_name) > 0 && field_json.count(fields::create_from) == 0) {
+ return Option<bool>(400, "Property `" + fields::model_name + "` can only be used with `" + fields::create_from + "`.");
}
if(field_json.count(fields::create_from) != 0) {
// If the model path is not specified, use the default model and set the number of dimensions to 384 (number of dimensions of the default model)
field_json[fields::num_dim] = static_cast<unsigned int>(384);
- if(!field_json[fields::model_path].is_null()) {
+ if(field_json.count(fields::model_name) != 0) {
unsigned int num_dim = 0;
- if(!field_json[fields::model_path].is_string()) {
- return Option<bool>(400, "Property `" + fields::model_path + "` must be a string.");
+ if(!field_json[fields::model_name].is_string()) {
+ return Option<bool>(400, "Property `" + fields::model_name + "` must be a string.");
}
- if(field_json[fields::model_path].get<std::string>().empty()) {
- return Option<bool>(400, "Property `" + fields::model_path + "` must be a non-empty string.");
+ if(field_json[fields::model_name].get<std::string>().empty()) {
+ return Option<bool>(400, "Property `" + fields::model_name + "` must be a non-empty string.");
}
- if(TextEmbedder::is_model_valid(field_json[fields::model_path].get<std::string>(), num_dim)) {
+ if(TextEmbedder::is_model_valid(field_json[fields::model_name].get<std::string>(), num_dim)) {
field_json[fields::num_dim] = num_dim;
} else {
- return Option<bool>(400, "Property `" + fields::model_path + "` must be a valid model path.");
+ return Option<bool>(400, "Property `" + fields::model_name + "` must be a valid model path.");
}
}
} else {
@@ -690,8 +690,8 @@ Option<bool> field::json_field_to_field(bool enable_nested_fields, nlohmann::jso
}
- if(field_json[fields::model_path].is_null()) {
- field_json[fields::model_path] = "";
+ if(field_json.count(fields::model_name) == 0) {
+ field_json[fields::model_name] = "";
}
@@ -776,7 +776,7 @@ Option<bool> field::json_field_to_field(bool enable_nested_fields, nlohmann::jso
field_json[fields::sort], field_json[fields::infix], field_json[fields::nested],
field_json[fields::nested_array], field_json[fields::num_dim], vec_dist,
field_json[fields::reference], field_json[fields::create_from].get<std::vector<std::string>>(),
- field_json[fields::model_path])
+ field_json[fields::model_name])
);
if (!field_json[fields::reference].get<std::string>().empty()) {
diff --git a/src/index.cpp b/src/index.cpp
index ce5fb65c..d0d0f4ec 100644
--- a/src/index.cpp
+++ b/src/index.cpp
@@ -3011,34 +3011,47 @@ Option<bool> Index::search(std::vector<query_tokens_t>& field_query_tokens, cons
curated_ids_sorted, all_result_ids, all_result_ids_len, groups_processed);
if(!vector_query.field_name.empty()) {
- VectorFilterFunctor filterFunctor(seq_ids->uncompress(), seq_ids->num_ids());
- auto& field_vector_index = vector_index.at(vector_query.field_name);
- std::vector<std::pair<float, size_t>> dist_labels;
- auto k = std::max<size_t>(vector_query.k, per_page * page);
-
- if(field_vector_index->distance_type == cosine) {
- std::vector<float> normalized_q(vector_query.values.size());
- hnsw_index_t::normalize_vector(vector_query.values, normalized_q);
- dist_labels = field_vector_index->vecdex->searchKnnCloserFirst(normalized_q.data(), k, filterFunctor);
- } else {
- dist_labels = field_vector_index->vecdex->searchKnnCloserFirst(vector_query.values.data(), k, filterFunctor);
+ // check at least one of sort fields is text match
+ bool has_text_match = false;
+ for(auto& sort_field : sort_fields_std) {
+ if(sort_field.name == sort_field_const::text_match) {
+ has_text_match = true;
+ break;
+ }
}
- for (const auto& dist_label : dist_labels) {
- uint32 seq_id = dist_label.second;
+ if(has_text_match) {
+ auto seq_ids_arr = seq_ids->uncompress();
+ VectorFilterFunctor filterFunctor(seq_ids_arr, seq_ids->num_ids());
+ auto& field_vector_index = vector_index.at(vector_query.field_name);
+ std::vector<std::pair<float, size_t>> dist_labels;
+ auto k = std::max<size_t>(vector_query.k, per_page * page);
- auto vec_dist_score = (field_vector_index->distance_type == cosine) ? std::abs(dist_label.first) :
- dist_label.first;
+ if(field_vector_index->distance_type == cosine) {
+ std::vector<float> normalized_q(vector_query.values.size());
+ hnsw_index_t::normalize_vector(vector_query.values, normalized_q);
+ dist_labels = field_vector_index->vecdex->searchKnnCloserFirst(normalized_q.data(), k, filterFunctor);
+ } else {
+ dist_labels = field_vector_index->vecdex->searchKnnCloserFirst(vector_query.values.data(), k, filterFunctor);
+ }
- auto score = (1.0 - vec_dist_score) * 100000000000.0;
-
- auto found = topster->kv_map.find(seq_id);
-
- if (found != topster->kv_map.end()) {
- found->second->scores[0] += score;
+ for (const auto& dist_label : dist_labels) {
+ uint32 seq_id = dist_label.second;
+
+ auto vec_dist_score = (field_vector_index->distance_type == cosine) ? std::abs(dist_label.first) :
+ dist_label.first;
+
+ auto score = (1.0 - vec_dist_score) * 100000000000.0;
+
+ auto found = topster->kv_map.find(seq_id);
+
+ if (found != topster->kv_map.end() && found->second->match_score_index >= 0 && found->second->match_score_index <= 2) {
+ found->second->scores[found->second->match_score_index] += score;
+ }
}
- }
+ delete [] seq_ids_arr;
+ }
}
/*auto timeMillis0 = std::chrono::duration_cast<std::chrono::milliseconds>(
diff --git a/src/text_embedder.cpp b/src/text_embedder.cpp
index 976da0f5..92b3ebfb 100644
--- a/src/text_embedder.cpp
+++ b/src/text_embedder.cpp
@@ -10,7 +10,7 @@ TextEmbedder::TextEmbedder(const std::string& model_path) {
// create environment
Ort::SessionOptions session_options;
- std::string abs_path = get_absolute_model_path(model_path);
+ std::string abs_path = TextEmbedderManager::get_absolute_model_path(model_path);
LOG(INFO) << "Loading model from: " << abs_path;
session_ = new Ort::Session(env_, abs_path.c_str(), session_options);
std::ifstream stream("vocab.txt");
@@ -21,11 +21,9 @@ TextEmbedder::TextEmbedder(const std::string& model_path) {
ustring("[CLS]"), ustring("[MASK]"), true, true, ustring("##"),512, std::string("longest_first"));
auto output_tensor_count = session_->GetOutputCount();
- bool found_output_tensor = false;
- for (int i = 0; i < output_tensor_count; i++) {
+ for (size_t i = 0; i < output_tensor_count; i++) {
auto shape = session_->GetOutputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
if (shape.size() == 3 && shape[0] == -1 && shape[1] == -1 && shape[2] > 0) {
- found_output_tensor = true;
Ort::AllocatorWithDefaultOptions allocator;
output_tensor_name = std::string(session_->GetOutputNameAllocated(i, allocator).get());
break;
@@ -57,8 +55,6 @@ std::vector<float> TextEmbedder::mean_pooling(const std::vector<std::vector<floa
}
std::vector<float> TextEmbedder::Embed(const std::string& text) {
-
- LOG(INFO) << "Embedding text: " << text;
auto encoded_input = Encode(text);
// create input tensor object from data values
Ort::AllocatorWithDefaultOptions allocator;
@@ -109,7 +105,7 @@ bool TextEmbedder::is_model_valid(const std::string& model_path, unsigned int& n
LOG(INFO) << "Loading model: " << model_path;
Ort::SessionOptions session_options;
Ort::Env env;
- std::string abs_path = get_absolute_model_path(model_path);
+ std::string abs_path = TextEmbedderManager::get_absolute_model_path(model_path);
if(!std::filesystem::exists(abs_path)) {
LOG(ERROR) << "Model file not found: " << abs_path;
@@ -142,7 +138,7 @@ bool TextEmbedder::is_model_valid(const std::string& model_path, unsigned int& n
auto output_tensor_count = session.GetOutputCount();
bool found_output_tensor = false;
- for (int i = 0; i < output_tensor_count; i++) {
+ for (size_t i = 0; i < output_tensor_count; i++) {
auto shape = session.GetOutputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
if (shape.size() == 3 && shape[0] == -1 && shape[1] == -1 && shape[2] > 0) {
num_dims = shape[2];
@@ -158,22 +154,3 @@ bool TextEmbedder::is_model_valid(const std::string& model_path, unsigned int& n
return true;
}
-
-std::string TextEmbedder::get_absolute_model_path(const std::string & model_path) {
- const std::string model_dir = TextEmbedderManager::model_dir;
-
- if(model_dir.back() != '/') {
- if(model_path.front() != '/') {
- return model_dir + "/" + model_path;
- } else {
- return model_dir + model_path;
- }
- } else {
- if(model_path.front() != '/') {
- return model_dir + model_path;
- } else {
- return model_dir + "/" + model_path;
- }
- }
-}
-
diff --git a/src/typesense_server_utils.cpp b/src/typesense_server_utils.cpp
index 44565b61..d924515e 100644
--- a/src/typesense_server_utils.cpp
+++ b/src/typesense_server_utils.cpp
@@ -459,12 +459,12 @@ int run_server(const Config & config, const std::string & version, void (*master
TextEmbedderManager::get_instance().set_model_dir(config.get_model_dir());
LOG(INFO) << "Downloading default model and vocab";
- long res = httpClient.download_file(TextEmbedderManager::DEFAULT_MODEL_URL, config.get_model_dir() + "/" + TextEmbedderManager::DEFAULT_MODEL_NAME);
+ long res = httpClient.download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
if(res != 200) {
LOG(INFO) << "Failed to download default model: " << res;
}
- res = httpClient.download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, config.get_model_dir() + "/" + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ res = httpClient.download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::get_absolute_vocab_path());
if(res != 200) {
LOG(INFO) << "Failed to download default vocab: " << res;
}
diff --git a/test/collection_all_fields_test.cpp b/test/collection_all_fields_test.cpp
index 43f8525d..8dc3a257 100644
--- a/test/collection_all_fields_test.cpp
+++ b/test/collection_all_fields_test.cpp
@@ -1650,7 +1650,7 @@ TEST_F(CollectionAllFieldsTest, ModelPathWithoutCreateFrom) {
nlohmann::json field_json;
field_json["name"] = "embedding";
field_json["type"] = "float[]";
- field_json["model_path"] = "model.onnx";
+ field_json["model_name"] = "model";
std::vector<field> fields;
std::string fallback_field_type;
@@ -1659,15 +1659,15 @@ TEST_F(CollectionAllFieldsTest, ModelPathWithoutCreateFrom) {
auto field_op = field::json_fields_to_fields(false, arr, fallback_field_type, fields);
ASSERT_FALSE(field_op.ok());
- ASSERT_EQ("Property `model_path` can only be used with `create_from`.", field_op.error());
+ ASSERT_EQ("Property `model_name` can only be used with `create_from`.", field_op.error());
}
TEST_F(CollectionAllFieldsTest, CreateFromBasicValid) {
TextEmbedderManager::model_dir = "./models/";
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_MODEL_NAME);
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::get_absolute_vocab_path());
field embedding = field("embedding", field_types::FLOAT_ARRAY, false);
embedding.create_from.push_back("name");
diff --git a/test/collection_test.cpp b/test/collection_test.cpp
index 9ca811cf..98ebe805 100644
--- a/test/collection_test.cpp
+++ b/test/collection_test.cpp
@@ -4622,8 +4622,8 @@ TEST_F(CollectionTest, SemanticSearchTest) {
})"_json;
TextEmbedderManager::model_dir = "./models/";
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_MODEL_NAME);
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::get_absolute_vocab_path());
auto op = collectionManager.create_collection(schema);
ASSERT_TRUE(op.ok());
@@ -4661,8 +4661,8 @@ TEST_F(CollectionTest, InvalidSemanticSearch) {
})"_json;
TextEmbedderManager::model_dir = "./models/";
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_MODEL_NAME);
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::get_absolute_vocab_path());
auto op = collectionManager.create_collection(schema);
LOG(INFO) << "op.error(): " << op.error();
@@ -4696,8 +4696,8 @@ TEST_F(CollectionTest, HybridSearch) {
})"_json;
TextEmbedderManager::model_dir = "./models/";
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_MODEL_NAME);
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::get_absolute_vocab_path());
auto op = collectionManager.create_collection(schema);
LOG(INFO) << collectionManager.get_collection("objects")->get_summary_json();
@@ -4737,8 +4737,8 @@ TEST_F(CollectionTest, EmbedFielsTest) {
})"_json;
TextEmbedderManager::model_dir = "./models/";
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_MODEL_NAME);
- HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::DEFAULT_VOCAB_NAME);
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_MODEL_URL, TextEmbedderManager::get_absolute_model_path(TextEmbedderManager::DEFAULT_MODEL_NAME));
+ HttpClient::get_instance().download_file(TextEmbedderManager::DEFAULT_VOCAB_URL, TextEmbedderManager::model_dir + TextEmbedderManager::get_absolute_vocab_path());
auto op = collectionManager.create_collection(schema);
ASSERT_TRUE(op.ok());