default task detection based on model config

src/BUILD

@@ -363,6 +363,7 @@ ovms_cc_library(
     srcs = ["cli_parser.cpp"],
     deps = [
         "@com_github_jarro2783_cxxopts//:cxxopts",
+        "@com_github_tencent_rapidjson//:rapidjson",
         "libovms_server_settings",
         "libovms_version",
         "//src/filesystem:libovmsfilesystem",
@@ -373,6 +374,8 @@ ovms_cc_library(
         "//src/graph_export:t2s_graph_cli_parser",
         "//src/graph_export:s2t_graph_cli_parser",
         "//src/graph_export:image_generation_graph_cli_parser",
+        "//src/pull_module:curl_downloader",
+        "//src/pull_module:hf_env_vars",
     ],
     visibility = ["//visibility:public",],
 )
@@ -2158,8 +2161,6 @@ cc_binary(
     linkstatic = True,
 )
 
-
-
 cc_binary(
     name = "optimum-cli",
     srcs = [
@@ -2273,6 +2274,7 @@ cc_test(
         "test/tensor_conversion_test.cpp",
         "test/tensorinfo_test.cpp",
         "test/tensorutils_test.cpp",
+        "test/task_determine_test.cpp",
         "test/test_http_utils.hpp",
         "test/tfs_rest_parser_binary_inputs_test.cpp",
         "test/tfs_rest_parser_column_test.cpp",
@@ -2467,7 +2469,7 @@ cc_test(
         "//src:libcustom_node_image_transformation.so",
         "//src:libcustom_node_add_one.so",
         "//src:libcustom_node_horizontal_ocr.so",
-    ],
+    ] + glob(["test/models_config_json/**"]),
     deps = [
             "optimum-cli",
             "//src:ovms_lib",

src/cli_parser.hpp

@@ -16,6 +16,7 @@
 #pragma once
 
 #include <memory>
+#include <optional>
 #include <string>
 #include <utility>
 #include <variant>
@@ -38,13 +39,16 @@ class CLIParser {
     std::unique_ptr<cxxopts::Options> options;
     std::unique_ptr<cxxopts::ParseResult> result;
     std::variant<GraphCLIParser, RerankGraphCLIParser, EmbeddingsGraphCLIParser, ImageGenerationGraphCLIParser, TextToSpeechGraphCLIParser, SpeechToTextGraphCLIParser> graphOptionsParser;
+    std::optional<std::string> inferredTaskParameter;
 
 public:
     CLIParser() = default;
+    static std::string determineDefaultTaskParameter(const std::optional<std::string>& modelPath, const std::optional<std::string>& sourceModel, const std::optional<std::string>& modelRepositoryPath);
     std::variant<bool, std::pair<int, std::string>> parse(int argc, char** argv);
     void prepare(ServerSettingsImpl*, ModelsSettingsImpl*);
 
 protected:
+    std::string getEffectiveTaskParameter() const;
     void prepareServer(ServerSettingsImpl& serverSettings);
     void prepareModel(ModelsSettingsImpl& modelsSettings, HFSettingsImpl& hfSettings);
     void prepareGraph(ServerSettingsImpl& serverSettings, HFSettingsImpl& hfSettings, const std::string& modelName);

src/llm/http_llm_calculator.cc

@@ -68,13 +68,13 @@ class HttpLLMCalculator : public CalculatorBase {
 
     absl::Status Close(CalculatorContext* cc) final {
         OVMS_PROFILE_FUNCTION();
-        SPDLOG_LOGGER_DEBUG(llm_calculator_logger, "LLMCalculator [Node: {} ] Close", cc->NodeName());
+        SPDLOG_LOGGER_TRACE(llm_calculator_logger, "LLMCalculator [Node: {} ] Close", cc->NodeName());
         return absl::OkStatus();
     }
 
     absl::Status Open(CalculatorContext* cc) final {
         OVMS_PROFILE_FUNCTION();
-        SPDLOG_LOGGER_DEBUG(llm_calculator_logger, "LLMCalculator  [Node: {}] Open start", cc->NodeName());
+        SPDLOG_LOGGER_TRACE(llm_calculator_logger, "LLMCalculator  [Node: {}] Open start", cc->NodeName());
         ovms::GenAiServableMap servableMap = cc->InputSidePackets().Tag(LLM_SESSION_SIDE_PACKET_TAG).Get<ovms::GenAiServableMap>();
         auto it = servableMap.find(cc->NodeName());
         RET_CHECK(it != servableMap.end()) << "Could not find initialized LLM node named: " << cc->NodeName();
@@ -90,7 +90,7 @@ class HttpLLMCalculator : public CalculatorBase {
         if (!this->executionContextHolder) {
             this->executionContext = servable->createExecutionContext();
         }
-        SPDLOG_LOGGER_DEBUG(llm_calculator_logger, "LLMCalculator [Node: {}] Open end", cc->NodeName());
+        SPDLOG_LOGGER_TRACE(llm_calculator_logger, "LLMCalculator [Node: {}] Open end", cc->NodeName());
         return absl::OkStatus();
     }
     absl::Status handleGenerationError(CalculatorContext* cc, const char* errorMessage) {

src/pull_module/BUILD

@@ -68,6 +68,13 @@ ovms_cc_library(
     visibility = ["//visibility:public"],
 )
 
+ovms_cc_library(
+    name = "hf_env_vars",
+    hdrs = ["hf_env_vars.hpp"],
+    deps = [],
+    visibility = ["//visibility:public"],
+)
+
 ovms_cc_library(
     name = "gguf_downloader",
     srcs = ["gguf_downloader.cpp"],
@@ -109,6 +116,7 @@ ovms_cc_library(
     srcs = ["hf_pull_model_module.cpp"],
     hdrs = ["hf_pull_model_module.hpp"],
     deps =  [
+        ":hf_env_vars",
         ":curl_downloader",
         ":libgit2",
         "gguf_downloader",

src/pull_module/hf_env_vars.hpp

@@ -0,0 +1,22 @@
+//****************************************************************************
+// Copyright 2025 Intel Corporation
+//
+// 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.
+//****************************************************************************
+#pragma once
+
+namespace ovms {
+inline constexpr const char* HF_TOKEN_ENV_VAR = "HF_TOKEN";
+inline constexpr const char* HF_ENDPOINT_ENV_VAR = "HF_ENDPOINT";
+inline constexpr const char* DEFAULT_HF_ENDPOINT = "https://huggingface.co";
+}  // namespace ovms

src/pull_module/hf_pull_model_module.cpp

@@ -30,6 +30,7 @@
 #include "optimum_export.hpp"
 #include "curl_downloader.hpp"
 #include "gguf_downloader.hpp"
+#include "hf_env_vars.hpp"
 #include "../graph_export/graph_export.hpp"
 #include "../logging.hpp"
 #include "../module_names.hpp"
@@ -293,11 +294,11 @@ const std::string HfPullModelModule::GetProxy() const {
 }
 
 const std::string HfPullModelModule::GetHfToken() const {
-    return getEnvReturnOrDefaultIfNotSet("HF_TOKEN");
+    return getEnvReturnOrDefaultIfNotSet(HF_TOKEN_ENV_VAR);
 }
 
 const std::string HfPullModelModule::GetHfEndpoint() const {
-    std::string hfEndpoint = getEnvReturnOrDefaultIfNotSet("HF_ENDPOINT", "https://huggingface.co");
+    std::string hfEndpoint = getEnvReturnOrDefaultIfNotSet(HF_ENDPOINT_ENV_VAR, DEFAULT_HF_ENDPOINT);
     if (!endsWith(hfEndpoint, "/")) {
         hfEndpoint.append("/");
     }

src/test/models_config_json/Kokoro/config.json

@@ -0,0 +1,9 @@
+{
+  "return_dict": true,
+  "output_hidden_states": false,
+  "dtype": null,
+  "chunk_size_feed_forward": 0,
+  "is_encoder_decoder": false,
+  "architectures": null,
+  "n_mels": 80
+}

src/test/models_config_json/NullArch/config.json

@@ -0,0 +1,8 @@
+{
+  "return_dict": true,
+  "output_hidden_states": false,
+  "dtype": null,
+  "chunk_size_feed_forward": 0,
+  "is_encoder_decoder": false,
+  "architectures": null
+}

src/test/models_config_json/Qwen3-8B/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Qwen3ForCausalLM"],
+  "model_type": "qwen3"
+}

src/test/models_config_json/Qwen3-Embedding-0.6B/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Qwen3ForCausalLM"],
+  "model_type": "qwen3"
+}

src/test/models_config_json/Qwen3-Reranker-0.6B/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Qwen3ForCausalLM"],
+  "model_type": "qwen3"
+}

src/test/models_config_json/bge/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["BertModel"],
+  "model_type": "bert",
+  "hidden_size": 768
+}

src/test/models_config_json/bge_reranker/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["XLMRobertaForSequenceClassification"],
+  "model_type": "xlm-roberta"
+}

src/test/models_config_json/cross_encoder/config.json

@@ -0,0 +1,6 @@
+{
+  "architectures": ["BertForSequenceClassification"],
+  "model_type": "bert",
+  "hidden_size": 768,
+  "num_labels": 2
+}

src/test/models_config_json/flux/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["FluxTransformer2DModel"],
+  "model_type": "flux",
+  "hidden_size": 3072
+}

src/test/models_config_json/flux_pipeline/model_index.json

@@ -0,0 +1,5 @@
+{
+  "_class_name": "FluxPipeline",
+  "_diffusers_version": "0.34.0",
+  "_name_or_path": "black-forest-labs/FLUX.1-dev"
+}

src/test/models_config_json/gemma4/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Gemma4ForConditionalGeneration"],
+  "model_type": "gemma4"
+}

src/test/models_config_json/invalid_architecture/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": [
+    "InvalidArchitecture"
+  ]
+}

src/test/models_config_json/lfm/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Lfm2MoeForCausalLM"],
+  "model_type": "lfm2_moe"
+}

src/test/models_config_json/llama/config.json

@@ -0,0 +1,6 @@
+{
+  "architectures": ["LlamaForCausalLM"],
+  "model_type": "llama",
+  "hidden_size": 4096,
+  "num_hidden_layers": 32
+}

src/test/models_config_json/no_architectures/config.json

@@ -0,0 +1,4 @@
+{
+  "model_type": "bert",
+  "hidden_size": 768
+}

src/test/models_config_json/parlertts/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["ParlerTTSForConditionalGeneration"],
+  "model_type": "parlertts",
+  "hidden_size": 768
+}

src/test/models_config_json/phi3/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["Phi3ForCausalLM"],
+  "model_type": "phi3",
+  "hidden_size": 3072
+}

src/test/models_config_json/qwen2_embedding/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["Qwen2Model"],
+  "model_type": "qwen2",
+  "hidden_size": 1024
+}

src/test/models_config_json/qwen2_rerank/config.json

@@ -0,0 +1,6 @@
+{
+  "architectures": ["Qwen2ForSequenceClassification"],
+  "model_type": "qwen2",
+  "hidden_size": 1024,
+  "num_labels": 2
+}

src/test/models_config_json/qwen3/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["Qwen3ForCausalLM"],
+  "model_type": "qwen3",
+  "hidden_size": 1024
+}

src/test/models_config_json/qwen3_6/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": [
+    "Qwen3_5MoeForConditionalGeneration"
+  ]
+}

src/test/models_config_json/qwen3_asr/config.json

@@ -0,0 +1,6 @@
+{
+  "architectures": ["Qwen3ASRForConditionalGeneration"],
+  "model_type": "qwen3_asr",
+  "return_dict": true,
+  "output_hidden_states": false
+}

src/test/models_config_json/qwen3_multi_arch/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["Qwen3ASRForConditionalGeneration", "Qwen3ConditionalGeneration"],
+  "model_type": "qwen3_multi_arch"
+}

src/test/models_config_json/sdxl/model_index.json

@@ -0,0 +1,5 @@
+{
+  "_class_name": "StableDiffusionXLPipeline",
+  "_diffusers_version": "0.34.0",
+  "_name_or_path": "stabilityai/stable-diffusion-xl-base-1.0"
+}

src/test/models_config_json/seamlessm4t/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["SeamlessM4TModel"],
+  "model_type": "seamless_m4t",
+  "hidden_size": 512
+}

src/test/models_config_json/speecht5/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["SpeechT5ForTextToSpeech"],
+  "model_type": "speecht5",
+  "hidden_size": 256
+}

src/test/models_config_json/stable_diffusion/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["UNet2DConditionModel"],
+  "model_type": "unet",
+  "in_channels": 9
+}

src/test/models_config_json/t5_encoder/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["T5EncoderModel"],
+  "model_type": "t5",
+  "d_model": 512
+}

src/test/models_config_json/trinity/config.json

@@ -0,0 +1,4 @@
+{
+  "architectures": ["AfmoeForCausalLM"],
+  "model_type": "afmoe"
+}

src/test/models_config_json/whisper/config.json

@@ -0,0 +1,5 @@
+{
+  "architectures": ["WhisperForConditionalGeneration"],
+  "model_type": "whisper",
+  "hidden_size": 512
+}

src/test/models_config_json/xlm_roberta/config.json

@@ -0,0 +1,6 @@
+{
+  "architectures": ["XLMRobertaForSequenceClassification"],
+  "model_type": "xlm-roberta",
+  "hidden_size": 768,
+  "num_labels": 2
+}