isa-model 0.4.3__py3-none-any.whl → 0.4.5__py3-none-any.whl

isa_model/core/config.py

@@ -54,7 +54,7 @@ class LocalGPUGlobalConfig:
     enable_local_gpu: bool = True
     auto_detect_gpu: bool = True
     workspace_dir: str = "./local_deployments"
-    preferred_backend: str = "transformers"  # vllm, tensorrt_llm, transformers
+    preferred_backend: str = "api"  # cloud api only
     
     # Default resource settings
     default_gpu_memory_fraction: float = 0.9
@@ -114,7 +114,7 @@ class GlobalConfig:
     local_gpu_memory_fraction: float = 0.9
     local_workspace_dir: str = "./local_deployments"
     auto_detect_gpu: bool = True
-    preferred_local_backend: str = "transformers"  # vllm, tensorrt_llm, transformers
+    preferred_local_backend: str = "api"  # cloud api only
     
     # Local service defaults
     local_health_check_interval: int = 30  # seconds
@@ -263,7 +263,7 @@ class ConfigManager:
             "enable_local_gpu": os.getenv("ISA_ENABLE_LOCAL_GPU", "true").lower() == "true",
             "auto_detect_gpu": os.getenv("ISA_AUTO_DETECT_GPU", "true").lower() == "true",
             "workspace_dir": os.getenv("ISA_LOCAL_WORKSPACE_DIR", "./local_deployments"),
-            "preferred_backend": os.getenv("ISA_PREFERRED_LOCAL_BACKEND", "transformers"),
+            "preferred_backend": os.getenv("ISA_PREFERRED_LOCAL_BACKEND", "api"),
             "default_gpu_memory_fraction": float(os.getenv("ISA_GPU_MEMORY_FRACTION", "0.9")),
             "health_check_interval": int(os.getenv("ISA_LOCAL_HEALTH_CHECK_INTERVAL", "30")),
             "max_concurrent_services": int(os.getenv("ISA_MAX_CONCURRENT_SERVICES", "3")),

isa_model/core/logging/__init__.py

@@ -3,21 +3,24 @@ Logging module for ISA Model
 
 Provides comprehensive logging capabilities including:
 - Loki-based centralized application logging (via loki_logger)
-- InfluxDB-based inference metrics logging (via influx_logger)
-- Real-time monitoring and alerting
+- Basic inference logging with request tracking
 
 Architecture:
 - Loki: General application logs (INFO, WARNING, ERROR, DEBUG)
-- InfluxDB: Inference metrics (tokens, costs, performance data)
+- Basic logging: Simple request tracking and basic metrics
 """
 
-# InfluxDB inference metrics logging
-from .influx_logger import (
-    InfluxInferenceLogger,
-    InferenceLogEntry,
-    get_inference_logger,
-    generate_request_id
-)
+# Basic inference logging (no external dependencies)
+import uuid
+import logging
+
+def generate_request_id():
+    """Generate unique request ID for tracking"""
+    return str(uuid.uuid4())
+
+def get_inference_logger():
+    """Get basic inference logger"""
+    return logging.getLogger("isa_model.inference")
 
 # Loki centralized application logging
 from .loki_logger import (
@@ -34,9 +37,7 @@ from .loki_logger import (
 )
 
 __all__ = [
-    # InfluxDB inference logging
-    'InfluxInferenceLogger',
-    'InferenceLogEntry',
+    # Basic inference logging
     'get_inference_logger',
     'generate_request_id',
 

isa_model/core/models/model_manager.py

@@ -2,8 +2,6 @@ from typing import Dict, Optional, List, Any
 import logging
 from pathlib import Path
 from datetime import datetime
-from huggingface_hub import hf_hub_download, snapshot_download
-from huggingface_hub.errors import HfHubHTTPError
 from .model_storage import ModelStorage, LocalModelStorage
 from .model_repo import ModelRegistry, ModelType, ModelCapability
 from .model_billing_tracker import ModelBillingTracker, ModelOperationType
@@ -102,73 +100,7 @@ class ModelManager:
             logger.warning(f"Failed to find cheapest model for {provider}: {e}")
             return None
 
-    async def get_model(self, 
-                       model_id: str, 
-                       repo_id: str,
-                       model_type: ModelType,
-                       capabilities: List[ModelCapability],
-                       revision: Optional[str] = None,
-                       force_download: bool = False) -> Optional[Path]:
-        """
-        Get model files, downloading if necessary
-        
-        Args:
-            model_id: Unique identifier for the model
-            repo_id: Hugging Face repository ID
-            model_type: Type of model (LLM, embedding, etc.)
-            capabilities: List of model capabilities
-            revision: Specific model version/tag
-            force_download: Force re-download even if cached
-            
-        Returns:
-            Path to the model files or None if failed
-        """
-        # Check if model is already downloaded
-        if not force_download:
-            model_path = await self.storage.load_model(model_id)
-            if model_path:
-                logger.info(f"Using cached model {model_id}")
-                return model_path
-        
-        try:
-            # Download model files
-            logger.info(f"Downloading model {model_id} from {repo_id}")
-            model_dir = Path(f"./models/temp/{model_id}")
-            model_dir.mkdir(parents=True, exist_ok=True)
-
-            snapshot_download(
-                repo_id=repo_id,
-                revision=revision,
-                local_dir=model_dir,
-                local_dir_use_symlinks=False
-            )
-
-            # Save model and metadata
-            metadata = {
-                "repo_id": repo_id,
-                "revision": revision,
-                "downloaded_at": str(Path(model_dir).stat().st_mtime)
-            }
-
-            # Register model
-            self.registry.register_model(
-                model_id=model_id,
-                model_type=model_type,
-                capabilities=capabilities,
-                metadata=metadata
-            )
-
-            # Save model files
-            await self.storage.save_model(model_id, str(model_dir), metadata)
-
-            return await self.storage.load_model(model_id)
-
-        except HfHubHTTPError as e:
-            logger.error(f"Failed to download model {model_id}: {e}")
-            return None
-        except Exception as e:
-            logger.error(f"Unexpected error downloading model {model_id}: {e}")
-            return None
+    # Local model download functionality removed - use cloud API services only
     
     async def list_models(self) -> List[Dict[str, Any]]:
         """List all downloaded models with their metadata"""

isa_model/core/models/model_storage.py

@@ -39,11 +39,13 @@ class ModelStorage(ABC):
 class LocalModelStorage(ModelStorage):
     """Local file system based model storage"""
     
-    def __init__(self, base_dir: str = "./models"):
+    def __init__(self, base_dir: str = "./models", auto_create: bool = False):
         self.base_dir = Path(base_dir)
         self.models_dir = self.base_dir / "models"
         self.metadata_file = self.base_dir / "model_metadata.json"
-        self._ensure_directories()
+        self.auto_create = auto_create
+        if auto_create:
+            self._ensure_directories()
         self._load_metadata()
     
     def _ensure_directories(self):

{isa_model-0.4.3.dist-info → isa_model-0.4.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: isa_model
-Version: 0.4.3
+Version: 0.4.5
 Summary: Unified AI model serving framework
 Author: isA_Model Contributors
 Classifier: Development Status :: 3 - Alpha
@@ -89,6 +89,11 @@ Requires-Dist: isa-model[cloud,k8s,monitoring,storage]; extra == "production"
 Provides-Extra: staging
 Requires-Dist: isa-model[cloud,langchain,monitoring,storage]; extra == "staging"
 Requires-Dist: python-consul>=1.1.0; extra == "staging"
+Provides-Extra: staging-minimal
+Requires-Dist: isa-model[cloud,langchain,storage]; extra == "staging-minimal"
+Requires-Dist: influxdb-client>=1.36.0; extra == "staging-minimal"
+Requires-Dist: python-logging-loki>=0.3.1; extra == "staging-minimal"
+Requires-Dist: python-consul>=1.1.0; extra == "staging-minimal"
 Provides-Extra: all
 Requires-Dist: isa-model[audio,cloud,gpu-cloud,k8s,langchain,local,monitoring,storage,training,vision]; extra == "all"
 

{isa_model-0.4.3.dist-info → isa_model-0.4.5.dist-info}/RECORD

@@ -1,6 +1,6 @@
 isa_model/__init__.py,sha256=lYYKstKw33oavW6xS0-9cpsdYq-h0cfV_ZlGAwICRaU,868
 isa_model/client.py,sha256=7OUA6yi0G94I8U4GOFcoXvCjWxk2ds1-3CDCivJXA6M,67628
-isa_model/core/config.py,sha256=9OL8_EkBcnAH-RgyWUi3jblKo42m7K1JDeHa9C5CPL4,19519
+isa_model/core/config.py,sha256=EWadZRo3LY2gZzL81MnqK-39lUDhaFtLiVE4aoEdXZI,19456
 isa_model/core/dependencies.py,sha256=2ZgGDjtYitBEVy8H3UppQSb_BId3--f2kQw-Lm4Umh8,10050
 isa_model/core/pricing_manager.py,sha256=NWQLhNIzUDqS5_jBfVcJGrdOdRasFyifSNCliaIDvqU,17122
 isa_model/core/types.py,sha256=jdO_q0FDuzvWURXZtxMV1Zj1XgARX9kopTviWFuq_FU,8713
@@ -14,39 +14,28 @@ isa_model/core/database/migrations.py,sha256=RpM1eqt7Chu736K0ij_42gKMYHg2HPR_qSt
 isa_model/core/database/supabase_client.py,sha256=waY0VQLy9VM6FMIoViDe5yAgNW78qJaI3Jb2ohTNj98,11124
 isa_model/core/discovery/__init__.py,sha256=U7YzSNqsyPYmT_TdMlLy9QMAotdfaFlDE-c9XgKk8-4,380
 isa_model/core/discovery/consul_discovery.py,sha256=UziTVGVbejh0zUvayhISkiRJ5KeFcxUkw8t4gTK0UVw,6631
-isa_model/core/logging/__init__.py,sha256=9PA5MJvn7y73aCUUqCgv5r2it2nNb38YXSG31oxNIDA,1213
-isa_model/core/logging/influx_logger.py,sha256=TsPnWeKgV6pl2cz9NoqM0OV9z9jI1_uVbn6nTblfM94,19483
+isa_model/core/logging/__init__.py,sha256=pYQtm4GhZTEODS3tWA8EGXYhJpqK-JoZyqSPIdAANoY,1237
 isa_model/core/logging/loki_logger.py,sha256=pFkKFK5ouKNa9dPehC_kV6af0urnAvzfvbYE5IFiwD0,4706
 isa_model/core/models/__init__.py,sha256=bDzyE0KHIJxxoN3q08pvW_hHBeHux2aMdeKY4GlDzmU,1286
 isa_model/core/models/config_models.py,sha256=Gy0oNndDsKVQ92Bz6aZp1JMqr4Z0jdoEJCCscIrpogI,22955
 isa_model/core/models/deployment_billing_tracker.py,sha256=hW2e3vUkASvCycYhBdrZcsEtBjYo3TLTVpfwfVkaXnQ,16406
 isa_model/core/models/model_billing_tracker.py,sha256=er35dsoKAGt8bjkQwO9f3MQ6U_NI6OIuhIn4PEOPEWU,17302
-isa_model/core/models/model_manager.py,sha256=vjFYWxmhRGUDlrHkwySjN5d0hfCx-VqBKiMhgTXjs8Y,17093
+isa_model/core/models/model_manager.py,sha256=N_Y3kRRfV_8MPIkfNz1aKcSW_j6tDvTJ2rXRPDrd31U,14611
 isa_model/core/models/model_metadata.py,sha256=C6ubW12qmXAuqKmE_2BaO4HMB7WJqBkYDVXYCINsc-4,25148
 isa_model/core/models/model_repo.py,sha256=1018Qi8fMfgXtU8DKfWvf-0hBE16Q7wJIzmEZa7x6bw,19868
 isa_model/core/models/model_statistics_tracker.py,sha256=4KoKawwtEDAx8FV9ysmZS4nvRqZAgRSSIa-32f_Jhwk,10561
-isa_model/core/models/model_storage.py,sha256=gpW7R_wDQh0WUo4CYkrQen9GMKn8Z8ys5iGQenaMmCM,4473
+isa_model/core/models/model_storage.py,sha256=nfzesUgImbdHqTL82fTxsAeFCIGMtG1y-WIVXIQznMY,4567
 isa_model/core/models/model_version_manager.py,sha256=20BwNbCg1NlcmHmCxK_zMvpPmVFHg0B6ZCFnPLY6Yj8,37563
 isa_model/core/models/system_models.py,sha256=I52nTi0UVft8tkJdb2LZrJ_Qxax-JE00_YKqnSa-P4E,32525
 isa_model/core/repositories/__init__.py,sha256=RRERY7mWZxhSAZa4m6493l6sFl3CPlyL2bW6qJMEzD8,172
 isa_model/core/repositories/config_repository.py,sha256=QlL22r_bGEV6mHfmztEIY5Zw3wIFoiR5IQJyIj36wXU,37428
 isa_model/core/resilience/circuit_breaker.py,sha256=Ccoh3O31xVFJO2A0flnc9SI-sRqQ3sGKbwv3WbgJxBc,12435
-isa_model/core/security/secrets.py,sha256=kzRjpSiGwY9z47NUlurK29uY_uMsA5lqk8_6Ywu8Zvw,13319
 isa_model/core/services/__init__.py,sha256=TEE58Vk8JKIaQx8ELeAaWo-WPz0hjck9x-ZK7pbfiIE,422
 isa_model/core/services/intelligent_model_selector.py,sha256=PPUWiMcV8DkCPMHhnIlsBgksUY8hKB4SjlFDW1zhLYY,29205
-isa_model/core/storage/hf_storage.py,sha256=k271Rg5G7qUJAJ6VXQBTUHGU6y2NYBNWKVeBJm02DRo,14736
 isa_model/core/storage/local_storage.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 isa_model/deployment/__init__.py,sha256=Y3IUEOriJYVZ-3ZEamMs4n6_X0OwtD2eguwBas5zgtg,345
 isa_model/deployment/core/__init__.py,sha256=TRJ4tNjNFub_ObhZy61iZZpqC0CYsnv1HV1Qp_XWhVI,119
 isa_model/deployment/core/deployment_manager.py,sha256=In2e5EuXwVJfg6ENjigOrgQyy19DCWX5uuwSa--Czzc,58337
-isa_model/deployment/local/__init__.py,sha256=Ld1QbaDHIHnbW2IkSXVTZeDcxnmUXBa074uOSLRu5t0,904
-isa_model/deployment/local/config.py,sha256=Kft5EORBcsO2HVizD0ct6VFIIs9sVBN-CjRnOrcm00g,9246
-isa_model/deployment/local/gpu_gateway.py,sha256=be6d9eSWRufXvJH9GyklBWlXhOukITY1lnXTM6RPcQs,21954
-isa_model/deployment/local/health_checker.py,sha256=_u2vwiwEGCbFA6laUu1JX6GfE6prrOSHiy6PclimGys,16392
-isa_model/deployment/local/provider.py,sha256=6E1WfTUvI32aeEQs13TIyuy9xQmZqJeaJULfE2KLe4E,22763
-isa_model/deployment/local/tensorrt_service.py,sha256=f05BkJMw2NhiMp18xW1RwRED4bIjZ0gmUS5OgEAGnk0,23026
-isa_model/deployment/local/transformers_service.py,sha256=pdC3KppUzSVrWd-CKA8fXPC1uzy45S8FTtQj9odAWpM,23937
-isa_model/deployment/local/vllm_service.py,sha256=zVuBopgzG6ulSvHnRE8h_dLQQpNqTDwHbXo88IKXrwk,18849
 isa_model/deployment/modal/__init__.py,sha256=ZBcToGNtL6ztWY5pvqM1YMiL_F-S1xx9b-uZd8cuajc,380
 isa_model/deployment/modal/config.py,sha256=8XhBMIbx6oDTf-P9ipQ58xmBYDbNZekZ4gixorBDIpw,4267
 isa_model/deployment/modal/deployer.py,sha256=YNCtbO8FTVstz8OG6Kh8p3AM05dtbg5i73-JsuNy4KM,31961
@@ -98,7 +87,6 @@ isa_model/inference/repositories/__init__.py,sha256=SYTQX1E5L6zTuo_p_KnDjYefoCKw
 isa_model/inference/repositories/inference_repository.py,sha256=QnfSzkcLQ5CPcABTmSYBRAv_5SVk0ayjVW6B1Q0SKaQ,31718
 isa_model/inference/services/__init__.py,sha256=yfLz0YGl8ixk6LfTRL6cRTvZMb9F_Pv1QRgGyNc9xYM,386
 isa_model/inference/services/base_service.py,sha256=NJIvq7YpGw55ah-axDR2hcu40B2gm6L_WYXyfX0rSaE,5816
-isa_model/inference/services/custom_model_manager.py,sha256=HUHSDOWArJYMfdvaI-gfCJkVRVFdftScOw7BgS-h3zo,10829
 isa_model/inference/services/audio/__init__.py,sha256=Hgtk3j5H4U3YxNlfG8UaU2eUNOWgrpSA8LN_tKEFWMk,616
 isa_model/inference/services/audio/base_realtime_service.py,sha256=hSP89_hnzLBnmBvFOQlU_tW8UT2QKWKVR9Z7fwsVPa8,8125
 isa_model/inference/services/audio/base_stt_service.py,sha256=qahYTLpf8gruvhEtS5bWDXPiYbgxXF3nYnqTq3Ckc0E,13361
@@ -128,7 +116,6 @@ isa_model/inference/services/llm/__init__.py,sha256=aiNdB692nopCy8QaopVDB0e-J7mg
 isa_model/inference/services/llm/base_llm_service.py,sha256=CUw24rMytVXAUcberTzogKRLwYZiYifyg-kcNfHiHkg,36402
 isa_model/inference/services/llm/cerebras_llm_service.py,sha256=8BU9I7HHO481nn7ddsiP4nl2ItYTCQJzJyaIArKA0ZA,25126
 isa_model/inference/services/llm/huggingface_llm_service.py,sha256=mWnOGh3OsRyaL002Ax71Mb7oXp254VDDdP0QiQ-p9Yk,22733
-isa_model/inference/services/llm/local_llm_service.py,sha256=_ILRD-oKcolf972aXe3zPS_tBu8SD-xH_Iw29alpkHM,27606
 isa_model/inference/services/llm/ollama_llm_service.py,sha256=78VNSspzlQrXDqAxUR52jLGIKnBw4e_4LT2unAFMiTk,17967
 isa_model/inference/services/llm/openai_llm_service.py,sha256=BpYugS2Vsrc-SS69cnW2VqFv4JXMbgglXvvbNgUZNZY,43874
 isa_model/inference/services/llm/yyds_llm_service.py,sha256=ZHl2ukcDVkwYahF4OV5etTvJKa9Ni6O1TkJp75pQWaA,12495
@@ -139,7 +126,6 @@ isa_model/inference/services/ml/base_ml_service.py,sha256=mLBA6ENowa3KVzNqHyhWxf
 isa_model/inference/services/ml/sklearn_ml_service.py,sha256=Lf9JrwvI25lca7JBbjB_e66eAUtXFbwxZ3Hs13dVGkA,5512
 isa_model/inference/services/vision/__init__.py,sha256=1GO2NoC7p8IJ92mI6fGcPaN4MeFzLhdNdNlAnFYpzpE,1839
 isa_model/inference/services/vision/base_vision_service.py,sha256=mjrfcUT01HBi0k1qeIL3CkpkvQIuL_jar-N03W8sMV8,10531
-isa_model/inference/services/vision/blip_vision_service.py,sha256=tmGCznQ9qBsidLV_mnKEtvpSUPvBUbwJdwviKYXrrkA,12020
 isa_model/inference/services/vision/isa_vision_service.py,sha256=OPuIZmG_lYOgajGfrQj4uLzVk5Y4H0PkeSNViIiL1O0,22960
 isa_model/inference/services/vision/openai_vision_service.py,sha256=LeD910WWyJd6QiJncSM3x_Whj-a32Vr1_2FG4gfjtc4,10179
 isa_model/inference/services/vision/replicate_vision_service.py,sha256=smRkSCTwk5mvyKVnvyplqPNuVYjRZngVBWxTCbFmrxA,20679
@@ -187,7 +173,7 @@ isa_model/serving/api/schemas/__init__.py,sha256=Tu_hzxoKW1ZHpww3-5ER4A2hNuDByZ0
 isa_model/serving/api/schemas/common.py,sha256=HVaAS7wlvqrwC1gMZ2Cvo0vzHB053x2uOTAwUoY2vsE,696
 isa_model/serving/api/schemas/ui_analysis.py,sha256=IpOcIvmUeXN1UtZsbGozMfV1vvz7AVF2PVXjjxYl_0k,4089
 isa_model/utils/gpu_utils.py,sha256=HbMvJzSsOCcjOJluUrszAJ58dC8LPnyA_nQn9s_1I6c,11730
-isa_model-0.4.3.dist-info/METADATA,sha256=2xHn4pAvo10QKBDQK34v0-acBWGBUZRR6ZUZa-xq7BU,15090
-isa_model-0.4.3.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-isa_model-0.4.3.dist-info/top_level.txt,sha256=eHSy_Xb3kNkh2kK11mi1mZh0Wz91AQ5b8k2KFYO-rE8,10
-isa_model-0.4.3.dist-info/RECORD,,
+isa_model-0.4.5.dist-info/METADATA,sha256=wepLxrhF_7CMPHcQRRXTvSkm8tkIv5fn997d5zl-c-8,15401
+isa_model-0.4.5.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+isa_model-0.4.5.dist-info/top_level.txt,sha256=eHSy_Xb3kNkh2kK11mi1mZh0Wz91AQ5b8k2KFYO-rE8,10
+isa_model-0.4.5.dist-info/RECORD,,