Оновлено моделі Gemini до версії 2.0 у всіх відповідних файлах, включаючи AGENTS.md, INSTRUCTION.md, app.py, requirements.txt та інші. Змінено залежність на google-genai для інтеграції з новим клієнтом. Поліпшено обробку запитів та візуалізацію аналізу.

AGENTS.md

@@ -83,8 +83,8 @@ class GeminiConnector:
     def __init__(self, api_key=None):
         self.client = self._init_gemini_client(api_key)
         self.models = {
-            'analysis': 'gemini-1.5-pro',
-            'summary': 'gemini-1.5-flash',
+            'analysis': 'gemini-2.0-pro-exp',
+            'summary': 'gemini-2.0-flash-exp',
             'vision': 'gemini-1.5-pro-vision'
         }
     
@@ -376,7 +376,7 @@ def create_analysis_visualization(analysis_results):
 ```txt
 gradio>=4.0.0
 markitdown[all]>=0.1.0
-google-generativeai>=0.3.0
+google-genai>=0.1.0
 plotly>=5.0.0
 python-multipart>=0.0.6
 aiofiles>=22.0.0
@@ -531,4 +531,4 @@ class HFResourceManager:
 - ✅ **HF Optimization**: Resource-aware processing
 - ✅ **Progressive Enhancement**: Core features first, advanced features additive
 
-This revised architecture prioritizes **deployment simplicity** while maintaining **functional richness** - perfect for HF Spaces environment with Gemini integration.
+This revised architecture prioritizes **deployment simplicity** while maintaining **functional richness** - perfect for HF Spaces environment with Gemini integration.

INSTRUCTION.md

@@ -527,7 +527,7 @@ PROCESSING_TIMEOUT_SECONDS=300
 ENABLE_DEBUG_LOGGING=false
 
 # AI Integration  
-GEMINI_DEFAULT_MODEL=gemini-1.5-pro
+GEMINI_DEFAULT_MODEL=gemini-2.0-pro-exp
 AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT=your-endpoint
 
 # Performance Tuning
@@ -590,4 +590,4 @@ A: Так, через Advanced Analytics tab можна обробляти кі
 
 **Версія документа**: 2.0.0 | **Остання редакція**: Вересень 2025
 
-*Це керівництво відображає current state платформи та буде оновлюватися з новими features та improvements.*
+*Це керівництво відображає current state платформи та буде оновлюватися з новими features та improvements.*

README.md

@@ -80,7 +80,7 @@ A comprehensive testing platform for Microsoft's MarkItDown document conversion
 ```python
 gradio>=4.0.0                    # UI framework
 markitdown[all]>=0.1.0         # Document conversion
-google-generativeai>=0.3.0       # Gemini integration  
+google-genai>=0.1.0             # Gemini integration (new client)  
 plotly>=5.17.0                   # Interactive visualizations
 pandas>=1.5.0                    # Data processing
 ```
@@ -258,4 +258,4 @@ A: Scores are based on structural analysis and AI evaluation. Use as guidelines
 
 **Built with ❤️ for enterprise document processing**
 
-*Last updated: September 2025*
+*Last updated: September 2025*

app.py

@@ -70,7 +70,7 @@ class ProcessingRequest:
     file_metadata: JSONDict
     gemini_api_key: Optional[str] = None
     analysis_type: str = "quality_analysis"
-    model_preference: str = "gemini-1.5-pro"
+    model_preference: str = GeminiModel.PRO.value
     enable_plugins: bool = False
     azure_endpoint: Optional[str] = None
     session_context: JSONDict = field(default_factory=dict)
@@ -316,7 +316,7 @@ class DocumentProcessingOrchestrator:
             analysis_request = AnalysisRequest(
                 content=conversion_result.content,
                 analysis_type=AnalysisType(request.analysis_type),
-                model=GeminiModel(request.model_preference)
+                model=GeminiModel.from_str(request.model_preference)
             )
             
             analysis_result = await engine.analyze_content(analysis_request)
@@ -645,10 +645,13 @@ class MarkItDownTestingApp:
                     
                     model_preference = gr.Dropdown(
                         choices=[
-                            ("Gemini 1.5 Pro (Best Quality)", "gemini-1.5-pro"),
-                            ("Gemini 1.5 Flash (Faster)", "gemini-1.5-flash")
+                            ("Gemini 2.0 Pro (Advanced Reasoning)", GeminiModel.PRO.value),
+                            ("Gemini 2.0 Flash (Fast Inference)", GeminiModel.FLASH.value),
+                            ("Gemini 2.5 Flash (Enhanced Quality)", GeminiModel.FLASH_25.value),
+                            ("Gemini 1.5 Pro (Legacy)", GeminiModel.LEGACY_PRO.value),
+                            ("Gemini 1.5 Flash (Legacy)", GeminiModel.LEGACY_FLASH.value)
                         ],
-                        value="gemini-1.5-pro",
+                        value=GeminiModel.PRO.value,
                         label="AI Model Preference"
                     )
                 

core/modules.py

@@ -26,7 +26,6 @@ from contextlib import asynccontextmanager
 
 import aiofiles
 from markitdown import MarkItDown
-import google.generativeai as genai
 from tenacity import retry, stop_after_attempt, wait_exponential
 try:
     import magic

examples/usage_examples.py

@@ -218,7 +218,11 @@ async def process_document(file_path, api_key):
                 ],
                 "ai_analysis": {
                     "enabled": True,
-                    "models": ["gemini-1.5-pro", "gemini-1.5-flash"],
+                    "models": [
+                        "gemini-2.0-pro-exp",
+                        "gemini-2.0-flash-exp",
+                        "gemini-2.5-flash"
+                    ],
                     "analysis_types": [
                         "quality_analysis",
                         "structure_review", 
@@ -1156,4 +1160,4 @@ async def main():
 
 
 if __name__ == "__main__":
-    asyncio.run(main())
+    asyncio.run(main())

llm/gemini_connector.py

@@ -19,8 +19,9 @@ from typing import Dict, Any, List, Optional, Union, AsyncGenerator
 from dataclasses import dataclass, asdict
 from enum import Enum
 
-import google.generativeai as genai
-from google.generativeai.types import HarmCategory, HarmBlockThreshold
+from google import genai
+from google.genai import types
+from google.genai.types import HarmCategory, HarmBlockThreshold
 from tenacity import retry, stop_after_attempt, wait_exponential, retry_if_exception_type
 from pydantic import BaseModel, Field, validator, JsonValue
 
@@ -40,9 +41,37 @@ class AnalysisType(Enum):
 
 class GeminiModel(Enum):
     """Available Gemini models with strategic use case mapping"""
-    PRO = "gemini-1.5-pro"              # Complex analysis, reasoning
-    FLASH = "gemini-1.5-flash"          # Fast processing, summaries
-    PRO_VISION = "gemini-1.5-pro-vision"  # Multimodal content analysis
+
+    PRO = "gemini-2.0-pro-exp"              # Latest high-accuracy reasoning model
+    FLASH = "gemini-2.0-flash-exp"          # Latest high-speed model
+    FLASH_25 = "gemini-2.5-flash"           # Enhanced quality flash model
+    LEGACY_PRO = "gemini-1.5-pro"           # Legacy compatibility
+    LEGACY_FLASH = "gemini-1.5-flash"       # Legacy compatibility
+    PRO_VISION = "gemini-1.5-pro-vision"    # Multimodal content analysis
+
+    @classmethod
+    def from_str(cls, value: Union[str, "GeminiModel", None]) -> "GeminiModel":
+        """Resolve string input to an enum member with graceful fallbacks"""
+
+        if isinstance(value, cls):
+            return value
+
+        if value in (None, ""):
+            return cls.PRO
+
+        try:
+            return cls(value)
+        except ValueError as exc:
+            legacy_aliases = {
+                "gemini-1.5-pro": cls.LEGACY_PRO,
+                "gemini-1.5-flash": cls.LEGACY_FLASH,
+                "gemini-1.5-pro-vision": cls.PRO_VISION,
+            }
+
+            if value in legacy_aliases:
+                return legacy_aliases[value]
+
+            raise ValueError(f"Unsupported Gemini model: {value}") from exc
 
 
 @dataclass
@@ -54,16 +83,28 @@ class GeminiConfig:
     temperature: float = 0.1  # Low temperature for consistent analysis
     timeout_seconds: int = 60
     max_retry_attempts: int = 3
-    safety_settings: Optional[Dict] = None
+    safety_settings: Optional[List[types.SafetySetting]] = None
     
     def __post_init__(self):
         if self.safety_settings is None:
-            self.safety_settings = {
-                HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
-                HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
-                HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
-                HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
-            }
+            self.safety_settings = [
+                types.SafetySetting(
+                    category=HarmCategory.HARM_CATEGORY_HATE_SPEECH,
+                    threshold=HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
+                ),
+                types.SafetySetting(
+                    category=HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
+                    threshold=HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
+                ),
+                types.SafetySetting(
+                    category=HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
+                    threshold=HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
+                ),
+                types.SafetySetting(
+                    category=HarmCategory.HARM_CATEGORY_HARASSMENT,
+                    threshold=HarmBlockThreshold.BLOCK_MEDIUM_AND_ABOVE,
+                ),
+            ]
 
 
 class AnalysisRequest(BaseModel):
@@ -81,6 +122,10 @@ class AnalysisRequest(BaseModel):
             raise ValueError("Content must be at least 10 characters long")
         return v
 
+    @validator('model', pre=True, always=True)
+    def validate_model(cls, value):
+        return GeminiModel.from_str(value)
+
 
 class AnalysisResponse(BaseModel):
     """Standardized analysis response structure"""
@@ -241,7 +286,7 @@ class GeminiAnalysisEngine:
         """Initialize Gemini Analysis Engine with configuration"""
         
         self.config = config
-        self.client = None
+        self.client: Optional[genai.Client] = None
         self._initialize_client()
         
         # Performance tracking
@@ -256,11 +301,16 @@ class GeminiAnalysisEngine:
             raise ValueError("Gemini API key is required")
         
         try:
-            genai.configure(api_key=self.config.api_key)
-            # Test client initialization with a simple call
-            models = genai.list_models()
-            logging.info(f"Gemini client initialized successfully. Available models: {len(list(models))}")
-            
+            self.client = genai.Client(api_key=self.config.api_key)
+
+            # Optional warm-up to validate credentials without incurring generation cost
+            try:
+                _ = next(self.client.models.list(page_size=1), None)
+            except Exception as list_error:  # pragma: no cover - defensive logging
+                logging.debug(f"Model listing skipped: {list_error}")
+
+            logging.info("Gemini client (google-genai) initialized successfully")
+
         except Exception as e:
             logging.error(f"Failed to initialize Gemini client: {e}")
             raise
@@ -287,15 +337,15 @@ class GeminiAnalysisEngine:
         try:
             # Prepare analysis prompt
             prompt = self._build_analysis_prompt(request)
-            
+
             # Select optimal model for analysis type
-            model_name = self._select_optimal_model(request.analysis_type, request.model)
-            
+            model_enum = self._select_optimal_model(request.analysis_type, request.model)
+
             # Execute analysis
-            response = await self._execute_analysis(model_name, prompt)
-            
+            response_text = await self._execute_analysis(model_enum.value, prompt)
+
             # Parse and structure response
-            analysis_content = self._parse_analysis_response(response.text, request.analysis_type)
+            analysis_content = self._parse_analysis_response(response_text, request.analysis_type)
             
             processing_time = (datetime.now() - start_time).total_seconds()
             self.total_processing_time += processing_time
@@ -303,7 +353,7 @@ class GeminiAnalysisEngine:
             return AnalysisResponse(
                 success=True,
                 analysis_type=request.analysis_type,
-                model_used=GeminiModel(model_name),
+                model_used=model_enum,
                 content=analysis_content,
                 metadata={
                     'processing_time': processing_time,
@@ -353,9 +403,9 @@ class GeminiAnalysisEngine:
         
         return f"{system_context}\n\n{main_prompt}"
     
-    def _select_optimal_model(self, analysis_type: AnalysisType, requested_model: GeminiModel) -> str:
+    def _select_optimal_model(self, analysis_type: AnalysisType, requested_model: GeminiModel) -> GeminiModel:
         """Select optimal Gemini model based on analysis requirements"""
-        
+
         # Strategic model selection based on analysis complexity
         model_recommendations = {
             AnalysisType.QUALITY_ANALYSIS: GeminiModel.PRO,      # Complex reasoning
@@ -364,38 +414,67 @@ class GeminiAnalysisEngine:
             AnalysisType.COMPARATIVE_ANALYSIS: GeminiModel.PRO, # Complex comparison
             AnalysisType.EXTRACTION_QUALITY: GeminiModel.PRO,   # Detailed quality assessment
         }
-        
-        # Use recommended model unless specifically overridden
-        recommended_model = model_recommendations.get(analysis_type, requested_model)
-        return recommended_model.value
+
+        # Respect explicit model choices outside default presets
+        default_overrides = {GeminiModel.PRO, GeminiModel.FLASH}
+        if requested_model not in default_overrides:
+            return requested_model
+
+        return model_recommendations.get(analysis_type, requested_model)
     
-    async def _execute_analysis(self, model_name: str, prompt: str):
+    async def _execute_analysis(self, model_name: str, prompt: str) -> str:
         """Execute analysis using Gemini API with timeout and error handling"""
-        
+
+        if not self.client:
+            raise RuntimeError("Gemini client is not initialized")
+
+        def _run_generation() -> str:
+            contents = [
+                types.Content(
+                    role="user",
+                    parts=[types.Part.from_text(text=prompt)],
+                )
+            ]
+
+            config_kwargs = {
+                "temperature": self.config.temperature,
+                "max_output_tokens": self.config.max_tokens,
+            }
+
+            if self.config.safety_settings:
+                config_kwargs["safety_settings"] = self.config.safety_settings
+
+            generation_config = types.GenerateContentConfig(**config_kwargs)
+
+            try:
+                stream = self.client.models.generate_content_stream(
+                    model=model_name,
+                    contents=contents,
+                    config=generation_config,
+                )
+
+                collected_chunks: List[str] = []
+                for chunk in stream:
+                    text_part = getattr(chunk, "text", None)
+                    if text_part:
+                        collected_chunks.append(text_part)
+
+                return "".join(collected_chunks)
+
+            except AttributeError:
+                response = self.client.models.generate_content(
+                    model=model_name,
+                    contents=contents,
+                    config=generation_config,
+                )
+                return getattr(response, "text", getattr(response, "output_text", ""))
+
         try:
-            model = genai.GenerativeModel(
-                model_name=model_name,
-                safety_settings=self.config.safety_settings
-            )
-            
-            # Configure generation parameters
-            generation_config = genai.GenerationConfig(
-                max_output_tokens=self.config.max_tokens,
-                temperature=self.config.temperature,
-            )
-            
-            # Execute with timeout
-            response = await asyncio.wait_for(
-                asyncio.to_thread(
-                    model.generate_content,
-                    prompt,
-                    generation_config=generation_config
-                ),
-                timeout=self.config.timeout_seconds
+            return await asyncio.wait_for(
+                asyncio.to_thread(_run_generation),
+                timeout=self.config.timeout_seconds,
             )
-            
-            return response
-            
+
         except asyncio.TimeoutError:
             raise TimeoutError(f"Gemini API request timed out after {self.config.timeout_seconds} seconds")
         except Exception as e:
@@ -666,7 +745,7 @@ class GeminiConnectionManager:
 def create_analysis_request(
     content: str,
     analysis_type: str,
-    model: str = "gemini-1.5-pro",
+    model: str = GeminiModel.PRO.value,
     custom_instructions: Optional[str] = None
 ) -> AnalysisRequest:
     """Factory function for creating analysis requests"""
@@ -674,7 +753,7 @@ def create_analysis_request(
     return AnalysisRequest(
         content=content,
         analysis_type=AnalysisType(analysis_type),
-        model=GeminiModel(model),
+        model=GeminiModel.from_str(model),
         custom_instructions=custom_instructions
     )
 

requirements.txt

@@ -6,7 +6,7 @@ gradio>=4.0.0,<5.0.0                    # UI framework - pinned major version fo
 markitdown[all]>=0.1.0                # Microsoft's document conversion engine
 
 # LLM Integration - Gemini Focus
-google-generativeai>=0.3.0,<1.0.0       # Google Gemini API client
+google-genai>=0.1.0                     # Google Gemini API client (latest)
 google-auth>=2.0.0                      # Authentication for Google services
 
 # Data Processing & Visualization
@@ -40,4 +40,4 @@ psutil>=5.9.0                           # System resource monitoring
 # Development & Testing Dependencies
 pytest>=7.0.0                           # Testing framework
 black>=23.0.0                           # Code formatting
-flake8>=6.0.0                           # Code linting
+flake8>=6.0.0                           # Code linting

utils/deployment.py

@@ -169,7 +169,7 @@ class SystemHealthChecker:
         required_packages = [
             'gradio',
             'markitdown',
-            'google-generativeai',
+            'google-genai',
             'plotly',
             'pandas',
             'numpy',
@@ -412,7 +412,7 @@ class DeploymentValidator:
                     requirements = f.read()
                 
                 # Check for essential packages
-                essential_packages = ['gradio', 'markitdown', 'google-generativeai']
+                essential_packages = ['gradio', 'markitdown', 'google-genai']
                 for package in essential_packages:
                     if package in requirements:
                         validation_results['checks'][f'req_{package}'] = True
@@ -606,4 +606,4 @@ def main():
 
 
 if __name__ == "__main__":
-    main()
+    main()