feat: Add support for Gemini internal tools (Google Search, Code Execution, URL Context)

- Add parameters to LLM class for Gemini internal tools configuration
- Support google_search_retrieval, enable_code_execution, dynamic_retrieval_config
- Handle both simplified boolean flags and detailed tool_config format
- Add comprehensive documentation and examples
- Maintain full backward compatibility

Fixes #867

Co-authored-by: Mervin Praison <MervinPraison@users.noreply.github.com>

Author: claude[bot]

examples/python/models/googlegemini/gemini-internal-tools.py

@@ -0,0 +1,75 @@
+"""
+Example demonstrating Gemini's internal tools:
+- Google Search Grounding for real-time information
+- Code Execution for computational tasks
+- URL Context for web content analysis
+"""
+
+from praisonaiagents import Agent
+
+def main():
+    # Example 1: Google Search for current information
+    print("=== Example 1: Google Search Grounding ===")
+    search_agent = Agent(
+        instructions="You are a helpful assistant that provides current information.",
+        llm={
+            "model": "gemini/gemini-1.5-flash",
+            "google_search_retrieval": True
+        }
+    )
+    
+    response = search_agent.start("What are the latest developments in quantum computing in 2024?")
+    print(f"Response: {response}\n")
+    
+    # Example 2: Code Execution for data analysis
+    print("=== Example 2: Code Execution ===")
+    code_agent = Agent(
+        instructions="You are a data analyst that can write and execute Python code.",
+        llm={
+            "model": "gemini/gemini-1.5-flash",
+            "enable_code_execution": True
+        }
+    )
+    
+    response = code_agent.start("""
+    Create a Python script that:
+    1. Generates a list of the first 20 Fibonacci numbers
+    2. Calculates their sum and average
+    3. Creates a simple ASCII bar chart showing their growth
+    """)
+    print(f"Response: {response}\n")
+    
+    # Example 3: Comprehensive research with all tools
+    print("=== Example 3: Combined Internal Tools ===")
+    research_agent = Agent(
+        instructions="""You are a comprehensive research assistant that can:
+        - Search the web for current information
+        - Analyze web page content
+        - Write and execute code for data processing""",
+        llm={
+            "model": "gemini/gemini-1.5-pro-latest",
+            "temperature": 0.7,
+            "tool_config": {
+                "google_search_retrieval": {
+                    "threshold": 0.8  # Higher threshold for more accurate results
+                },
+                "code_execution": {},
+                "dynamic_retrieval_config": {
+                    "mode": "grounded",
+                    "dynamic_threshold": 0.6
+                }
+            }
+        },
+        verbose=True  # Show detailed execution
+    )
+    
+    response = research_agent.start("""
+    Please help me understand the current state of renewable energy:
+    1. Search for the latest statistics on global renewable energy adoption
+    2. If you find any relevant URLs, analyze their content
+    3. Write Python code to visualize the key statistics you found
+    """)
+    print(f"Response: {response}")
+
+if __name__ == "__main__":
+    main()

examples/python/models/googlegemini/gemini-tools-configuration.py

@@ -0,0 +1,131 @@
+"""
+Different ways to configure Gemini internal tools in PraisonAI
+"""
+
+from praisonaiagents import Agent
+from praisonaiagents.llm import LLM
+
+# Method 1: Simple boolean flags
+print("=== Method 1: Simple Boolean Flags ===")
+agent1 = Agent(
+    instructions="Assistant with search and code capabilities",
+    llm={
+        "model": "gemini/gemini-1.5-flash",
+        "google_search_retrieval": True,
+        "enable_code_execution": True
+    }
+)
+print("✓ Agent created with boolean flags\n")
+
+# Method 2: Detailed configuration with thresholds
+print("=== Method 2: Detailed Configuration ===")
+agent2 = Agent(
+    instructions="Research assistant with fine-tuned settings",
+    llm={
+        "model": "gemini/gemini-1.5-pro",
+        "google_search_retrieval": {
+            "threshold": 0.9  # High confidence threshold
+        },
+        "dynamic_retrieval_config": {
+            "mode": "grounded",
+            "dynamic_threshold": 0.7
+        }
+    }
+)
+print("✓ Agent created with detailed configuration\n")
+
+# Method 3: Using tool_config parameter
+print("=== Method 3: Tool Config Parameter ===")
+agent3 = Agent(
+    instructions="Advanced assistant with all tools",
+    llm={
+        "model": "gemini/gemini-1.5-pro-latest",
+        "tool_config": {
+            "google_search_retrieval": {
+                "threshold": 0.8
+            },
+            "code_execution": {},
+            "dynamic_retrieval_config": {
+                "mode": "grounded",
+                "dynamic_threshold": 0.6
+            }
+        }
+    }
+)
+print("✓ Agent created with tool_config\n")
+
+# Method 4: Direct LLM instance with internal tools
+print("=== Method 4: Direct LLM Instance ===")
+llm = LLM(
+    model="gemini/gemini-1.5-flash",
+    google_search_retrieval=True,
+    enable_code_execution=True,
+    temperature=0.7,
+    max_tokens=2000
+)
+
+# Use the LLM instance with an agent
+agent4 = Agent(
+    instructions="Assistant using pre-configured LLM",
+    llm=llm  # Pass the LLM instance directly
+)
+print("✓ Agent created with pre-configured LLM instance\n")
+
+# Method 5: Combining with custom tools
+print("=== Method 5: Internal + Custom Tools ===")
+
+def calculate_compound_interest(principal: float, rate: float, time: int) -> float:
+    """Calculate compound interest
+    
+    Args:
+        principal: Initial amount
+        rate: Annual interest rate (as decimal)
+        time: Time period in years
+    """
+    return principal * (1 + rate) ** time
+
+agent5 = Agent(
+    instructions="Financial assistant with search and custom calculation tools",
+    llm={
+        "model": "gemini/gemini-1.5-flash",
+        "google_search_retrieval": True,  # For market research
+        "enable_code_execution": True     # For complex calculations
+    },
+    tools=[calculate_compound_interest]  # Custom tool
+)
+print("✓ Agent created with both internal and custom tools\n")
+
+# Method 6: Environment-specific configuration
+print("=== Method 6: Environment-Specific Config ===")
+import os
+
+# You can set defaults via environment
+config = {
+    "model": os.getenv("GEMINI_MODEL", "gemini/gemini-1.5-flash"),
+    "temperature": float(os.getenv("GEMINI_TEMPERATURE", "0.7"))
+}
+
+# Add internal tools based on environment
+if os.getenv("ENABLE_SEARCH", "true").lower() == "true":
+    config["google_search_retrieval"] = True
+
+if os.getenv("ENABLE_CODE_EXEC", "true").lower() == "true":
+    config["enable_code_execution"] = True
+
+agent6 = Agent(
+    instructions="Environment-configured assistant",
+    llm=config
+)
+print("✓ Agent created with environment-based configuration\n")
+
+# Example usage demonstrating the tools
+print("\n=== Example Usage ===")
+print("Testing search capability...")
+response = agent1.start("What's the current temperature in Tokyo?")
+print(f"Search result preview: {response[:100]}...\n")
+
+print("Testing code execution...")
+response = agent1.start("Write and run a Python function to check if 2024 is a leap year")
+print(f"Code execution result preview: {response[:100]}...\n")
+
+print("Configuration examples complete!")