Initial LlamaIndex Agent Setup with Multi-LLM

Set up a LlamaIndex ReActAgent. Configure it to use both Claude 4 Sonnet for general text understanding and Gemini 3 Flash for numerical reasoning and complex logical evaluations. Define a custom tool, `load_financial_report(file_path)`, that simulates loading and parsing a financial report from a local file. Ensure your setup demonstrates how the agent would choose between LLMs based on the task. ```python
from llama_index.core.agent import ReActAgent
from llama_index.llms.anthropic import Anthropic # For Claude 4 Sonnet
from llama_index.llms.gemini import Gemini # For Gemini 3 Flash
from llama_index.core.tools import FunctionTool # Initialize LLMs
claude_sonnet = Anthropic(model="claude-4-sonnet")
gemini_flash = Gemini(model="gemini-3-flash") # Define a simple function tool
def load_financial_report(file_path: str) -> str: """Loads and returns content from a simulated financial report file.""" # Simulate reading from a file path if file_path == "./reports/Q2_2024_Financials.txt": return "Revenue: 10M, Expenses: 8M, Profit: 2M. Anomaly detected: Expenses 20% higher than Q1 without explanation." return "File not found." financial_tool = FunctionTool.from_defaults(fn=load_financial_report) # Your task: Initialize ReActAgent with the appropriate LLM strategy and the financial_tool.
# How would you configure LlamaIndex to use different LLMs for different parts of a reasoning chain?
```