feat(docs): add cultural agents tutorial for multi-agent systems

[anderson-ufrj]

Summary

This PR adds a comprehensive tutorial demonstrating how to build multi-agent systems with distinct cultural personalities using DSPy. The tutorial is based on Cidadao.AI, a Brazilian government transparency platform that uses 22 AI agents with unique cultural identities.

Live Demo: https://cidadao-ai-frontend.vercel.app/

Tutorial Contents (15 Parts)

1. Agent Architecture Design - Patterns for multi-agent systems with cultural personalities
2. Meet the Agents - 5 example agents with images (Zumbi, Anita, Drummond, Tiradentes, Santos-Dumont)
3. Setup & Configuration - Multiple LLM providers (OpenAI, Anthropic, Ollama, Maritaca AI)
4. Personality Architecture - Enum + Dataclass + Registry patterns
5. DSPy Signatures - AgentChat, InvestigationAnalysis, ReportGeneration, IntentRouter
6. Custom Modules - ChainOfThought for natural, reasoned responses
7. Service Layer - Unified API with fallback handling
8. Testing Agents - Demonstrating personality-driven responses
9. Specialized Investigation - Fraud detection use case
10. Report Generation - Citizen-friendly reports
11. Intelligent Routing - Automatic agent selection based on intent
12. Multi-Agent Collaboration - Chaining agents (Investigate → Analyze → Report)
13. Streaming Responses - Real-time response generation
14. Optimization - Using DSPy optimizers with custom metrics
15. Save/Load - Production deployment patterns

Why This Tutorial?

• Unique use case: Cultural personalities make agents more engaging and relatable
• Production-ready patterns: Based on a real deployed system with 22 agents
• Comprehensive coverage: From basic signatures to multi-agent collaboration
• International appeal: Demonstrates how to localize AI for different cultures

Files Added

• docs/docs/tutorials/cultural_agents/index.ipynb - Main tutorial notebook
• docs/docs/tutorials/cultural_agents/*.webp - Agent images (5 files)
• Updated docs/docs/tutorials/index.md - Added tutorial to Real-World Examples

Test Plan

• Verify notebook renders correctly in Jupyter
• Test code cells execute without errors
• Verify images display in the notebook
• Check links to external resources work