New Example: Add validation for structured outputs from LLM

[pguso]

Implement validation system to ensure LLM outputs conform to expected schemas/formats. Should include JSON schema validation, type checking, and fallback handling for malformed responses.

[LifeHashed]

@pguso Hi I would like to work on this issue. My approach would be intercept the llm response. And 1. Pass it through a validator function. 2. Pass it through the llm to create suggestion to create a feedback loop.

[pguso]

@LifeHashed you are very welcome to work on this. Please read this file before starting https://github.com/pguso/ai-agents-from-scratch/blob/main/CONTRIBUTING.md

This would have been me idea around implementing this:

1. Define schemas upfront
2. Validation pipeline:
   a) Try to parse JSON
   b) Validate against schema
   c) Type check fields
3. If validation fails:
   a) Retry with error message in prompt (1-2 times)
   b) Apply fallback strategy (default values, partial data, or error)
4. Log validation failures for monitoring