ai-agents-structured-output-pa-when-using-llama-index-s-structured-output-with-a--3e419363

THEN Wrap the structured LLM chat call in a retry loop that catches AttributeError or pydantic.ValidationError. On failure, log the original error message and retry up to a maximum number of attempts (e.g., 3). Use a short delay between retries to avoid rate limits.

from pydantic import ValidationError
from llama_index.core.llms import LLM

max_retries = 3
for attempt in range(max_retries):
    try:
        response = sllm.chat([system_prompt, user_prompt])
        break
    except (AttributeError, ValidationError) as e:
        print(f"Attempt {attempt+1} failed: {e}")
        if attempt == max_retries - 1:
            raise