Ability to retry with multiple LLM providers if call fails

[aavetis]

https://twitter.com/DennisHXu/status/1637499076496461827?s=20

Instead of passing a single llm to chains and agents, imagine passing llms as a list of LLMs providers? Thinking about how to also include a config that says how many times to retry with a given LLM before moving on to the next one.

e.g.

from langchain.llms import AzureOpenAI
from langchain.llms import OpenAI

azLLM = AzureOpenAI(deployment_name="text-davinci-003-prod", model_name="text-davinci-003")
oaLLM = OpenAI(...)

llms = [ azLLM, oaLLM ]

# just thinking out loud with a retry_config ?
llm_chain = LLMChain(prompt=prompt, llms=llms, retry_config={'retries_each': 2, 'max_retries': 5 ...}

[cktang88]

Does the retry of OutputParser work for this?