Add ReliableGPT to gpt-engineer - Handle OpenAI Errors

[ishaan-jaff]

Add https://github.com/BerriAI/reliableGPT to gpt-engineer to handle the following errors:

• OpenAI APIError, OpenAI Timeout, OpenAI Rate Limit Errors, OpenAI Service UnavailableError / Overloaded
• Context Window Errors
• Invalid API Key errors

gpt-engineer users can specify fallback strategies for error handling - eg. if GPT-4 is overloaded use GPT 3.5 Turbo, Anthropic etc

cc @patillacode @AntonOsika

[patillacode]

My two cents:

Devil's advocate: Are we sure we want to add this dependency? The idea is good, I like it, just not sure this is for everyone that might use GPT Engineer.

The email thing alone calls for each user's email (reminiscence of the .env file convo 😅) or a fake one, not in love with that part.

The keys also might be a bit confusing for some users.

[AntonOsika]

Great idea but with suggested config I don't see the problem we're solving.

If it would indeed have a specified retry strategy (which I think openai package has?) then it would be different thing

Closing for now!

[AntonOsika]

Happy if you reopen when problem is more clearly stated 🙏

There are many issues that would get a lof to praise if they got solved!